This PDF file contains the front matter associated with SPIE Proceedings Volume 9069, including the Title Page, Copyright information, Table of Contents, Introduction, and Conference Committee listing.

Breast cancer occurs with high frequency among women. In most cases, the main early signs appear as mass and calcification. Distinguishing masses from normal tissues is still a challenging work as mass varies with shapes, margins and sizes. In this paper, a novel method for mass detection in mammograms was presented. First, morphology operators are employed to locate mass candidates. Then anisotropic diffusion was applied to make mass region display better multiple concentric layers (MCL). Finally an extended concentric morphology model (ECMM) criterion combining MCL criterion and template matching was proposed to detect masses. This method was examined on 170 images from Digital Database for Screening Mammography (DDSM) database. The detection rate is 93.92% at 1.88 false positives per image (FPs/I), demonstrating the effectiveness of the proposed method.

Facial expressions reflect internal emotional states of a character or in response to social communications. Though much effort has been taken to generate realistic facial expressions, it still remains a challenging topic due to human being’s sensitivity to subtle facial movements. In this paper, we present a method for facial animation generation, which reflects true facial muscle movements with high fidelity. An intermediate model space is introduced to transfer captured static AU peak frames based on FACS to the conformed target face. And then dynamic parameters derived using a psychophysics method is integrated to generate facial animation, which is assumed to represent natural correlation of multiple AUs. Finally, the animation sequence in the intermediate model space is mapped to the target face to produce final animation.

The ink diffusion on Xuan paper is essentially a particle diffusion process, and the diffusion result is mainly determined by paper structure and ink attributes. This paper introduces weighting fiber structure to model the Xuan paper, which regards the paper fiber as the hindrance to ink diffusion. Based on the paper model, we present a novel simulation method of ink diffusion using the diffusion equation with variable coefficient. The diffusion coefficient consists of several factors including the fiber’s weights and the ink quantity in the current diffusion location. To efficiently solve the ink diffusion equation, we also propose a new implicit difference method with high accuracy and linear time complexity. Compared with the previous similar methods, our method is able to describe the spontaneous gray evolution and generate a more natural diffusion boundary. The experimental results demonstrate that our approach can realistically simulate different diffusion effects on different kinds of Xuan paper.

In this paper, an adaptive edge enhancement algorithm is proposed to reconstruct a super resolution image from a single low resolution one. In order to improve the results of the high resolution reconstruction, edge statistics is learned from the scenes using a statistical analysis of the maximum likelihood estimation to approximate edge boosting weight that helps to significantly enhance edge information in the high frequency area. The edge sketch image will be adaptively combined with the results of wiener filter according to the values of the local variance. The experimental results on several test images show the success in reconstructing the super resolution both quantitatively and perceptually.

Image completion solves the problem of filling missing region by using the information from the same or another image. It is difficult to maintain a balance between visual plausibility and efficiency among the existing algorithms. In this paper, we first propose a novel graph-based approach combining patch offsets and structure feature to get more coherent completion result. We further put forward creatively using a few dominate offsets with an adaptive mechanism of labels and formulate image completion to be a graph-cut optimization problem. Experiments on a wide variety of images show our method yields better results in various challenging cases than state-of-art methods both on visual impact and efficiency.

Fingertips of human hand play an important role in hand-based interaction with computers. Identification of fingertips’ positions in hand images is vital for developing a human computer interaction system. This paper proposes a novel method for detecting fingertips of a hand image analyzing the concept of the geometrical structural information of fingers. The research is divided into three parts: First, hand image is segmented for detecting hand; Second, invariant features (curvature zero-crossing points) are extracted from the boundary of the hand; Third, fingertips are detected. Experimental results show that the proposed approach is promising.

The anti-counterfeiting watermarked images of lithographic prints must be transformed into digital images by scanner or camera before authenticating. The obtained images usually happen to geometric distortion so as to influence the watermark extraction. A geometric distortion correction method for the lithographic watermarked authentication images is proposed in this paper. Hough transform is used to detect the edge lines or corner points, and then unites with the nature of the oblique distortion or the perspective transformation model to correct the distortion. Finally, morphological filtering is used to revise the corrected images. Experiments results demonstrate the method can correct the geometric distortion well and the watermarking image can be extracted and clear enough to authentication.

Restoration of a degraded image from motion blurring is highly dependent on the estimation of the point spread function, which depends on the blur orientation and the blur extent. In this paper, we present a spectral analysis of image gradients, which has more strong periodic strips compare the spectrum of image. The experiments on simulated images show that our algorithm is capable of accurately identifying the blurring kernel for a wider range of blur orientation and blur extent. Furthermore, it’s been proved to be more robust to noise over available techniques.

In this paper, we design a DCT-based model for estimating spatio-temporal just noticeable distortion (JND) profiles of color image/video. Based on a mathematical model of measuring the base detection threshold for each DCT coefficient in the color component of color images, the masking adjustments for luminance component and chrominance components are utilized for estimating the spatial JND profiles. The above spatial JND profiles are extended to video signals by incorporating the proposed block-based temporal masking adjustment mainly considering local temporal statistics in luminance component. The model is verified by designing a subjective viewing test of evaluating the visual quality under the specified viewing condition. In the experiment, the test video is contaminated by the estimated JND profiles in the DCT domain and is compared with the original video. The simulation results show that the JNDcontaminated color video has nearly perceptual lossless visual quality and the model is able to estimate the JND profiles inherent in color videos.

In this work, we propose a feasible 3D video generation method to enable high quality visual perception using a monocular uncalibrated camera. Anthropometric distances between face standard landmarks are approximated based on the person's age and gender. These measurements are used in a 2-stage approach to facilitate the construction of binocular stereo images. Specifically, one view of the background is registered in initial stage of video shooting. It is followed by an automatically guided displacement of the camera toward its secondary position. At the secondary position the real-time capturing is started and the foreground (viewed person) region is extracted for each frame. After an accurate parallax estimation the extracted foreground is placed in front of the background image that was captured at the initial position. So the constructed full view of the initial position combined with the view of the secondary (current) position, form the complete binocular pairs during real-time video shooting. The subjective evaluation results present a competent depth perception quality through the proposed system.

Progressive photon mapping solves the memory limitation problem of traditional photon mapping. It gives the correct radiance with a large passes, but it converges slowly. We propose an anisotropic progressive photon mapping method to generate high quality images with a few passes. During the rendering process, different from standard progressive photon mapping, we store the photons on the surfaces. At the end of each pass, an anisotropic method is employed to compute the radiance of each eye ray based on the stored photons. Before move to a new pass, the photons in the scene are cleared. The experiments show that our method generates better results than the standard progressive photon mapping in both numerical and visual qualities.

This paper presents a simple and fast algorithm to extract the skeleton of vascular structures from segmented vessel datasets. Our algorithm is based on a step by step approach to move a small volume of interest along the vessel tree. With the introduction of Signed Distance Function (SDF), the moving sphere along the vessel tree can easily and automatically detect bifurcations and predict the location of next axis point. Some experiments have been carried out to demonstrate the strengths of our proposed method.

Since “cloud computing” was put forward by Google , it quickly became the most popular concept in IT industry and widely permeated into various areas promoted by IBM, Microsoft and other IT industry giants. In this paper the methods of bibliometric analysis were used to investigate the global cloud computing research trend based on Web of Science (WoS) database and the Engineering Index (EI) Compendex database. In this study, the publication, countries, institutes, keywords of the papers was deeply studied in methods of quantitative analysis, figures and tables are used to describe the production and the development trends of cloud computing.

Visualization and reconstruction of blood vessel from standard medical datasets play an important role in many clinical situations. This paper presents a survey on the visualization and reconstruction of vascular structures. Firstly, the visualization techniques of vasculatures are introduced, which includes volume rendering and surface rendering of vasculatures. Then, we focus on the reconstruction techniques of vascular structures, which can be classified into two categories: explicit reconstruction and implicit reconstruction of vascular structures. With reconstructed vascular geometry, it is quite easy to produce smooth visualization of vessel surfaces. In addition, finding the accurate geometric representation of vascular structures is crucial in developing computer aided vascular surgery systems.

Automated processing and quantification of biological images have been rapidly increasing the attention of researchers in image processing and pattern recognition because the roles of computerized image and pattern analyses are critical for new biological findings and drug discovery based on modern high-throughput and highcontent image screening. This paper presents a study of the automated detection of regions of mitochondria, which are a subcellular structure of eukaryotic cells, in microscopy images. The automated identification of mitochondria in intracellular space that is captured by the state-of-the-art combination of focused ion beam and scanning electron microscope imaging reported here is the first of its type. Existing methods and a proposed algorithm for texture analysis were tested with the real intracellular images. The high correction rate of detecting the locations of the mitochondria in a complex environment suggests the effectiveness of the proposed study.

This paper presents a robust adaptive embedding scheme using a modified Spatio-Temporal noticeable distortion (JND) model that is designed for tracing the distribution of the H.264/AVC video content and protecting them from unauthorized redistribution. The Embedding process is performed during coding process in selected macroblocks type Intra 4x4 within I-Frame. The method uses spread-spectrum technique in order to obtain robustness against collusion attacks and the JND model to dynamically adjust the embedding strength and control the energy of the embedded fingerprints so as to ensure their imperceptibility. Linear and non linear collusion attacks are performed to show the robustness of the proposed technique against collusion attacks while maintaining visual quality unchanged.

In order to accurately realize the defect detection of pipeline inwall, this paper proposes a measurement system made up of cross structured light, single CCD camera and a smart car, etc. Based on structured light measurement technology, this paper mainly introduces the structured light measurement system, the imaging mathematical model, and the parameters and method of camera calibration. Using these measuring principles and methods, the camera in remote control car platform achieves continuous shooting of objects and real-time rebound processing as well as utilizing established model to extract 3D point cloud coordinate to reconstruct pipeline defects, so it is possible to achieve 3D automatic measuring, and verifies the correctness and feasibility of this system. It has been found that this system has great measurement accuracy in practice.

Fluorescent lamps are becoming popular as an indoor light source in many commercial sites. While high speed video cameras can detect variations in the illumination levels from such light sources, they can be quite costly to operate. Moreover, most popular security video cameras can only operate up to a maximum speed of 30 frames per second. Would the camera be able to detect such variations in the scene caused by the inconsistent level of illumination from the fluorescent lamps? This paper examines several image quality measures to detect flickering on video signals from the video cameras at low frame rates. Future research is outlined at the end of the paper.

This paper proposes a three-dimensional(3D) segmentation algorithm using hyper-complex edge detection operator and applies the new algorithm to three-dimensional hepatic vessel segmentation from computed tomography (CT) volumetric data. A 3D hyper-complex edge detection operator is constructed by combining octonion and gradient operator. We replace every voxel of the volumetric data by one octonion which consist of its gray-level and its 6 neighborhoods' gray-level. Via this the original volumetric data is defined as octonion volumetric data. Similar to the Sobel operator, there are three principal directions (coordinate axes) in 3D hyper-complex edge detection operator, and each element in this operator is a octonion. The operator is circularly convoluted with octonion volumetric data to get the value of matching response. If matched, this voxel is the edge of vessel. Experimental results show that the algorithm can effectively segment small vascular tree branches.

It is one of useful and interesting applications to discriminate the kind of the flower or recognize the name of the flower, as example of retrieving flower database. As its contour line of the petal region of flower is useful for such problems, it is important to extract the precise region of the petal of a flower picture. In this paper, the method which extracts petal regions on a flower picture using HSV color information is proposed, such to discriminate the kind of the flower. The experiments show that the proposed method can extract petal regions at the success rate of about 90%, which is thought to be satisfied. In detail, the success rates of one-colored flower, plural-colored flower, and white flower are about 98%, 85%, and 83%, respectively.

Most general purpose no-reference image quality assessment algorithms need prior knowledge about anticipated distortions and their corresponding human opinion scores. One or more of distortion types may not be available when creating the model. In this paper, we develop a blind/no-reference opinion unaware distortion unaware image quality assessment algorithm based on natural scenes. The proposed approach extracts features in spatial domain for both natural images and distorted image at two scales, where locally normalized luminance values are modeled in two forms: pointwise for single pixels and pair-wise based log-derivative for the relationship of adjacent pixels. Then two sharpness functions are applied whose their outputs represent the extracted features of the proposed approach. Results show that the proposed algorithm correlates well with subjective opinion scores. They also show that the proposed algorithm outperforms the full-reference PSNR and SSIM methods. Not only do the results compete well with the recently developed NIQE model, but also outperform it.

In this paper, we propose a novel iterative active contour algorithm, i.e. Iterative Contextual CV Model (ICCV), and apply it to automatic liver segmentation from 3D CT images. ICCV is a learning-based method and can be divided into two stages. At the first stage, i.e. the training stage, given a set of abdominal CT training images and the corresponding manual liver labels, our task is to construct a series of self-correcting classifiers by learning a mapping between automatic segmentations (in each round) and manual reference segmentations via context features. At the second stage, i.e. the segmentation stage, first the basic CV model is used to segment the image and subsequently Contextual CV Model (CCV), which combines the image information and the current shape model, is iteratively performed to improve the segmentation result. The current shape model is obtained by inputting the previous automatic segmentation result into the corresponding self-correcting classifier. The proposed method is evaluated on the datasets of MICCAI 2007 liver segmentation challenge. The experimental results show that we obtain more and more accurate segmentation results by the iterative steps and satisfying results are obtained after about six iterations. Also, our method is comparable to the state-of-the-art work on liver segmentation.

In recent years, research on human emotion estimation using thermal infrared (IR) imagery has appealed to many researchers due to its invariance to visible illumination changes. Although infrared imagery is superior to visible imagery in its invariance to illumination changes and appearance differences, it has difficulties in handling transparent glasses in the thermal infrared spectrum. As a result, when using infrared imagery for the analysis of human facial information, the regions of eyeglasses are dark and eyes’ thermal information is not given. We propose a temperature space method to correct eyeglasses’ effect using the thermal facial information in the neighboring facial regions, and then use Principal Component Analysis (PCA), Eigen-space Method based on class-features (EMC), and PCA-EMC method to classify human emotions from the corrected thermal images. We collected the Kotani Thermal Facial Emotion (KTFE) database and performed the experiments, which show the improved accuracy rate in estimating human emotions.

Reversible watermarking is used to hide information in images for medical and military uses. Reversible watermarking in images using distortion compensation proposed by Vasily et al [5] embeds each pixel twice such that distortion caused by the first embedding is reduced or removed by the distortion introduced by the second embedding. In their paper, because it is not applied in its most basic form, it is not clear whether improving it can achieve better results than the existing state of the art techniques. In this paper we first provide a novel basic distortion compensation technique that uses same prediction method as Tian’s [2] difference expansion method (DE), in order to measure the effect of the distortion compensation more accurately. In the second part, we will analyze what kind of improvements can be made in distortion compensation.

Reversible watermarking can embed data into the cover image and extract data from stego image, where the original cover image can be recovered perfectly after the extraction of data. Difference expansion (DE) and prediction error expansion (PEE) are two popular reversible watermarking methods. DE has the advantage of small distortion while PEE has the advantage of large embedding capacity and smaller prediction error compared with pixel difference. In this paper, we proposed a novel method that combines the advantages of DE and PEE, where the difference calculated between two pixels is combined with the edge information near this pixel pair. The proposed difference calculation can produce smaller pixel difference compared with the original simple pixel difference calculation. Overlapping embedding is then used to increase the embedding capacity. Our proposed method gives excellent results which is shown by several experiments.

This paper presents a novel low-rank and sparse decomposition (LSD) based model for anomaly detection in hyperspectral images. In our model, a local image region is represented as a low-rank matrix plus spares noises in the spectral space, where the background can be explained by the low-rank matrix, and the anomalies are indicated by the sparse noises. The detection of anomalies in local image regions is formulated as a constrained LSD problem, which can be solved efficiently and robustly with a modified “Go Decomposition” (GoDec) method. To enhance the validity of this model, we adapts a “simple linear iterative clustering” (SLIC) superpixel algorithm to efficiently generate homogeneous local image regions i.e. superpixels in hyperspectral imagery, thus ensures that the background in local image regions satisfies the condition of low-rank. Experimental results on real hyperspectral data demonstrate that, compared with several known local detectors including RX detector, kernel RX detector, and SVDD detector, the proposed model can comfortably achieves better performance in satisfactory computation time.

This paper proposes a new curve smoothing method invariant to affine transformation. Curve smoothing is one of the important challenges in computer vision as a procedure for noise suppression in shape analysis such as Curvature Scale Space (CSS). Currently, Gaussian filtering is widely used among a lot of smoothing methods. However Gaussian filtering is not affine invariant. This paper proposes a new method for curve smoothing that is invariant under affine transformation such that area of any region in the image does not change. Specifically, we introduce an affine invariant evaluate function with a metric tensor. The original curve is smoothed by minimizing the evaluation function. We mathematically prove that this method is affine invariant. Further, experimental results show that the proposed method is almost never affected by affine transformation different from usual Gaussian filtering. In the proposed method, processing results are expected to be not affected much by variation of the viewpoint.

We develop an augmented reality (AR) environment with hidden-marker via touch interface using Kinect device, and then also set up a touch painting game with the AR environment. This environment is similar to that of the touch screen interface which allows user to paint picture on a tabletop with his fingers, and it is designed with depth image information from Kinect device setting up above a tabletop. We incorporate support vector machine (SVM) to classify painted pictures which correspond to the inner data and call out its AR into the tabletop in color images information from Kinect device. Because users can utilize this similar touch interface to control AR, we achieve a marker-less AR and interactive environment.

This paper presents a robust color image watermarking algorithm, which embeds a grayscale image into a color image using the higher order singular value decomposition (HOSVD). We look the color image in the RGB color space upon as a tensor rather than three independent channels. The color image is partitioned into non-overlapped patches (subtensors), and their HOSVDs are computed. Moreover, a subtle preprocessing step, block Arnold transform, is designed to improve the robustness to cropping attack. Experimental results show that the proposed algorithm makes the watermarking invisible effectively and is robust against wide variety of non-geometric and geometric attacks.

Short-term forecasting of cloud distribution within a sequence of all-sky images is an important issue in meteorological area. In this work, a cloud image forecasting system is designed, which includes three steps---cloud detection, cloud matching and motion estimation. We treat cloud detection as a classification problem based on Linear Discriminant Analysis. During the matching, a set of Speed Up Robust Features (SURF) are extracted to represent the cloud, then clouds are matched by computing correspondences between SURF features. Finally, affine transform is applied to estimate the motion of cloud. This local features based method is capable of predicting the rotation and scaling of cloud, while the traditional method is only limited to translational motion. Objective evaluation results show higher accuracy of the proposed method compared with some other algorithms.

This paper presents a Belief Propagation (BP) stereo matching algorithm using ground control points. The proposed algorithm combines local and global stereo methods, which first utilizes local stereo method to obtain an initial disparity map, then the ground control points are selected from the initial disparity and used on belief propagation algorithm for global stereo matching. Since using ground control points, the proposed algorithm improves BP algorithm in convergence speed. Moreover, this paper proposes a color constraint voting method to optimize the disparity in postprocessing. Experimental results show that the proposed algorithm shares low computational complexity but high matching accuracy.

In this paper, we propose a new algorithm for simplifying line drawing sketches. First, we segment the strokes at the points of large curvature if desired. Then, we perform a low-pass filter and use the result to assign a weight to every stroke. The strokes are moved to the position of the higher weight. After that, we find the stroke pairs and combine them to reduce the total number of the strokes, resulting in a cleaner line drawing art. This system also cuts down the disordered and confusing small strokes and combines them to form long strokes.

This paper studies a three-dimensional non-contact attitude and deformation measuring system of the wind tunnel model. By projecting the single laser speckle on the measured three-dimensional space, it receives a random optical encoding. In addition, correlation search matching for collected image is conducted by matlab, the measurement data results of the existing PIV equipment are compared, and the spin binocular measurement system is employed to obtain the three-dimensional coordinate of the model. As a result, the feasibility of the system is verified. the author grasped the principle of 3D non-contact coding and recognition.

Appearance modeling is an important and yet challenging issue for online visual tracking due to the accumulation of errors which is prone to potential drifting during the self-updating with newly obtained results. In this paper, we propose a novel online tracking algorithm using spatio-temporal cue integration. Specifically, the object is represented as a set of local patches with respect to the spatial cue. In terms of the temporal cue, we keep the appearance models at different time and do appearance updating alternately. Taking full advantage of both historical and current information of the tracked object, the drift problem is alleviated. We also develop an effective cue quality measurement that combines similarity and motion information. Both qualitative and quantitative evaluations on challenging video sequences demonstrate that the proposed algorithm performs comparable against the state-of-the-art methods.

Demosaicing of the color filter array is one of the most important parts of the image processing pipeline for single sensor digital cameras. In recent years, one of the most successful algorithms is the multiscale gradients (MSG) algorithm. In this paper, several modifications were made to the MSG algorithm such that the computational complexity is significantly reduced while maintaining image quality.

Embedded block coding with optimized truncation (EBCOT) is a key algorithm in JPEG 2000 image compression system. Recently, the bit-plane coder architectures are capable of producing symbols at a higher rate than the capability of the existing MQ arithmetic coders. To solve this problem, a design of a multiple-symbol processor for statistical MQ coder architecture on FPGA is proposed. The proposed architecture takes advantage of simplicity of single-symbol architecture while integrates several techniques in order to increase the coding rate (more than one symbol per clock), reduce critical path, thus accelerate the coding speed. The repeated symbol statistics has been analyzed prior to the proposed architecture using lookahead technique. This allows the proposed architecture to support encoding rate of maximum 8 symbols per clock cycle without stalls and without excessively increasing the hardware cost. This helps to accelerate encoding process, which leads to greatly increase throughput. From the experiments, for lossy wavelet transform, the proposed architecture offers high throughput of at least 233.07 MCxD/S with effectively reducing the number of clock cycles more than 35.51%.

In image processing and computational photography, automatic image enhancement is one of the long-range objectives. Recently the automatic image enhancement methods not only take account of the globe semantics, like correct color hue and brightness imbalances, but also the local content of the image, such as human face and sky of landscape. In this paper we describe a new scheme for automatic image enhancement that considers both global semantics and local content of image. Our automatic image enhancement method employs the multi-scale edge-aware image decomposition approach to detect the underexposure regions and enhance the detail of the salient content. The experiment results demonstrate the effectiveness of our approach compared to existing automatic enhancement methods.

Signature verification holds a significant place in today's world as most of the bank transactions, stock trading etc. are validated via signatures. Signatures are considered as one of the most effective biometric identity but unfortunately signature forgery attempts are quite rampant. To prevent this, a robust signature verification mechanism is essential. In this paper, a new method has been proposed which uses Local Binary Pattern and geometrical features. A new geometric property has been devised i.e. Octave Pattern. Performance is analyzed by comparing random, semi-skilled and skilled forgeries with the genuine signature.

In a previous study, we examined color spaces to read resistor lines of 11 different colors (black, brown, red, orange, yellow, green, blue, purple, gray, white, and gold). However, the color classification experiment was carried out only under a certain illumination. In order effectively to classify real resistor color lines, the color classification under various illumination conditions must be considered. In this paper, we examine 10 color features (RGB, XYZ, YCbCr, YIQ, HSI, HSV, HLS, L*u*v*, L*a*b*, and I1I2I3) under the various illumination. The experimental results show the effectiveness of the L*u*v* color feature. Furthermore, in a very small training sample size situation, the classification performance of the u*v* feature vector except for an intensity element L* outperforms that of L*u*v*.

Chaotic theory has been used in cryptography application for generating a sequence of data that is close to pseudorandom number based on an adjusted initial condition and a parameter. However, data recovery becomes a crucial problem due to the precision of the parameters. This difficulty leads to limited usage of Chaotic-based cryptography especially for error sensitive applications such as voice cryptography. In order to enhance the encryption security and overcome this limitation, an Adaptive Pixel-Selection using Chaotic Map Lattices (APCML) is proposed. In APCML, the encryption sequence has been adaptively selected based on chaos generator. Moreover, the chaotic transformation and normalization boundary have been revised to alleviate the rounding error and inappropriate normalization boundary problems. In the experiments, the measurement indices of originality preservation, visual inspection, and statistical analysis are used to evaluate the performance of the proposed APCML compared to that of the original CML. Consequently, the APCML algorithm offers greater performance with full recovery of the original message.

In this paper, we propose a robust and fast line segment detector, which achieves accurate results with a controlled number of false detections and requires no parameter tuning. It consists of three steps: first, we propose a novel edge point chaining method to extract Canny edge segments (i.e., contiguous chains of Canny edge points) from the input image; second, we propose a top-down scheme based on smaller eigenvalue analysis to extract line segments within each obtained edge segment; third, we employ Desolneux et al.’s method to reject false detections. Experiments demonstrate that it is very efficient and more robust than two state of the art methods—LSD and EDLines.

This paper proposes a novel face super-resolution reconstruction (hallucination) technique for YCbCr color space. The underlying idea is to learn with an error regression model and multi-linear principal component analysis (MPCA). From hallucination framework, many color face images are explained in YCbCr space. To reduce the time complexity of color face hallucination, we can be naturally described the color face imaged as tensors or multi-linear arrays. In addition, the error regression analysis is used to find the error estimation which can be obtained from the existing LR in tensor space. In learning process is from the mistakes in reconstruct face images of the training dataset by MPCA, then finding the relationship between input and error by regression analysis. In hallucinating process uses normal method by backprojection of MPCA, after that the result is corrected with the error estimation. In this contribution we show that our hallucination technique can be suitable for color face images both in RGB and YCbCr space. By using the MPCA subspace with error regression model, we can generate photorealistic color face images. Our approach is demonstrated by extensive experiments with high-quality hallucinated color faces. Comparison with existing algorithms shows the effectiveness of the proposed method.

Image matching plays an important role in a variety of applications. For matching images, the features are needed to describe the local content of images. In this paper, we proposed novel features (Vector Features) for image matching and image registration. The vector features are based on the intensity differences of the pixels around the interesting points and characterize the spatial distribution of gray values in an effective and efficient way. In the experiments, the invariance of vector features to image rotation is verified. The vector features perform well on the images of different focal lengths and the multi-view images. The experiment on image registration demonstrates a good result.

With the rapid development of information techniques, data mining approaches have become one of the most important tools to discover the in-deep associations of tuples in large-scale database. Hence how to protect the private information is quite a huge challenge, especially during the data mining procedure. In this paper, a new method is proposed for privacy protection which is based on fuzzy theory. The traditional fuzzy approach in this area will apply fuzzification to the data without considering its readability. A new style of obscured data expression is introduced to provide more details of the subsets without reducing the readability. Also we adopt a balance approach between the privacy level and utility when to achieve the suitable subgroups. An experiment is provided to show that this approach is suitable for the classification without a lower accuracy. In the future, this approach can be adapted to the data stream as the low computation complexity of the fuzzy function with a suitable modification.

Mammogram is currently the best way for early detection of breast cancer. Mass is a typical sign of breast cancer, and the classification of masses as malignant or benign may assist radiologists in reducing the biopsy rate without increasing false negatives. Typically, different geometry and texture features are extracted and utilized to train a classifier to classify a mass. However, not each feature is equally important for a classifier, and some features may indeed decrease the performance of a classifier. In this paper, we investigated the usage of semi-supervised feature selection method for classification. After a mass is extracted from a ROI (region of interest) with level set method. Morphological and texture features are extracted from the segmented regions and surrounding regions. SSLFE (Semi- Supervised Local Feature Extraction, proposed in our previous work) is utilized to select important features for KNN classifier. Mammography images from DDSM were used for experiment. The experimental result shows that by incorporating information embedded in unlabeled data, SSLFE can improve the performance compared to the method without feature selection and traditional Relief method.

Applying the structure element (SE)-based morphology to detect the edges of the image corrupted by noise may obtain the irregular shape, and hence the precision of the object identification may decrease. Lately, the multi-SE morphology was introduced to detect the edge of gray-scale image, which yields the more integral and continual final result than that of the single-structure element. For color images, the robust color morphology gradient (RCMG) edge-detection, extended from CMG, was proposed by using a square SE along with the removing outliner technique to solve the noisesensitivity problem. In this paper, in order to solve such a noise problem on color images, we propose an efficient color edge-detection algorithm, called the ECMSEM (Efficient Color Multi-SE Morphology) by combining the advantage of the multi-SE morphology and the removing (noise) outliner. In our ECMSEM approach, the color multi-SE morphology with using a synthetic weighted method is introduced to detect only edge (except noise) and using multi-scale SEs to increase the precision of each SE-direction, and hence improve the edge-detection results. In performance evaluation, the edge-detection results (of synthetic color tested images) showed that our ECMSEM yielded the higher FOM than those of the RCMG approach.

In this paper, we proposed a two-step algorithm based on the combination of the exemplar-based algorithm and the illumination model to deal with specular images, especially those contain saturated pixels in the highlight areas. First the proposed modified exemplar-based algorithm is employed to process the unsaturated specular pixels under the supervision of illumination model. Then we inpaint the rest regions in which the pixels are saturated with original exemplar-based algorithm to obtain the final result. Experimental results demonstrate that the proposed algorithm performs better on the images with saturated pixels in the highlight areas compared with classical highlight removal and image inpainting algorithms.

A real-time orientation feature descriptor for portable devices is introduced. The descriptor requires very low computational resources and has 16 dimensions shorter than all existing methods. The patch of a candidate feature is firstly segmented into polar arranged sub-regions, which enables us to achieve rotation invariance rapidly. Furthermore, the principal orientation is used to describe each sub-region. The computations can be considerably accelerated by using integral image. The descriptor is used for object tracking and achieves 25 fps frame rate on mobile phone. Experimental results demonstrate that the proposed method offers sufficient matching performance.

According to the requirement of RoboCup Standard Platform League (NAO) Rule Book, the author did some experiments to finish the League’s tasks based on the humanoid robot NAO. The image captured by a moving camera will have a distortion which is affected by the so called “rolling shutter”. The rolling shutter means that not all scanlines are exposed over the same time interval. This paper presents the model of rolling shutter of cameras and principle of capturing images. Then corrections to image distortion found on the NAO robots is described. At last the cases of reversing the effects of a rolling shutter on the images taken by the robot’s moving camera are presented. This approach improves the effectiveness of shape recognition and the accuracy of Tilt rate of field line.

Liaohe river basin mountain torrent disaster serious, this paper tells a system of flood loss assessment which would be helpful for the decrease of the geologic disastrous loss. Firstly, it mainly analyzes four uncertainties aspects of flood loss: precision of prediction and simulation, accounting standard of asset, asset vulnerability and flood prevention ability. Secondly, EasyDHM has been selected for forecast flood, the simulation forecast time is 6h before real flood peak coming; flood inundation model is selected for flood submerged level information extracted and flood submerged map. In the end, the model of flood Lose calculation is used to calculation the loss by the extracted information from stacking social economic data and water data. The feasibility of flood loss assessment system has been proved by flood simulation in 1998, 2003 and 2005. The unified assessment criteria make the assessment result difference, but this system of flood loss assessment has some value in system integration.

This paper presents a supervised smoke detection method that uses local and global features. This framework integrates and extends notions of many previous works to generate a new comprehensive method. First chrominance detection is used to screen areas that are suspected to be smoke. For these areas, local features are then extracted. The features are among homogeneity of GLCM and energy of wavelet. Then, global feature of motion of the smoke-color areas are extracted using a space-time analysis scheme. Finally these features are used to train an artificial intelligent. Here we use neural network, experiment compares importance of each feature. Hence, we can really know which features among those used by many previous works are really useful. The proposed method outperforms many of the current methods in the sense of correctness, and it does so in a reasonable computation time. It even has less limitation than conventional smoke sensors when used in open space. Best method for the experimental results is to use all the mentioned features as expected, to insure which is the best experiment result can be achieved. The achieved with high accuracy of result expected output is high value of true positive and low value of false positive. And show that our algorithm has good robustness for smoke detection.

In this study a Computer-Aided Microscopy (CAM) system is proposed for investigating the importance of the histological criteria involved in diagnosing of cancers in microscopy in order to suggest the more informative features for discriminating low from high-grade brain tumours. Four families of criteria have been examined, involving the greylevel variations (i.e. texture), the morphology (i.e. roundness), the architecture (i.e. cellularity) and the overall tumour qualities (expert’s ordinal scale). The proposed CAM system was constructed using a modified Seeded Region Growing algorithm for image segmentation, and the Probabilistic Neural Network classifier for image classification. The implementation was designed on a commercial Graphics Processing Unit card using parallel programming. The system’s performance using textural, morphological, architectural and ordinal information was 90.8%, 87.0%, 81.2% and 88.9% respectively. Results indicate that nuclei texture is the most important family of features regarding the degree of malignancy, and, thus, may guide more accurate predictions for discriminating low from high grade gliomas. Considering that nuclei texture is almost impractical to be encoded by visual observation, the need to incorporate computer-aided diagnostic tools as second opinion in daily clinical practice of diagnosing rare brain tumours may be justified.

Investigating an efficient method for plant propagation can help not only prevent extinction of plants but also facilitate the development of botanical industries. In this paper, we propose to use image processing techniques to determine the cutting-line for the propagation of two kinds of plants, i.e. Melaleuca alternifolia Cheel and Cinnamomum kanehirai Hay, which have quite different characteristics in terms of shape, structure, and propagation way (e.g. propagation by seeding and rooting, respectively). The proposed cutting line determination methods can be further applied to develop an automatic control system to reduce labor cost and increase the effectiveness of plant propagation.

In this paper, we propose a novel image retrieval algorithm using local spatial binary patterns (LSBP) for contentbased image retrieval. The traditional local binary pattern (LBP) encodes the relationship between the referenced pixel and its surrounding neighbors by calculating gray-level difference, but LBP lacks the spatial distribution information of texture direction. The proposed method encodes spatial relationship of the referenced pixel and its neighbors, based on the gray-level variation patterns of the horizontal, vertical and oblique directions. Additionally, variation between center pixel and its surrounding neighbors is calculated to reflect the magnitude information of the whole image. We compare our method with LBP, uniform LBP (ULBP), completed LBP (CLBP), local ternary pattern (LTP) and local tetra patterns (LTrP) based on three benchmark image databases including, Brodatz texture database(DB1), Corel database(DB2), and MIT VisTex database(DB3). Experiment analysis shows that the proposed method improves the retrieval results from 70.49%/41.30% to 73.26%/46.26% in terms of average precision/average recall on database DB2, from 79.02% to 85.92% and 82.14% to 90.88% in terms of average precision on databases DB1 and DB3, respectively, as compared with the traditional LBP.

Pleural thickenings can be found in asbestos exposed patient’s lung. Non-invasive diagnosis including CT imaging can detect aggressive malignant pleural mesothelioma in its early stage. In order to create a quantitative documentation of automatic detected pleural thickenings over time, the differences in volume and thickness of the detected thickenings have to be calculated. Physicians usually estimate the change of each thickening via visual comparison which provides neither quantitative nor qualitative measures. In this work, automatic spatiotemporal matching techniques of the detected pleural thickenings at two points of time based on the semi-automatic registration have been developed, implemented, and tested so that the same thickening can be compared fully automatically. As result, the application of the mapping technique using the principal components analysis turns out to be advantageous than the feature-based mapping using centroid and mean Hounsfield Units of each thickening, since the resulting sensitivity was improved to 98.46% from 42.19%, while the accuracy of feature-based mapping is only slightly higher (84.38% to 76.19%).

In this paper an approach based on particle filtering to automatic extract power lines from aerial images is presented. We integrate the similarity of grey value of power lines into particle filtering to track the points on power lines, and use those extracted points to fit the power line as a parabola. Moreover, a fully automatic initialization strategy is used. Experimental results show that the proposed approach is a promising and fully automatic method for extracting power lines from very complex background. This algorithm will play an important role in the exact 3D-reconstructions of power lines that can help the power grid company to ensure the safety of the power lines.

Edge-preserving smoothing is crucial for image decomposition to extract the base layer. However, current methods fail to smooth high-contrast details or preserve thin edges due to their single criterion for distinguishing edges and details. In this paper, we present a hybrid definition for salient edges using two properties: intensity amplitude and oscillations density. Based on this definition, we propose an edge-preserving image smoothing algorithm. Firstly local extrema of the input image are located. Then these extrema points are classified into edge or detail points by the two properties. Thirdly, max and min envelops are obtained by an optimizing process with edge points as constrains. Lastly, the smoothing result is obtained by an averaging operation. Experimental results show that the proposed method can preserve salient step edges while smoothing high-contrast details and is useful in many applications such as image enhancement and hatch-to-tone mapping.

In this paper, we present a fully automatic facial expression recognition system using support vector machines, with geometric features extracted from the tracking of facial landmarks. Facial landmark initialization and tracking is performed by using an elastic bunch graph matching algorithm. The facial expression recognition is performed based on the features extracted from the tracking of not only individual landmarks, but also pair of landmarks. The recognition accuracy on the Extended Kohn-Kanade (CK+) database shows that our proposed set of features produces better results, because it utilizes time-varying graph information, as well as the motion of individual facial landmarks.

AdaBoost is a machine learning technique which integrates many weak classifiers into one strong classifier to enhance its classification performance. Gentle AdaBoost is a variant of AdaBoost which introduces Newton steps to the boosting process. It is proved that the overall performance considering both the training error and generalization error of Gentle AdaBoost is better than other AdaBoost variants on low-noise data. However, it suffers from overfitting problem when the training data include high noise. To solve this problem, we propose a new approach to limit the weight distortion according to a stretched distribution of the whole sample weights. Experimental results have shown that our algorithm obtains a better generalization error on both standard and noise-input datasets. Moreover, our method does not increase the calculation time compared with Gentle AdaBoost.

The advances of computer games have shown their potentials for developing edutainment content and services. Current cultural heritages often make use of games in order to complement existing presentations and to create a memorable exhibition. It offers opportunities to reorganize and conceptualize historical, cultural and technological information about the exhibits. To demonstrate the benefits of serious games in terms of facilitating the learning activities in a constructive and meaningful way, we designed a video game about the Heerlen bathhouse heritage. This paper explains the design considerations of this Roman bathhouse game, with a particular focus on the link between game play and learning.

In order to improve the accuracy of geometrical defect detection, this paper presented a method based on HU moment invariants of skeleton image. This method have four steps: first of all, grayscale images of non-silicon MEMS parts are collected and converted into binary images, secondly, skeletons of binary images are extracted using medialaxis- transform method, and then HU moment invariants of skeleton images are calculated, finally, differences of HU moment invariants between measured parts and qualified parts are obtained to determine whether there are geometrical defects. To demonstrate the availability of this method, experiments were carried out between skeleton images and grayscale images, and results show that: when defects of non-silicon MEMS part are the same, HU moment invariants of skeleton images are more sensitive than that of grayscale images, and detection accuracy is higher. Therefore, this method can more accurately determine whether non-silicon MEMS parts qualified or not, and can be applied to nonsilicon MEMS part detection system.

Normal estimation is an essential step in point cloud based geometric processing, such as high quality point based rendering and surface reconstruction. In this paper, we present a clustering based method for normal estimation which preserves sharp features. For a piecewise smooth point cloud, the k-nearest neighbors of one point lie on a union of multiple subspaces. Given the PCA normals as input, we perform a subspace clustering algorithm to segment these subspaces. Normals are estimated by the points lying in the same subspace as the center point. In contrast to the previous method, we exploit the low-rankness of the input data, by seeking the lowest rank representation among all the candidates that can represent one normal as linear combinations of the others. Integration of Low-Rank Representation (LRR) makes our method robust to noise. Moreover, our method can simultaneously produce the estimated normals and the local structures which are especially useful for denoise and segmentation applications. The experimental results show that our approach successfully recovers sharp features and generates more reliable results compared with the state-of-theart.

Today, digital multimedia messages have drawn more and more attention due to the great achievement of computer and network techniques. Nevertheless, text is still the most popular media for people to communicate with others. Many fonts have been developed so that product designers can choose unique fonts to demonstrate their idea gracefully. It is commonly believed that handwritings can reflect one’s personality, emotion, feeling, education level, and so on. This is especially true in Chinese calligraphy. However, it is not easy for ordinary users to customize a font of their personal handwritings. In this study, we performed a process reengineering in font generation. We present a new method to create font in a batch mode. Rather than to create glyphs of characters one by one according to their codepoints, people create glyphs incrementally in an on-demand manner. A Java Implementation is developed to read a document image of user handwritten Chinese characters, and make a vector font of these handwritten Chinese characters. Preliminary experiment result shows that the proposed method can help ordinary users create their personal handwritten fonts easily and quickly.

In this paper, a novel method of conformal parameterization for triangular meshes is presented. Firstly, based on geodesic on a mesh, an algorithm constructing local barycentric coordinates is proposed. Then, these local coordinates are merged via a linear system to form a global conformal parameterization of the mesh. The conformal mesh parameterization method here can be viewed as a development of the shape-preserving method proposed by M. S. Floater. It avoids error of locally approximating the so-called geodesic polar mapping and hence giving better results. Experimental results are given to illustrate the effectiveness of proposed methods.

Considering that existing line segment detection algorithms may detect a long line segment as several short fragmented segments, a novel line segment linking algorithm is proposed in this paper to improve the performance of line segment detection. Since the gradient orientations of points on the Right Linking Segments (RLSs) have better consistency than those on Wrong Linking Segments (WLSs), a feature descriptor is designed for each candidate linking segment based on the gradient orientation information that can effectively distinguish RLSs from WLSs. Experiment results of the testing images show that the proposed method can greatly improve the original line segment detection results by connecting most fragmented line segments more accurately.

Strand structures such as the tails or feelers are common to man-made or natural shapes. The knowledge of strand structures which a shape possesses can be exploited in its matching, recognition, retrieval etc. Although a variety of methods in shape decomposition have been presented, there is still a need for a robust and versatile method to detect the strand structures, especially when the shapes have large deformation or noise. Based on the visibility of points, we design a shape descriptor and propose an effective method to detect the strand structures possessing in 2D shapes. An intuitive idea is that the points in strand structures can only see small number of points from their reference points inside the shape. Meanwhile, the visibility of a point is more robust than its convex-concave features. Extensive experiments have been done on shapes with various kinds of deformation and large noise, demonstrating the robustness and effectiveness of our strand structures detection method.

The brush plays an important role in creating Chinese calligraphy. We regard a single bristle of a writing brush as an elastic rod and the brush tuft absorbing ink as an elastic cone, which naturally deforms according to the force exerted on it when painting on a paper, and the brush footprint is formed by the intersection region between the deformed tuft and the paper plane. To efficiently generate brush strokes, this paper introduces interpolation and texture mapping approach between two adjacent footprints, and automatically applies bristle-splitting texture to the stroke after long-time painting. Experimental results demonstrate that our method is effective and reliable. Users can create realistic calligraphy in real time.

This paper offers an algebraic explanation for the phenomenon of a new and prosperous branch of evolutionary metaheuristics – “skeletal algorithms”. We show how this explanation gives rise to algorithms for recognition of algebraic theories and present sample applications.

This paper presents a novel algorithm for selecting random features via compressed sensing to improve the performance of Normalized Cuts in image segmentation. Normalized Cuts is a clustering algorithm that has been widely applied to segmenting images, using features such as brightness, intervening contours and Gabor filter responses. Some drawbacks of Normalized Cuts are that computation times and memory usage can be excessive, and the obtained segmentations are often poor. This paper addresses the need to improve the processing time of Normalized Cuts while improving the segmentations. A significant proportion of the time in calculating Normalized Cuts is spent computing an affinity matrix. A new algorithm has been developed that selects random features using compressed sensing techniques to reduce the computation needed for the affinity matrix. The new algorithm, when compared to the standard implementation of Normalized Cuts for segmenting images from the BSDS500, produces better segmentations in significantly less time.

The booming electronic books (e-books), as an extension to the paper book, are popular with readers. Recently, many efforts are put into the realistic page-turning simulation o f e-book to improve its reading experience. This paper presents a new 3D page-turning simulation approach, which employs piecewise time-dependent cylindrical surfaces to describe the turning page and constructs smooth transition method between time-dependent cylinders. The page-turning animation is produced by sequentially mapping the turning page into the cylinders with different radii and positions. Compared to the previous approaches, our method is able to imitate various effects efficiently and obtains more natural animation of turning page.

Distance measure of point to segment is one of the determinants which affect the efficiency of DP (Douglas-Peucker) polyline simplification algorithm. Zone-divided distance measure instead of only perpendicular distance is proposed by Dan Sunday [1] to improve the deficiency of the original DP algorithm. A new efficiency zone-divided distance measure method is proposed in this paper. Firstly, a rotating coordinate is established based on the two endpoints of curve. Secondly, the new coordinate value in the rotating coordinate is computed for each point. Finally, the new coordinate values are used to divide points into three zones and to calculate distance, Manhattan distance is adopted in zone I and III, perpendicular distance in zone II. Compared with Dan Sunday’s method, the proposed method can take full advantage of the computation result of previous point. The calculation amount basically keeps for points in zone I and III, and the calculation amount reduces significantly for points in zone II which own highest proportion. Experimental results show that the proposed distance measure method can improve the efficiency of original DP algorithm.

Flood protection of Fuhe river basin has been payed high attention after Changkai-levee crevasse in 2010. This paper constructions a model of flood disaster lose calculation considering flood disaster and social economic developing based on GIS. Firstly social economic indexes have been selected according to characteristics of the urban and the rural. Secondly a mathematical model of flood routing using Finite Volume Method has been made in spacial information grids, the data of inundated depth and flood duration can be extracted from the grids. In the end ,wo calculate the loss by flood disaster losses calculation process model. This paper solves the stacking problem of flood characteristic and administrative boundaries effectively, which makes a development on accuracy of flood disaster assessment.

This paper presents a novel edge-guided filtering scheme for decomposition-based tone mapping, whose superiority is to prevent two major defects in filter-driven multi-scale decomposition: halo artifact and over-smoothing distortion. First, we calculate an edge-preserving smoothing by gradient domain reconstruction with given edges. Then we apply this output in high dynamic range tone mapping to address aforementioned problems. At last, some experimental results are presented to demonstrate the effectiveness of our method in producing high-quality low dynamic range outputs.

Image texture analysis plays an important role in object detection and recognition in image processing. The texture analysis can be used for early detection of breast cancer by classifying the mammogram images into normal and abnormal classes. This study investigates breast cancer detection using texture features obtained from the grey level cooccurrence matrices (GLCM) of curvelet sub-band levels combined with texture feature obtained from the image itself. The GLCM were constructed for each sub-band of three curvelet decomposition levels. The obtained feature vector presented to the classifier to differentiate between normal and abnormal tissues. The proposed method is applied over 305 region of interest (ROI) cropped from MIAS dataset. The simple logistic classifier achieved 86.66% classification accuracy rate with sensitivity 76.53% and specificity 91.3%.

Accurate assessment on the quality of color images is an important step to many image processing systems that convey visual information of the reproduced images. An accurate objective image quality assessment (IQA) method is expected to give the assessment result highly agreeing with the subjective assessment. To assess the quality of color images, many approaches simply apply the metric for assessing the quality of gray scale images to each of three color channels of the color image, neglecting the correlation among three color channels. In this paper, a metric for assessing color images’ quality is proposed, in which the model of variable just-noticeable color difference (VJNCD) is employed to estimate the visibility thresholds of distortion inherent in each color pixel. With the estimated visibility thresholds of distortion, the proposed metric measures the average perceptible distortion in terms of the quantized distortion according to the perceptual error map similar to that defined by National Bureau of Standards (NBS) for converting the color difference enumerated by CIEDE2000 to the objective score of perceptual quality assessment. The perceptual error map in this case is designed for each pixel according to the visibility threshold estimated by the VJNCD model. The performance of the proposed metric is verified by assessing the test images in the LIVE database, and is compared with those of many well-know IQA metrics. Experimental results indicate that the proposed metric is an effective IQA method that can accurately predict the image quality of color images in terms of the correlation between objective scores and subjective evaluation.