The aim of this paper is to remove speckle noise and enhance the laminar bone structure in ultrasound images in order to aid the anesthesiologist during an epidural anesthesia needle insertion. In this regard, a direction and scale sensitive anisotropic diffusion (AD) algorithm called the complex wavelet diffusion algorithm is proposed. The proposed algorithm is tested on synthetic and real ultrasound images of the laminar bone and the results are compared with the speckle-reducing AD method. Results indicate that the proposed method can better highlight the diagonal-shaped laminar bones while removing speckle noise. Furthermore, the proposed despeckling algorithm is very fast providing the anesthesiologist with real-time information while an epidural injection is performed.

Image matching has been one of the most fundamental issues in computer vision over the decades. We propose a method based on utilizing feature lines in order to achieve more robust image matching, which includes feature line detection, feature vector description and matching, and the devised rotation invariant feature line transform. The feature vectors have the properties of rotation and scaling invariance. Experimental results demonstrate the effectiveness and efficiency of the proposed method. Compared with the famous powerful algorithm scale invariant feature transform, the proposed method is more insensitive to noise, and the selected distinctive locations of features are more disperse. For a certain sequence of images, which contain strong lines, the proposed method is more efficient. Using the feature lines obtained by our method, it is possible to match two scene images with different rotation angles, scales, and light distortion, and the steps of matching are simpler.

Despite the approval of a new standard in 2009, JPEG-extended range lossless, current digital products still employ previous standards for lossless image compression, such as JPEG, JPEG2000, JPEG-LS, etc. Wavelet-based codecs can provide abundant functionalities and excellent compression efficiency. Among them, the backward coding of wavelet trees (BCWT) algorithm offers lower complexity and consumes less internal buffer memory without sacrificing quality at similar compression ratios (CR) when compared to other wavelet-based codecs, such as JPEG2000 and set partitioning in hierarchical trees (SPIHT). A line-based BCWT was developed for further reduction of internal buffer memory. A very efficient line-based lossless BCWT compression algorithm is presented. Lossless color and lossless wavelet transform are employed and the original BCWT algorithm is modified for lossless operation, including incorporation of adaptive arithmetic coding. In order to eliminate coding redundancies, a set to zeros method and a zero tree detection algorithm are proposed, which significantly enhance the boundary condition CR performance while reserving the algorithm’s advantages. Tests and analysis results show that the lossless BCWT algorithm requires less memory and computational resources than SPIHT and JPEG2000, while retaining image quality comparable to the standard image codecs, therefore, lossless BCWT is quite suitable for implementation in modern digital technologies.

A hand vein recognition method using local Gabor ordinal measure (OM) is presented. Gabor OM uses eight encoding masks to extract four types of features, which are derived from the magnitude, phase, real, and imaginary components of vein image after Gabor filtering, respectively, and then concatenates these feature histograms. Block-based pattern matching introduced with a Fisher linear discriminant adopts the “divide and conquer” strategy to alleviate the effect of noise and to further enhance the discriminative power of the feature descriptor. The proposed method is evaluated on our hand vein image database and HK PolyU database. The results with an error equation rate of 0.53% and 0.06% on the two databases, respectively, demonstrate the good performance of our approach.

The quad-tree-based picture partition scheme in high-efficiency video coding (HEVC) provides significant coding efficiency improvements, particularly for high-resolution videos. However, its flexible block partitions and numerous intraprediction modes result in a substantial increase in computational complexity compared with previous video coding standards because of the necessity of determining the best coding unit (CU) partitions. A fast HEVC intracomputation method that uses image textural feature information instead of the traditional rate-distortion (R-D) cost value is proposed. This information is used to control the CU partition at the exterior of the xCompressCU function in the HEVC model coding framework. The method is comprised of two key steps: (1) simulation of the CU partition of the current frame image and calculation of the strength value of the textural features of each sub-block; and (2) disregard of some CU partition depths in the xCompressCU function, with no R-D cost calculation being required in the corresponding prediction mode. By leveraging these two steps, the computational complexity of the encoder is effectively lowered. Experimental results indicate that the proposed method reduces the coding time computational complexity of the current HEVC model by approximately 39.08% with negligible loss of coding efficiency.

We proposed a superresolution (SR) method based on example-learning framework. In our framework, the relationship between the output high-resolution (HR) estimation and the HR training images is approximated by the relationship between the low-resolution (LR) test image and the HR training images. To effectively capture the strong correlation between LR and HR images, the LR and HR images are mapped onto a common feature space. Furthermore, in order to maintain their original two-dimensional (2-D) spatial structure, the original LR and HR patches are mapped onto the underlying common feature space using 2-D canonical correlation analysis. Later, the relationship between HR and LR features is established by partial least squares (PLS) with low regression errors on the derived feature space. In addition, a steering kernel regression (SKR) constraint is integrated into patch aggregation to improve the quality of the recovered images. Finally, the effectiveness of our approach is validated by extensive experimental comparisons with several SR algorithms for the natural image superresolution both quantitatively and qualitatively.

In general, context-based adaptive variable length coding (CAVLC) decoding in H.264/AVC standard requires frequent access to the unstructured variable length coding tables (VLCTs) and significant memory accesses are consumed. Heavy memory accesses will cause high power consumption and time delays, which are serious problems for applications in portable multimedia devices. We propose a method for high-efficiency CAVLC decoding by using a program instead of all the VLCTs. The decoded codeword from VLCTs can be obtained without any table look-up and memory access. The experimental results show that the proposed algorithm achieves 100% memory access saving and 40% decoding time saving without degrading video quality. Additionally, the proposed algorithm shows a better performance compared with conventional CAVLC decoding, such as table look-up by sequential search, table look-up by binary search, Moon’s method, and Kim’s method.

Automated personal identification using the shape of the human ear is emerging as an appealing modality in biometric and forensic domains. This is mainly due to the fact that the ear pattern can provide rich and stable information to differentiate and recognize people. In the literature, there are many approaches and descriptors that achieve relatively good results in constrained environments. The recognition performance tends, however, to significantly decrease under illumination variation, pose variation, and partial occlusion. In this work, we investigate the use of local texture descriptors, namely local binary patterns, local phase quantization, and binarized statistical image features for robust human identification from two-dimensional ear imaging. In contrast to global image descriptors which compute features directly from the entire image, local descriptors representing the features in small local image patches have proven to be more effective in real-world conditions. Our extensive experimental results on the benchmarks IIT Delhi-1, IIT Delhi-2, and USTB ear databases show that local texture features in general and BSIF in particular provide a significant performance improvement compared to the state-of-the-art.

A video stabilization method based on a new concept inspired by the human visual system is presented. The human eye provides a stable scene by continuously changing the eye’s orientation in a way that always places the focused target in the center of one’s view. Similar to the human eye, the proposed algorithm focuses only on a single target object within a scene and stabilizes the target on the two-dimensional image plane by rotating a camera in three-dimensional space while most previous methods consider all objects in a video. The rotational angles of the camera along the x and y axes are simply predicted from a translational motion vector of the target object on the image plane. Hence, the proposed algorithm can provide a vivid video as if it was seen through the eye. Efficient video stabilization by approximating the human visual system is introduced, a practical method for real-time devices. Experimental results demonstrate that the visual feelings of the compensated videos are different depending on the selected target object and the approximating method provides a reasonable performance.

We propose a contrast enhancement algorithm considering surrounding information by illumination image. Conventional contrast enhancement techniques can be classified as a retinex-based method and a tone mapping function-based method. However, many retinex methods suffer from high-computational costs or halo artifacts. To cope with these problems, efficient edge-preserving smoothing methods have been researched. Tone mapping function-based methods are limited in terms of enhancement since they are applied without considering surrounding information. To solve these problems, we estimate an illumination image with local adaptive smoothness, and then utilize it as surrounding information. The local adaptive smoothness is calculated by using illumination image properties and an edge-adaptive filter based on the just noticeable difference model. Additionally, we employ a resizing method instead of a blur kernel to reduce the computational cost of illumination estimation. The estimated illumination image is incorporated with the tone mapping function to address the limitations of the tone mapping function-based method. With this approach, the amount of local contrast enhancement is increased. Experimental results show that the proposed algorithm enhances both global and local contrasts and produces better performance in objective evaluation metrics while preventing a halo artifact.

Design of fusion rule is an important step in fusion process. Traditional single fusion rules are inflexible when they are being used to fuse feature-rich images. To address this problem, an adaptive multistrategy image fusion method is proposed. Its flexibility lies in the combination of a choose-max strategy and a weighted average strategy. Moreover, the region-based characteristics and the shift-invariant shearlet transform (SIST)-based activity measures are proposed to guide the selection of strategies. The key points of our method are: (1) Window-based features are extracted from the source images. (2) Use of the fuzzy c-means clustering algorithm to construct a region map in the feature difference space. (3) The dissimilarity between corresponding regions is employed to quantify the characteristic of regions and the local average variance of the SIST coefficients are considered as activity measures to evaluate the salience of the related coefficient. (4) The adaptive multistrategy selection scheme is achieved by a sigmoid function. Experimental results show that the proposed method is superior to the conventional image fusion methods both in subjective and objective evaluations.

The pseudo-log image transform belongs to a class of image processing kernels that generate memory references which are nonlinear functions of loop indices. Due to the nonlinearity of the memory references, the usual design methodologies do not allow efficient hardware implementation for nonlinear kernels. For optimized hardware implementation, these kernels require the creation of a customized memory hierarchy and efficient data/memory management strategy. We present the design and real-time hardware implementation of a pseudo-log image transform IP (hardware image processing engine) using a memory management framework. The framework generates a controller which efficiently manages input data movement in the form of tiles between off-chip main memory, on-chip memory, and the core processing unit. The framework can jointly optimize the memory hierarchy and the tile computation schedule to reduce on-chip memory requirements, to maximize throughput, and to increase data reuse for reducing off-chip memory bandwidth requirements. The algorithmic C++ description of the pseudo-log kernel is profiled in the framework to generate an enhanced description with a customized memory hierarchy. The enhanced description of the kernel is then used for high-level synthesis (HLS) to perform architectural design space exploration in order to find an optimal implementation under given performance constraints. The optimized register transfer level implementation of the IP generated after HLS is used for performance estimation. The performance estimation is done in a simulation framework to characterize the IP with different external off-chip memory latencies and a variety of data transfer policies. Experimental results show that the designed IP can be used for real-time implementation and that the generated memory hierarchy is capable of feeding the IP with a sufficiently high bandwidth even in the presence of long external memory latencies.

An efficient method for expression-invariant three-dimensional (3-D) face reconstruction from a frontal face image with a variety of facial expressions (FE) using the FE generic elastic model (GEM) is proposed. Three generic models are employed for FE modeling in the generic elastic model (GEM) framework, which are combined based on the similarity distance around the lips. Exclusively, FE-GEM demonstrated that it is more precisely able to estimate a 3-D model of a frontal face, attaining a more robust and better quality 3-D face reconstruction under a variety of FEs compared to the original GEM approach. It is tested on an available 3-D face database and its accuracy and robustness are demonstrated compared to the GEM approach under a variety of FEs. Also, the FE-GEM method is tested on available two-dimensional face databases and a new synthesized pose is generated from gallery images for handling pose variations in face recognition.

Scene-based nonuniformity correction (SBNUC) techniques provide a means of identifying and correcting focal plane array nonuniformity (NU) through algorithmic analysis of the camera output. SBNUC techniques rely almost universally on camera motion to provide a means of separating the scene from the NU pattern. A simulation is developed that is used to explore the role and effect of camera motion on two representative registration-based SBNUC algorithms: interframe registration-based least mean square (IRLMS) by Zuo et al. and feedback-integrated scene-cancellation (FiSC) by Black and Tyo. The effect of camera motion velocity and direction between frames is examined. The high spatial frequency portion of NU is shown to be corrected by both IRLMS and FiSC, and this correction is relatively indifferent to nonzero camera motion parameters. The FiSC algorithm was specifically designed to incorporate the low spatial frequency component into registration-based SBNUC, but demonstrates a strong dependency on camera motion. Techniques for mitigation of camera motion parameter effects through tradespace and buffering are presented and tested. With proper mitigation and camera motion, FiSC is shown to correct most high and low spatial frequency NU with fewer than 100 framepairs repeatedly processed using techniques suitable for real-time processing implementation.

This paper describes a method for moving shadow detection using the joint histogram of multifeatures. In our method, we first obtain the moving region by background subtraction. Then, based on the intensity feature, candidate shadow regions are extracted. Moreover, the joint histogram of intensity, color, and gradient features is constructed in candidate background and foreground regions. Furthermore, the joint histogram is backprojected to the foreground regions to yield the moving shadow likelihood image. In the end, the adaptive threshold is derived by the joint histogram of the foreground and background, and accurate shadow regions are extracted by segmenting the shadow likelihood image with this threshold. The main contribution of this paper is twofold. First, multifeatures are fused together by the joint histogram, which is a unified and simple description method for shadow detection. Second, the histogram of background and foreground was compared with backprojection. Moreover, the final result only depends on a few parameters. Unlike other approaches, our method does not make any assumption and moving shadow regions can be detected fast and accurately. Experimental results show that the proposed method is efficient and robust over a broad range of shadow types and challenging video conditions.

A dictionary-learning-based image denoising algorithm is proposed in this paper. Since traditional methods seldom take into account the rotational invariance of dictionaries learned from image patches, an improved K-means singular value decomposition algorithm is developed. In our method, the rotational version of atoms is introduced to greedily match the noisy image in a sparse coding procedure. On the other hand, in a dictionary learning procedure, to maximize the diversity of atoms, a rotational operation on the residual error is adopted such that the rotational correlation among atoms is reduced. As the strategy exploits the rotational invariance of atoms, more intrinsic features existing in image patches can be effectively extracted. Experiments illustrate that the proposed method can achieve a better performance than some other well-developed denoising methods.

In the test model of high efficiency video coding (HEVC) standard–based three-dimensional (3-D) video coding (3-D-HEVC), the variable size motion estimation (ME) and disparity estimation (DE) have been employed to select the best coding mode for each treeblock in the encoding process. This technique achieves the highest possible coding efficiency, but it brings extremely high computational complexity that limits 3-D-HEVC from practical applications. An early SKIP mode decision algorithm based on spatial and interview correlations is proposed to reduce the computational complexity of the ME/DE procedures. The basic idea of the method is to utilize the spatial and interview properties of coding information in previous coded frames to predict the current treeblock prediction mode and early skip unnecessary variable-size ME and DE. Experimental results show that the proposed algorithm can significantly reduce computational complexity of 3-D-HEVC while maintaining nearly the same rate distortion performance as the original encoder.

With a particular function in plate-type structure tomography, computed laminography (CL) has received increasing attention in industrial nondestructive testing and become an important branch of computed tomography (CT). For the reconstruction algorithm of CL, center of rotation (COR) is the most important parameter determining the reconstruction accuracy and must be located precisely. Otherwise, even a tiny error of COR will cause obvious artifacts in reconstructed images. In order to realize measurement of COR with high accuracy and efficiency, a feasible calibration method was proposed to determine the position of COR without dedicated phantoms. According to this method, when a sample fixed on the rotational stage turns around the rotational axis, the locus of the sample’s projection on the imaging plane will be an ellipse. Consistent with the symmetrical property of an elliptical image, a cross-correlation operation is adopted to determine the position of COR by locating the peak value of the cross-correlation function. The computer simulation and experimental results demonstrate that this method has high accuracy, and strong anti-noise and anti-wobble ability. In particular, this method does not need a dedicated phantom to perform the calibration, but rather uses projections of the inspected sample to calculate COR directly.

The matching between an open contour and a closed contour is a basis for the alignment of their common part and a similarity measure. We propose a coarse-to-fine method for partial shape matching, which does not need to scan the target shape, construct a codebook of model contour fragments, or depend on background support domains. For this purpose, a linearization algorithm for partial shapes is introduced to extract the initial shape segments from the closed contour those possibly match with the open contour. The next refining procedure of the coarse matching eliminates significantly dissimilar shape segments to reduce the further processing of fine matching. We propose a shape similarity description to finely describe the similarity between the open contour and the remaining shape segments. Finally, an order-preserving point injection between the open contour and the closed contour is established. Valuations of the proposed method on a benchmark dataset and real images demonstrate that the overall and component performances are excellent and robust to various disturbances and similarity transformations. Last, a gesture recognition application is implemented.

We present a multi-line-scan light-field image acquisition and processing system designed for 2.5/3-D inspection of fine surface structures in industrial environments. The acquired three-dimensional light field is composed of multiple observations of an object viewed from different angles. The acquisition system consists of an area-scan camera that allows for a small number of sensor lines to be extracted at high frame rates, and a mechanism for transporting an inspected object at a constant speed and direction. During acquisition, an object is moved orthogonally to the camera’s optical axis as well as the orientation of the sensor lines and a predefined subset of lines is read out from the sensor at each time step. This allows for the construction of so-called epipolar plane images (EPIs) and subsequent EPI-based depth estimation. We compare several approaches based on testing a set of slope hypotheses in the EPI domain. Hypotheses are derived from block matching, namely the sum of absolute differences, modified sum of absolute differences, normalized cross correlation, census transform, and modified census transform. Results for depth estimation and all-in-focus image generation are presented for synthetic and real data.

In order to address the challenges that both the training and testing images are contaminated by random pixels corruption, occlusion, and disguise, a robust face recognition algorithm based on two-stage sparse representation is proposed. Specifically, noises in the training images are first eliminated by low-rank matrix recovery. Then, by exploiting the first-stage sparse representation computed by solving a new extended ℓ¹-minimization problem, noises in the testing image can be successfully removed. After the elimination, feature extraction techniques that are more discriminative but are sensitive to noise can be effectively performed on the reconstructed clean images, and the final classification is accomplished by utilizing the second-stage sparse representation obtained by solving the reduced ℓ¹-minimization problem in a low-dimensional feature space. Extensive experiments are conducted on publicly available databases to verify the superiority and robustness of our algorithm.

Due to the security and privacy-preserving requirements for cloud data management, it is sometimes desired that video content is accessible in an encrypted form. Reversible data hiding in the encrypted domain is an emerging technology, as it can perform data hiding in encrypted videos without decryption, which preserves the confidentiality of the content. Furthermore, the original cover can be losslessly restored after decryption and data extraction. An efficient reversible data hiding scheme for encrypted H.264/AVC videos is proposed. During H.264/AVC encoding, the intraprediction mode, motion vector difference, and the sign bits of the residue coefficients are encrypted using a standard stream cipher. Then, the data-hider who does not know the original video content, may reversibly embed secret data into the encrypted H.264/AVC video by using a modified version of the histogram shifting technique. A scale factor is utilized for selecting the embedding zone, which is scalable for different capacity requirements. With an encrypted video containing hidden data, data extraction can be carried out either in the encrypted or decrypted domain. In addition, real reversibility is realized so that data extraction and video recovery are free of any error. Experimental results demonstrate the feasibility and efficiency of the proposed scheme.

Based on the fact that human attention is more likely to be attracted by different objects or statistical outliers of a scene, a bottom-up saliency detection model is proposed. Our model regards the saliency patterns of an image as the outliers in a dataset. For an input image, first, each image element is described as a feature vector. The whole image is considered as a dataset and an image element is classified as a saliency pattern if its corresponding feature vector is an outlier among the dataset. Then, a binary label map can be built to indicate the salient and the nonsalient elements in the image. According to the Boolean map theory, we compute multiple binary maps as a set of Boolean maps which indicate the outliers in multilevels. Finally, we linearly fused them into the final saliency map. This saliency model is used to predict the human eye fixation, and has been tested on the most widely used three benchmark datasets and compared with eight state-of-the-art saliency models. In our experiments, we adopt the shuffled the area under curve metric to evaluate the accuracy of our model. The experimental results show that our model outperforms the state-of-the-art models on all three datasets.

The normalized cut (Ncut) method is a popular method for segmenting images and videos. The Ncut method segments an image into two disjoint regions, each segmented by the same method. After the Ncut method has been recursively applied to an image, its final segmented image is obtained. The main drawback of the Ncut method is that a user cannot easily determine the stop criteria because users have no idea about the number of regions in an image. This work proposes the genetic cut (Gcut) algorithm to resolve this shortcoming. Users do need not to specify thresholds in the Gcut algorithm, which automatically segments an image into the proper number of regions. Also, the neighbor-merging (NM) algorithm is proposed for preprocessing the images and improves the performance of the Gcut algorithm. Thus, the proposed Gcut method combines the NM and Gcut algorithms. Furthermore, a heuristic method is proposed to identify a good segment for the Gcut method. In all experiments, the proposed Gcut method outperforms traditional Ncut methods.

Accurate recognition of airborne pollen taxa is crucial for understanding and treating allergic diseases which affect an important proportion of the world population. Modern computer vision techniques enable the detection of discriminant characteristics. Apertures are among the important characteristics which have not been adequately explored until now. A flexible method of detection, localization, and counting of apertures of different pollen taxa with varying appearances is proposed. Aperture description is based on primitive images following the bag-of-words strategy. A confidence map is estimated based on the classification of sampled regions. The method is designed to be extended modularly to new aperture types employing the same algorithm by building individual classifiers. The method was evaluated on the top five allergenic pollen taxa in Germany, and its robustness to unseen particles was verified.