To improve the performance of a top-hat transformation for infrared small target enhancement, a new class of top-hat transformation through structuring element construction and operation reorganization is proposed. The structuring element construction and operation reorganization are based on the property of the infrared small target image and thus can greatly improve the performance of small target enhancement. Experimental results verified that it was very efficient.

Automated detection of the clinically relevant image planes in a three-dimensional echocardiographic (3-DE) data set has potential applications for diagnosis of pediatric congenital heart disease. We propose a method based on template matching to automatically detect the four-chamber image plane (4CIP) in a 3-DE data set. First, a normal four-chamber image is chosen as the template. Second, to find the 4CIP in a 3-D volume, a series of cross sections are extracted from this volume. Then, a coarse-to-fine retrieval is applied to the stack of slices and the image most similar to the template is proposed as the 4CIP of this volume. We tested the method on 28 data sets of normal children and 22 data sets of children suffering from congenital heart disease, and the error rates are 7 and 9%, respectively. As will be shown, the autodetection algorithm can be straightforwardly implemented, has low computational complexity, and is robust to noise.

We present a method of introducing rotation invariance in texture features based on a local Fourier histogram (LFH) computed using a 1-D discrete Fourier transform (DFT). To compensate for image rotation, a local image-gradient angle at each image pixel is found from within one of the 1-D DFT coefficients. The rotation invariance is established theoretically, analytically as well as empirically. The rotation-compensated features extracted from the same texture image oriented at different angles exhibit very high cross correlation. Therefore, the proposed texture features are expected to yield very high accuracies for a variety of image data and applications. The improved LFH-based features outperform the earlier version of the features and the features based on Gabor filters in texture recognition on 8560 images from the Brodatz album.

Twenty five years after the seminal work of Jean Morlet, the wavelet transform, multiresolution analysis, and other space-frequency or space-scale approaches are considered standard tools by researchers in image processing. Many applications that point out the interest of these techniques have been proposed. We review the recent published work dealing with industrial applications of the wavelet and, more generally speaking, multiresolution analysis. We present more than 190 recent papers.

3D modeling is an important topic concerning visual inspection in automatic quality control. Through visual inspection, it is possible to determine whether a product fulfills the required specifications or whether it contains surface or volume imperfections. Although some processes such as color analysis can be achieved by 2D techniques, more challenging tasks such as volume inspection of large and complex objects/scenes may require the use of accurate 3D registration techniques. 3D simultaneous localization and mapping has become a very important research topic not only in the robotics field for solving problems such as robot navigation and mapping of 2D/3D scenarios but also in the computer vision community for the estimation of the camera pose and the registration of large-scale objects. Although their techniques differ slightly depending on the application, both communities tend to solve similar problems by means of different approaches. We present a survey of the techniques used by the robotics and computer vision communities, pointing out the pros and cons and potential applications of each approach. Furthermore, the most representative techniques have been programmed and tested, obtaining experimental results that provide an accurate comparison of the methods in the presence of noise and outliers.

Defect detection in images is a current task in quality control and is often integrated in partially or fully automated systems. Assessing the performances of defect detection algorithms is thus of great interest. However, because this is application- and context-dependent, it remains a difficult task. We describe a methodology to measure the performances of such algorithms on large images in a semi-automated defect inspection situation. Considering standard problems occurring in real cases, we compare typical performance evaluation methods. This analysis leads to the construction of a simple and practical receiver operating characteristic (ROC) based method. This method extends the pixel-level ROC analysis to an object-based approach by dilating the ground truth and the set of detected pixels before calculating the true-positive and false-positive rates. These dilations are computed thanks to the a priori knowledge of a human-defined ground truth and gives to true-positive and false-positive rates more consistent values in the semi-automated inspection context. Moreover, the dilation process is designed to be automatically suited to the object's shape in order to be applied on all types of defects without any parameter to be tuned.

We present a new efficient method for calibration of catadioptric sensors. The method is based on an accurate measurement of the three-dimensional parameters of the mirror through polarization imaging. While inserting a rotating polarizer between the camera and the mirror, the system is automatically calibrated without any calibration patterns. Moreover, this method permits most of the constraints related to the calibration of catadioptric systems to be relaxed. We show that, contrary to our system, the traditional methods of calibration are very sensitive to misalignment of the camera axis and the symmetry axis of the mirror. From the measurement of three-dimensional parameters, we apply the generic calibration concept to calibrate the catadioptric sensor. We also show the influence of the disturbed measurement of the parameters on the reconstruction of a synthetic scene. Finally, experiments prove the validity of the method with some preliminary results on three-dimensional reconstruction.

We present an innovative way to simultaneously perform feature extraction and classification for the quality-control issue of surface grading by applying two multivariate statistical projection methods: SIMCA and PLS-DA. These tools have been applied to compress the color texture data that describe the visual appearance of surfaces (soft color texture descriptors) and to directly perform classification using statistics and predictions from the projection models. Experiments have been carried out using an extensive ceramic images database (VxC TSG) comprised of 14 different models, 42 surface classes, and 960 pieces. A factorial experimental design evaluated all the combinations of several factors affecting the accuracy rate. These factors include the tile model, color representation scheme (CIE Lab, CIE Luv, and RGB), and compression/classification approach (SIMCA and PLS-DA). Moreover, a logistic regression model is fitted from the experiments to compute accuracy estimates and study the effect of the factors on the accuracy rate. Results show that PLS-DA performs better than SIMCA, achieving a mean accuracy rate of 98.95%. These results outperform those obtained in a previous work where the soft color texture descriptors in combination with the CIE Lab color space and the k-NN classifier achieved an accuracy rate of 97.36%.

We propose an evaluation system that assigns each line-type thin-film transistor liquid-crystal display defect a corresponding level that objectively agrees with human visual perception. By “objective,” we mean that the evaluation corresponds, on average, with the assessment of a group of inspectors. The basic idea is to use the human visual perception to evaluate defects. Crucial features of defects are selected to represent the human visual perception for the line-type defect. In the process, we define the “just-noticeable difference surface” (JND) and evaluate the level of defect as the distance from a feature point consisting of selected features of the JND.

We introduce an algorithm dedicated to the detection of the axes of cylindrical objects in a 3D block. The proposed algorithm performs 3D axis detection without prior segmentation of the block. This approach is specifically appropriate when the gray levels of the cylindrical objects are not homogeneous and are thus difficult to distinguish from the background. The method relies on gradient and curvature estimation and operates in two main steps. The first one selects candidate voxels for the axes, and the second one refines the determination of the axis of each cylindrical object. Applied to fiber-reinforced composite materials, this algorithm detects the axes of fibers in order to obtain the geometrical characteristics of the reinforcement. Knowing the reinforcement characteristics is an important issue in the quality control of the material but also in the prediction of the thermal and mechanical performances. We detail the various steps of the algorithm and then present some results, obtained with both synthetic blocks and real data acquired by synchrotron X-ray micro-tomography on carbon-fiber-reinforced carbon composites.

The management of mineral fertilization using centrifugal spreaders calls for the development of spread pattern characterization devices to improve the quality of fertilizer spreading. In order to predict spread pattern deposition using a ballistic flight model, several parameters need to be determined, in particular, the velocity of the granules when they leave the spinning disc. We demonstrate that a motion-blurred image acquired in the vicinity of the disc by a low-cost imaging system can provide the three-dimensional components of the outlet velocity of the particles. A binary image is first obtained using a recursive linear filter. Then an original method based on the Hough transform is developed to identify the particle trajectories and to measure their horizontal outlet angles, not only in the case of horizontal motion but also in the case of three-dimensional motion. The method combines a geometric approach and mechanical knowledge derived from spreading analysis. The outlet velocities are deduced from outlet angle measurements using kinematic relationships. Experimental results provide preliminary validations of the technique.

We present the design and realization of an imaging system intended for use as a reference method for the accurate measurement of cotton fiber length. The prototype system is composed of an off-the-shelf scanner that generates a grayscale image of multiple individualized fibers, followed by customized image processing algorithms that compute the length of each fiber in the image. Although the system requires some degree of separation between the individual fibers at scan time, it is shown to produce highly accurate length measurements that are invariant to fiber orientation, shape, interfiber intersections, and intrafiber crimps and crossovers. Hence, in its present state, the proposed system serves as an excellent reference method for assessing the efficacy of commercially available length measurement systems.

We propose a method using a machine vision technique to detect the defects caused by spatially non-uniform brightness on LCD panels. The detection method is based on human vision, so subjective experiments were conducted to investigate the correlation between the parameters related to non-uniformity and the degree of easy observability. The visibility of the defects depends mainly on the spatial gradient of luminance variation. The proposed method uses the spatial gradient calculated using extracted contours to detect the defects caused by non-uniform brightness. The detection method is composed of four parts: the extraction of contours, the calculation of spatial gradient, the decision about defects, and the display of defects. Finally, the spatial gradient has a near-logarithmic relationship to the observability in judging the degree of how easily the detected defects are observed by the human inspector.

The importance of the perceived quality measurement is fundamental for many image processing applications, such as compression, acquisition, restoration, enhancement, and reproduction. Color information is also of great importance for the perceived image quality, although perceived information is mainly represented by luminance. We present a computational and memory-efficient no-reference image quality assessment model independent of JPEG and JPEG2000 coded color images based on local regions. We also present the discrimination algorithm for these two types of coded images. The features of local regions are blockiness around the block boundary, average absolute difference between adjacent pixels within the block, and zero crossing rate within the block of the image. We validate the performance of our model on our subjective database, which shows good quality prediction performance, and the model's generalization ability is also verified on the other database.

Quality metrics, within the field of laser range imaging, are used to quantify by how much some aspect of a measurement deviates from a predefined standard. Measurement quality evaluations are becoming increasingly important in laser range imaging for range image registration, merging measurements, and planning the next best view. Spatial uncertainty and resolution are the primary metrics of image quality; however, spatial uncertainty is affected by a variety of environmental factors. A review how contemporary researchers have attempted to quantify these environmental factors is presented, along with spatial uncertainty and resolution, resulting in a wide range of quality metrics.

We introduce a novel workflow that will hopefully open new directions of processing and improvement in image reproduction. Existing gamut mapping algorithms can be classified into two basic categories: image-independent algorithms and image-dependent algorithms. The latter algorithms produce better reproduction; however, because they are time consuming and mathematically complex, the image-independent approach is commonly used in most imaging workflows. We suggest a new workflow that attempts to approach the image-dependent mapping method without incurring significant computational drawbacks nor requiring changes in the imaging industrial standards. The proposed method attempts to choose an appropriate gamut mapping per image without reconstructing the image gamut itself and without constructing an image-specific mapping on the fly, as required by image-dependent gamut mapping methods. Specifically, image characteristics are exploited for selection of a source gamut and a gamut mapping most appropriate for a given input image from a set of available mappings. Accordingly the proposed method is named image-guided gamut mapping. We show the practicability and advantages of the suggested workflow in several specific cases. We show that better image quality is achieved for 87% of the tested images when using the suggested workflow.

Custom color transformations for images or video can be learned from a small set of sample color pairs by estimating a look-up table (LUT) to describe the enhancement and storing the LUT in an International Color Consortium profile, which is a standard tool for color management. Estimating an accurate LUT from a small set of sample color pairs is challenging. Local linear and ridge regression are tested on six definitions of neighborhoods for twenty color enhancements and twenty-five color images. Excellent results were obtained with local ridge regression over proposed enclosing neighborhoods, including a variant of Sibson's natural neighbors. The evaluation of the different estimation methods for this task compared the fidelity of the learned color enhancement to the original sample color pairs and the presence of objectionable artifacts in enhanced images. These metrics show that enclosing neighborhoods are promising adaptive neighborhood definitions for local classification and regression.

Shu-Mei Guo and Bing-Hui Wu National Cheng Kung University, Department of Computer Science and Information Engineering, Tainan, Taiwan Jason Sheng-Hong Tsai National Cheng Kung University, Department of Electric Engineering, Tainan, Taiwan In the last few years, wavelet-based still image compression schemes have become one of the most popular fields of research. Several powerful methods of encoding the wavelet coefficients have been proposed and set-partitioning embedded block (SPECK) is one of the most popular techniques. We propose a method called HSPECK (Huffman SPECK), which is based on SPECK to further improve the compression ratio. By observation, we are able to establish the probability distribution over the significance pattern of four subblocks generated successively along the wavelet decomposition process. Therefore, the Huffman coding scheme can then be employed for coding wavelet coefficients. The experimental results show that under the same peak signal-to-noise ratio, the bit rates via the proposed approach improve between 6.1% and 14.9%.

The dyadic wavelet-based set partitioning in hierarchical trees (SPIHT) is highly efficient in coding nontextured images while the wavelet packet transform (WPT) is able to provide an optimal representation for textured images. However, incorporating the WPT with SPIHT to improve the performance of SPIHT is not as straightforward as it would seem. Although previous solutions generally adapt the zero-tree structure to the WPT, we introduce a method for selecting an optimal WPT basis for SPIHT, which efficiently compacts the high-frequency subband energy into as few trees as possible and avoids parental conflicts. The proposed SPIHT-WPT coder achieves improved coding gains, especially for highly textured images. Furthermore, the selected WPT is compatible with the region-of-interest and error resilient SPIHT coding schemes.

A removable visible watermarking scheme, which operates in the discrete cosine transform (DCT) domain, is proposed for combating copyright piracy. First, the original watermark image is divided into 16×16 blocks and the preprocessed watermark to be embedded is generated by performing element-by-element matrix multiplication on the DCT coefficient matrix of each block and a key-based matrix. The intention of generating the preprocessed watermark is to guarantee the infeasibility of the illegal removal of the embedded watermark by the unauthorized users. Then, adaptive scaling and embedding factors are computed for each block of the host image and the preprocessed watermark according to the features of the corresponding blocks to better match the human visual system characteristics. Finally, the significant DCT coefficients of the preprocessed watermark are adaptively added to those of the host image to yield the watermarked image. The watermarking system is robust against compression to some extent. The performance of the proposed method is verified, and the test results show that the introduced scheme succeeds in preventing the embedded watermark from illegal removal. Moreover, experimental results demonstrate that legally recovered images can achieve superior visual effects, and peak signal-to-noise ratio values of these images are >50 dB.

This paper proposes a novel fragile watermarking algorithm, designated as the compression-watermarking algorithm (CWA), which inserts watermark information in a JPEG image by modifying the last nonzero coefficient in each discrete cosine transform (DCT) quantized block. The proposed algorithm not only provides an authentication capability, but also decreases the size of JPEG compressed-domain images. The experimental results show that CWA has virtually no impact on the visual quality of the watermarked images and is highly sensitive to image modifications. Furthermore, it is found that CWA reduces the size of watermarked image by as much as 6.3% when applied to the watermarking of standard JPEG test images. Therefore, CWA provides a feasible solution for image authentication and data reduction in DCT-based domains such as JPEG and MPEG-family coders/decoders.

With the wide application of multimedia data, multimedia content protection becomes urgent. Until now, various means have been reported that can be classified into several types according to their functionalities, such as data encryption, digital watermarking, or data authentication. They are used to protect multimedia data's confidentiality, ownership, and integrity, respectively. A wavelet-based multifeature semifragile multimedia authentication scheme is proposed. According to the approximation component and the energy relationship between the subbands of the detail component, global feature and local feature are both generated. Then, the global watermark and local watermark are generated from global feature and local feature, respectively. Two watermarks are then embedded into the multimedia data themselves, in the wavelet domain. Both the feature extraction and embedding processes are controlled by secret keys to improve the security of the proposed scheme. On the receiver side, the extracted watermark and the one generated from the received image are compared to determine the tampered location. A new authentication method is designed, and it is proved valid in the experiments. This authentication scheme is robust to general compression; sensitive to cutting, pasting, or modification; efficient in real-time operation; and secure for practical applications.

We report on three-dimensional image reconstruction from a limited set of computed tomography projections. We focus on configurations with very limited angle of view and on applications in which the image to be reconstructed is composed of one or several localized objects laying in a known background. We propose an original method based on the detection of the localized object voxels and on a sparse modeling of the image. Reconstruction is done by computing the maximum a posteriori estimator of the image parameters. To implement image reconstruction, we adopt a multigrid strategy in which coarse-to-fine resoluted images are successively reconstructed. This strategy provides detection of localized object voxels as well as accurate initial solutions at each resolution level. Each optimization stage is carried out by using an iterative deterministic descent algorithm. We propose a convergent single-site update algorithm that consists of successive constrained optimizations with respect to one voxel at a time. We show the performance of the multigrid method on simulated data corresponding to a set of limited angle cone-beam projections of a synthetic image. The results are accurate, while both memory storage and numerical time of computation are dramatically reduced compared to the monogrid reconstruction method.

Image filtering is a very important task in any image-processing system, since the output of the filtering constitutes the primary input to high-level vision, which then utilizes domain-specific knowledge to interpret and analyze the image contents. We propose a new approach to filtering the noise of a stream of thermographic line scans. The filter is designed to be applied in real time and is divided in two components: an intrascan filter and an interscan filter. The intrascan filter is based on spatial overlapping which occurs when using high acquisition rates. The interscan filter is based on edge detection and is able to adapt the way it produces the filtered output based on the detected edges. A procedure to automatically tune the parameters of the interscan filter is described and applied. The results of the proposed filters are compared to those obtained using the average and the median filter. In addition, the real-time performance of the proposed filters is analyzed by measuring the time taken by the tasks involved in their application.

An image persistence framework is developed to analyze azimuthally varying synthetic aperture radar (SAR) data. The model focuses on cases containing rich aspect data from a single depression angle. The goal is to replace the data's intrinsic viewing geometry dependencies with target-specific dependencies. Both direct mapping functions and cost functions are presented for data transformation. An intensity-only mapping function is realized to illustrate the persistence model in terms of a canonical example, visualization, and classification. The limitations of an intensity-only mapping function are also discussed. Because the new image-space output from the persistence model is closely tied to the radio frequency (RF) characteristics of two different targets and not to the collection geometry, it shows promise for integration into various automated target recognition (ATR) algorithms. Target classification potential was tested using the MSTAR database and simplified template matching schemes. The intensity-only transformation function permits a very low number of persistence images to represent the general RF target characteristics of an assortment of vehicles. The ability to represent target characteristics with relatively low resources could also be beneficial in ATR applications. Overall, the persistence framework shows strong potential as a new tool that can be used in the analysis of multiaspect SAR images.

An alternative solution for surface inspection is being presented. It is based on a concerted combination of an adapted stripe-illumination principle together with an image processing approach specialized on the analysis of the obtained stripe images. This approach is capable of detecting, segmenting, and classifying nondefective surfaces, as well as three- and two-dimensional defective surfaces from perturbations in the stripe illumination. In contrast to alternative procedures, no calibration of illumination or camera is necessary. The principle of the proposed method using a concrete industrial application for the inspection of cylindrical metallic surfaces under structured lighting is explained. Furthermore, based on several examples involving different surface types, we demonstrate the broad range of applications for the proposed algorithm.

Motion blur is a result of finite acquisition time of practical cameras and the relative motion between the camera and moving objects. We present a method for speed measurement of spherical objects using motion blurred images captured by a digital camera. The object is assumed under a straight line uniform-velocity motion, and the speed is calculated according to the imaging geometry and blur extent estimates. We have established a link between the motion blur information of a 2-D image and the speed information of a moving object. Experimental results are presented for the real-scene images.

Images included in documents usually provide information that may not be readily expressible by words. For example, academic articles with similar pictures may be of interest for researchers. We deal with the problem of extracting images in digital document. Given a digital document, the optimal block size is first determined by finding the best fit of the horizontally projected gray-level pattern to a set of orthogonal basis vectors. Because the block with the optimal size is supposed to contain sufficient information to identify text regions, the proposed method is font-size independent regardless of the size of the words in the text lines. The blocks divided by the optimal block size are classified into one of image, text, and background blocks. This block classification result, in turn, is used for the initial configuration for blockwise document segmentation. The blockwise segmentation method is based on the maximum a posteriori (MAP) framework with a deterministic relaxation algorithm. After the blockwise segmentation, each boundary block in the image region is further divided into four subblocks and the class labels for these subblocks are updated. These subdivision and class updating processes are executed recursively until we have a pixel-level segmentation. Experimental results show that the proposed image extraction method yields 2.9% error rates for 232 documents in the Oulu database.

The C-fuzzy decision tree (CFDT)–based on the fuzzy C-means (FCM) algorithm has been proposed recently. In many experiments, the CFDT performs better than the “standard” decision tree, namely, the C4.5. A new C-fuzzy decision tree (NCFDT) is proposed, and it outperforms the CFDT. Two design issues for NCFDT are as follows. First, the growing method of NCFDT is based on both classification error rate and the average number of comparisons for the decision tree, whereas that of CFDT only addresses classification error rate. Thus, the proposed NCFDT performs better than the CFDT. Next, the classified point replaces the cluster center to classify the input vector in the NCFDT. The classified-points searching algorithm is proposed to search for one classified point in each cluster. The classification error rate of the NCFDT with classified points is smaller than that of CFDT with cluster centers. Furthermore, these classified points can be applied to the CFDT to reduce classification error rate. The performance of NCFDT is compared to CFDT and other methods in experiments.

In this paper, a multiple description image coding scheme is proposed to facilitate the transmission of images over media with possible packet loss. The proposed method is based on finding the optimal reconstruction filter coefficients that will be used to reconstruct lost descriptions. For this purpose initially, the original image is downsampled and each subimage is coded using standard JPEG. These decoded images are then mapped to the original image size using the optimal filters. Multiple descriptions consist of coded down-sampled images and the corresponding optimal reconstruction filter coefficients. It is shown that the proposed method provided better results compared to standard interpolation filters (i.e., bicubic and bilinear).

When sequence noise increases, the prediction process in the H.264 advanced video coding (AVC) encoder becomes less efficient. The residue resulting from the difference between the matching macroblocks in the frame coded in predictive mode and the best reference frame increases dramatically. To solve this problem, we propose to embed in the H.264/AVC encoder a motion detection technique based on the Markov random fields algorithm, which relies on robust moving pixel segmentation and reduces the luminance variation that results more often from noise rather than motion. This algorithm needs three reference frames to detect real motion between matching macroblocks, and it allows a decrease in bit rate while maintaining the compressed sequence quality.

We perform a statistical analysis of curvelet coefficients, distinguishing between two classes of coefficients: those that contain a significant noise-free component, which we call the “signal of interest,” and those that do not. By investigating the marginal statistics, we develop a prior model for curvelet coefficients. The analysis of the joint intra- and inter-band statistics enables us to develop an appropriate local spatial activity indicator for curvelets. Finally, based on our findings, we present a novel denoising method, inspired by a recent wavelet domain method called ProbShrink. The new method outperforms its wavelet-based counterpart and produces results that are close to those of state-of-the-art denoisers.