This PDF file contains the editorial “Changing Times” for JEI Vol. 25 Issue 02

In diffraction grating, at times, there are defective pixels on the focal plane array; this results in horizontal lines of corrupted pixels in some channels. Since only a few such pixels exist, the corruption/noise is sparse. Studies on sparse noise removal from hyperspectral noise are parsimonious. To remove such sparse noise, a prior work exploited the interband spectral correlation along with intraband spatial redundancy to yield a sparse representation in transform domains. We improve upon the prior technique. The intraband spatial redundancy is modeled as a sparse set of transform coefficients and the interband spectral correlation is modeled as a rank deficient matrix. The resulting optimization problem is solved using the split Bregman technique. Comparative experimental results show that our proposed approach is better than the previous one.

Stereoscopic image quality assessment (IQA) plays a vital role in stereoscopic image/video processing systems. We propose a new quality assessment for stereoscopic image that uses disparity-compensated view filtering (DCVF). First, because a stereoscopic image is composed of different frequency components, DCVF is designed to decompose it into high-pass and low-pass components. Then, the qualities of different frequency components are acquired according to their phase congruency and coefficient distribution characteristics. Finally, support vector regression is utilized to establish a mapping model between the component qualities and subjective qualities, and stereoscopic image quality is calculated using this mapping model. Experiments on the LIVE 3-D IQA database and NBU 3-D IQA databases demonstrate that the proposed method can evaluate stereoscopic image quality accurately. Compared with several state-of-the-art quality assessment methods, the proposed method is more consistent with human perception.

The existing Krawtchouk moment invariants are derived by a linear combination of geometric moment invariants. This indirect method cannot achieve perfect performance in rotation, scale, and translation (RST) invariant image recognition since the derivation of these invariants are not built on Krawtchouk polynomials. A direct method to derive RST invariants from Krawtchouk moments, named explicit Krawtchouk moment invariants, is proposed. The proposed method drives Krawtchouk moment invariants by algebraically eliminating the distorted (i.e., rotated, scaled, and translated) factor contained in the Krawtchouk moments of distorted image. Experimental results show that, compared with the indirect methods, the proposed approach can significantly improve the performance in terms of recognition accuracy and noise robustness.

A robust object tracking system that is invariant to object appearance variations and background clutter is proposed. Multiple instance learning with a boosting algorithm is applied to select discriminant texture information between the object and background data. Additionally, depth information, which is important to distinguish the object from a complicated background, is integrated. We propose two depth-based models that can compensate texture information to cope with both appearance variants and background clutter. Moreover, in order to reduce the risk of drifting problem increased for the textureless depth templates, an update mechanism is proposed to select more precise tracking results to avoid incorrect model updates. In the experiments, the robustness of the proposed system is evaluated and quantitative results are provided for performance analysis. Experimental results show that the proposed system can provide the best success rate and has more accurate tracking results than other well-known algorithms.

Today there is need for no reference (NR) objective perceived image quality measurement techniques as conducting subjective experiments and making reference image available is a very difficult task. Very few NR perceived image quality measurement algorithms are available for color distortions like chromatic aberration (CA), color quantization with dither, and color saturation. We proposed NR image quality assessment (NR-IQA) algorithms for images distorted with CA. CA is mostly observed in images taken with digital cameras, having higher sensor resolution with inexpensive lenses. We compared our metric performance with two state-of-the-art NR blur techniques, one full reference IQA technique and three general-purpose NR-IQA techniques, although they are not tailored for CA. We used a CA dataset in the TID-2013 color image database to evaluate performance. Proposed algorithms give comparable performance with state-of-the-art techniques in terms of performance parameters and outperform them in terms of monotonicity and computational complexity. We have also discovered that the proposed CA algorithm best predicts perceived image quality of images distorted with realistic CA.

It is a challenging task to develop an effective visual tracking algorithm due to factors such as pose variation, rotation, and so on. Combined discriminative global and generative local appearance models are proposed to address this problem. Specifically, we develop a compact global object representation by extracting the low-frequency coefficients of the color and texture of the object based on two-dimensional discrete cosine transform. Then, with the global appearance representation, we learn a discriminative metric classifier in an online fashion to differentiate the target object from its background, which is very important to robustly indicate the changes in appearance. Second, we develop a new generative local model that exploits the scale invariant feature transform and its spatial geometric information. To make use of the advantages of the global discriminative model and the generative local model, we incorporate them into Bayesian inference framework. In this framework, the complementary models help the tracker locate the target more accurately. Furthermore, we use different mechanisms to update global and local templates to capture appearance changes. The experimental results demonstrate that the proposed approach performs favorably against state-of-the-art methods in terms of accuracy.

In an effort to improve the tracking accuracy of the ground maneuvering target in infrared images, a method is proposed based on the strong tracking filter (STF) and the cubature Kalman filter (CKF) algorithms. In this method, the fading factor is introduced from the STF algorithm and is calculated by transforming the nonlinear measurement variance matrix to be linear approximately, and then the fading factor is used to correct the prediction error covariance matrix (PECM) of CKF, so that the gain matrix can be adjusted at real time and hence the tracking ability of the maneuvering target could be improved. After the digital simulation experiment, it is shown that, comparing with CKF and the unscented Kalman filter algorithms, the average tracking accuracy of the location is increased by more than 20% with the target velocity under 20  m/s and acceleration under 5  m/s², and it can even be increased by 50% when the target step maneuver occurs. With the tracking experiment on the real infrared tank images, it can be concluded that the target could be tracked stably by the proposed method, and the maximum tracking error is not more than 8 pixels even though the 180 deg turning takes place.

The low quality of an image is often characterized by low contrast and blurred edge details. Gradients have a direct relationship with image edge details. More specifically, the larger the gradients, the clearer the image details become. Robust image reconstruction enhancement based on Gaussian mixture model estimation is proposed here. First, image is transformed to its gradient domain, obtaining the gradient histogram. Second, the gradient histogram is estimated and extended using a Gaussian mixture model, and the predetermined function is constructed. Then, using histogram specification technology, the gradient field is enhanced with the constraint of the predetermined function. Finally, a matrix sine transform-based method is applied to reconstruct the enhanced image from the enhanced gradient field. Experimental results show that the proposed algorithm can effectively enhance different types of images such as medical image, aerial image, and visible image, providing high-quality image information for high-level processing.

Typically, the initial task of classifying indoor scenes is challenging, because the spatial layout and decoration of a scene can vary considerably. Recent efforts at classifying object relationships commonly depend on the results of scene annotation and predefined rules, making classification inflexible. Furthermore, annotation results are easily affected by external factors. Inspired by human cognition, a scene-classification framework was proposed using the empirically based annotation (EBA) and a match-over rule-based (MRB) inference system. The semantic hierarchy of images is exploited by EBA to construct rules empirically for MRB classification. The problem of scene classification is divided into low-level annotation and high-level inference from a macro perspective. Low-level annotation involves detecting the semantic hierarchy and annotating the scene with a deformable-parts model and a bag-of-visual-words model. In high-level inference, hierarchical rules are extracted to train the decision tree for classification. The categories of testing samples are generated from the parts to the whole. Compared with traditional classification strategies, the proposed semantic hierarchy and corresponding rules reduce the effect of a variable background and improve the classification performance. The proposed framework was evaluated on a popular indoor scene dataset, and the experimental results demonstrate its effectiveness.

A simple technique to avoid color limitations in image capture systems based on chroma key video composition using retroreflective screens and light-emitting diodes (LED) rings is proposed and demonstrated. The combination of an asynchronous temporal modulation onto the background illumination and simple image processing removes the usual restrictions on foreground colors in the scene. The technique removes technical constraints in stage composition, allowing its design to be purely based on artistic grounds. Since it only requires adding a very simple electronic circuit to widely used chroma keying hardware based on retroreflective screens, the technique is easily applicable to TV and filming studios.

A near-infrared (NIR) band provides information invisible to human eyes for discriminating and recognizing objects more clearly under low lighting conditions. To capture color and NIR images simultaneously, a multispectral filter array (MSFA) sensor is used. However, because lenses have different refractive indices for different wavelengths, lenses may fail to focus all rays to the same convergence. This is the reason an out-of-focus problem occurs and images are blurred. In this paper, a demosaicking algorithm that considers the out-of-focus problem is proposed. This algorithm is used by the MSFA of a red-green-blue-NIR image sensor to obtain color and NIR images. After the energies of the multispectral (MS) channels in the MSFA image are balanced to minimize aliasing, that image is filtered by the estimated low-pass kernel to generate a panchromatic (PAN) image. When an image is acquired, the out-of-focus problem and the formation process of the PAN image are modeled. The desired MS image is estimated by solving the least squares approach of the difference between the PAN and MS images based on the models. The experimental results show that the proposed algorithm performs well in estimating high-quality MS images and reduces the out-of-focus problem.

Visual tracking is very challenging due to the existence of several sources of variations, such as partial occlusion, deformation, scale variation, rotation, and background clutter. A model-free tracking method based on fusing accelerated features using fast explicit diffusion in nonlinear scale spaces (AKAZE) and KLT features is presented. First, matching-keypoints are generated by finding corresponding keypoints from the consecutive frames and the object template, then tracking-keypoints are generated using the forward–backward flow tracking method, and at last, credible keypoints are obtained by AKAZE-KLT tracking (AKT) algorithm. To avoid the instability of a statistical method, the median method is adopted to compute the object’s location, scale, and rotation in each frame. The experimental results show that the AKT algorithm has strong robustness and can achieve accurate tracking especially under conditions of partial occlusion, scale variation, rotation, and deformation. The tracking performance shows higher robustness and accuracy in a variety of datasets and the average frame rate reaches 78 fps, showing good performance in real time.

Evaluating algorithms used to assess image and video quality requires performance measures. Traditional performance measures (e.g., Pearson’s linear correlation coefficient, Spearman’s rank-order correlation coefficient, and root mean square error) compare quality predictions of algorithms to subjective mean opinion scores (mean opinion score/differential mean opinion score). We propose a subjective root-mean-square error (SRMSE) performance measure for evaluating the accuracy of algorithms used to assess image and video quality. The SRMSE performance measure takes into account dispersion between observers. The other important property of the SRMSE performance measure is its measurement scale, which is calibrated to units of the number of average observers. The results of the SRMSE performance measure indicate the extent to which the algorithm can replace the subjective experiment (as the number of observers). Furthermore, we have presented the concept of target values, which define the performance level of the ideal algorithm. We have calculated the target values for all sample sets of the CID2013, CVD2014, and LIVE multiply distorted image quality databases.The target values and MATLAB implementation of the SRMSE performance measure are available on the project page of this study.

We establish a practical regularized least-squares model with adaptive regularization for dealing with blur and mixed noise in images. This model has some advantages, such as good adaptability for edge restoration and noise suppression due to the application of a priori spatial information obtained from a polluted image. We further focus on finding an important feature of image restoration using an adaptive restoration model with different regularization parameters in polluted images. A more important observation is that the gradient of an image varies regularly from one regularization parameter to another under certain conditions. Then, a modified graduated nonconvexity approach combined with a median filter version of a spatial information indicator is proposed to seek the solution of our adaptive image-restoration model by applying variable splitting and weighted penalty techniques. Numerical experiments show that the method is robust and effective for dealing with various blur and mixed noise levels in images.

While three-dimensional (3-D) imaging systems are entering hospitals, no study to date has explored the luminance calibration needs of 3-D stereoscopic diagnostic displays and if they differ from two-dimensional (2-D) displays. Since medical display calibration incorporates the human contrast sensitivity function (CSF), we first assessed the 2-D CSF for benchmarking and then examined the impact of two image parameters on the 3-D stereoscopic CSF: (1) five depth plane (DP) positions (between DP: −171 and DP: 2853 mm), and (2) three 3-D inclinations (0 deg, 45 deg, and 60 deg around the horizontal axis of a DP). Stimuli were stereoscopic images of a vertically oriented 2-D Gabor patch at one of seven frequencies ranging from 0.4 to 10  cycles/deg. CSFs were measured for seven to nine human observers with a staircase procedure. The results indicate that the 2-D CSF model remains valid for a 3-D stereoscopic display regardless of the amount of disparity between the stereo images. We also found that the 3-D CSF at DP≠0 does not differ from the 3-D CSF at DP=0 for DPs and disparities which allow effortless binocular fusion. Therefore, the existing 2-D medical luminance calibration algorithm remains an appropriate tool for calibrating polarized stereoscopic medical displays.

Three-dimensional (3-D) simultaneous localization and mapping (SLAM) is a crucial technique for intelligent robots to navigate autonomously and execute complex tasks. It can also be applied to shape measurement, reverse engineering, and many other scientific or engineering fields. A widespread SLAM algorithm, named KinectFusion, performs well in environments with complex shapes. However, it cannot handle translation uncertainties well in highly structured scenes. This paper improves the KinectFusion algorithm and makes it competent in both structured and unstructured environments. 3-D line features are first extracted according to both color and depth data captured by Kinect sensor. Then the lines in the current data frame are matched with the lines extracted from the entire constructed world model. Finally, we fuse the distance errors of these line-pairs into the standard KinectFusion framework and estimate sensor poses using an iterative closest point-based algorithm. Comparative experiments with the KinectFusion algorithm and one state-of-the-art method in a corridor scene have been done. The experimental results demonstrate that after our improvement, the KinectFusion algorithm can also be applied to structured environments and has higher accuracy. Experiments on two open access datasets further validated our improvements.

Deblurring images captured from low-illumination conditions is a challenging task, because these images contain few useful structures for kernel estimation. However, these images usually contain some light streaks, which are beneficial for estimating the blur kernel. One of our key observations is that these light streaks can provide a good initial value for a nonconvex problem in kernel estimation. The other one is that they record the track of the blur kernel at the moment when images are taken. Therefore, we propose a new prior for kernel estimation based on light streaks in this paper. Moreover, in order to ensure the shape of the blur kernel to be similar to that of light streaks during the updating, a new method is proposed to refine the shape of light streaks. With the help of the refined shape, our kernel estimation process does not require heuristic coarse-to-fine strategy, which is widely used in image deblurring methods. Quantitative experimental results show the effectiveness of the proposed method. In addition, we also demonstrate that the proposed method can be applied to the existing deblurring methods to achieve better performance.

Image enhancement techniques are intended to improve the quality of an image without any kind of distortion or degradation. The literature is rich enough in this area, but there also exist some limitations. A technique is proposed for image enhancement by combining anisotropic diffusion with dynamic stochastic resonance in discrete wavelet transform domain. The method increases mean and variance of the image by the optimum iterations on low coefficients of images, which improves contrast and brightness, respectively, and simultaneously, edges also become sharper. It is well demonstrated by performing on various test images. Specifically, the adaptation and efficiency of the proposed technique for medical images are shown, because generally medical images appear contaminated with noise in terms of low illumination.

Convolutional autoencoders (CAEs) have been widely used as unsupervised feature extractors for high-resolution images. As a key component in CAEs, pooling is a biologically inspired operation to achieve scale and shift invariances, and the pooled representation directly affects the CAEs’ performance. Fine-grained pooling, which uses small and dense pooling regions, encodes fine-grained visual cues and enhances local characteristics. However, it tends to be sensitive to spatial rearrangements. In most previous works, pooled features were obtained by empirically modulating parameters in CAEs. We see the CAE as a whole and propose a fine-grained representation learning law to extract better fine-grained features. This representation learning law suggests two directions for improvement. First, we probabilistically evaluate the discrimination-invariance tradeoff with fine-grained granularity in the pooled feature maps, and suggest the proper filter scale in the convolutional layer and appropriate whitening parameters in preprocessing step. Second, pooling approaches are combined with the sparsity degree in pooling regions, and we propose the preferable pooling approach. Experimental results on two independent benchmark datasets demonstrate that our representation learning law could guide CAEs to extract better fine-grained features and performs better in multiclass classification task. This paper also provides guidance for selecting appropriate parameters to obtain better fine-grained representation in other convolutional neural networks.

Object recognition is considered a necessary part in many computer vision applications. Recently, sparse coding methods, based on representing a sparse feature from an image, show remarkable results on several object recognition benchmarks, but the precision obtained by these methods is not yet sufficient. Such a problem arises where there are few training images available. As such, using multiple features and multitask dictionaries appears to be crucial to achieving better results. We use multitask joint sparse representation, using dynamic coefficients to connect these sparse features. In other words, we calculate the importance of each feature for each class separately. This causes the features to be used efficiently and appropriately for each class. Thus, we use variance of features and particle swarm optimization algorithms to obtain these dynamic coefficients. Experimental results of our work on Caltech-101 and Caltech-256 databases show more accuracy compared with state-of-the art ones on the same databases.

Suppressing nonsalient patterns by smoothing the amplitude spectrum at an appropriate scale has been shown to effectively detect the visual saliency in the frequency domain. Different filter scales are required for different types of salient objects. We observe that the optimal scale for smoothing amplitude spectrum shares a specific relation with the size of the salient region. Based on this observation and the bottom-up saliency detection characterized by spectrum scale-space analysis for natural images, we propose to detect visual saliency, especially with salient objects of different sizes and locations via automatic adaptive amplitude spectrum analysis. We not only provide a new criterion for automatic optimal scale selection but also reserve the saliency maps corresponding to different salient objects with meaningful saliency information by adaptive weighted combination. The performance of quantitative and qualitative comparisons is evaluated by three different kinds of metrics on the four most widely used datasets and one up-to-date large-scale dataset. The experimental results validate that our method outperforms the existing state-of-the-art saliency models for predicting human eye fixations in terms of accuracy and robustness.

Low-rank matrix approximation and nonlocal means (NLM) are two popular techniques for image restoration. Although the basic principle for applying these two techniques is the same, i.e., similar image patches are abundant in the image, previously published related algorithms use either low-rank matrix approximation or NLM because they manipulate the information of similar patches in different ways. We propose a method for image restoration by jointly using low-rank matrix approximation and NLM in a unified minimization framework. To improve the accuracy of determining similar patches, we also propose a patch similarity measurement based on curvelet transform. Extensive experiments on image deblurring and compressive sensing image recovery validate that the proposed method achieves better results than many state-of-the-art algorithms in terms of both quantitative measures and visual perception.

Recently, global and local features have demonstrated excellent performance in image retrieval. However, there are some problems in both of them: (1) Local features particularly describe the local textures or patterns. However, similar textures may confuse these local feature extraction methods and get irrelevant retrieval results. (2) Global features delineate overall feature distributions in images, and the retrieved results often appear alike but may be irrelevant. To address problems above, we propose a fusion framework through the combination of local and global features, and thus obtain higher retrieval precision for color image retrieval. Color local Haar binary patterns (CLHBP) and the bag-of-words (BoW) of local features are exploited to capture global and local information of images. The proposed fusion framework combines the ranking results of BoW and CLHBP through a graph-based fusion method. The average retrieval precision of the proposed fusion framework is 83.6% on the Corel-1000 database, and its average precision is 9.9% and 6.4% higher than BoW and CLHBP, respectively. Extensive experiments on different databases validate the feasibility of the proposed framework.

Non-negative matrix factorization (NMF) as a popular technique for finding parts-based, linear representations of non-negative data has been successfully applied in a wide range of applications, such as feature learning, dictionary learning, and dimensionality reduction. However, both the local manifold regularization of data and the discriminative information of the available label have not been taken into account together in NMF. We propose a new semisupervised matrix decomposition method, called manifold regularized non-negative matrix factorization (MRNMF) with label information, which incorporates the manifold regularization and the label information into the NMF to improve the performance of NMF in clustering tasks. We encode the local geometrical structure of the data space by constructing a nearest neighbor graph and enhance the discriminative ability of different classes by effectively using the label information. Experimental comparisons with the state-of-the-art methods on theCOIL20, PIE, Extended Yale B, and MNIST databases demonstrate the effectiveness of MRNMF.

The latest video compression standard, high-efficiency video coding (HEVC), provides quad-tree structures of coding units (CUs) and four coding tree depths to facilitate coding efficiency. The HEVC encoder considerably increases the computational complexity to levels inappropriate for video applications of power-constrained devices. This work, therefore, proposes a complexity control method for the low-delay P-frame configuration of the HEVC encoder. The complexity control mechanism is among the group of pictures layer, frame layer, and CU layer, and each coding layer provides a distinct method for complexity allocation. Furthermore, the steps in the prediction unit encoding procedure are reordered. By allocating the complexity to each coding layer of HEVC, the proposed method can simultaneously satisfy the entire complexity constraint (ECC) for entire sequence encoding and the instant complexity constraint (ICC) for each frame during real-time encoding. Experimental results showed that as the target complexity under both the ECC and ICC was reduced to 80% and 60%, respectively, the decrease in the average Bjøntegaard delta peak signal-to-noise ratio was ∼0.1  dB with an increase of 1.9% in the Bjøntegaard delta rate, and the complexity control error was ∼4.3% under the ECC and 4.3% under the ICC.

It is well known that higher level features can represent the abstract semantics of original data. We propose a multiple scales combined deep learning network to learn a set of high-level feature representations through each stage of convolutional neural network for face recognition, which is named as multiscaled principle component analysis (PCA) Network (MS-PCANet). There are two main differences between our model and the traditional deep learning network. On the one hand, we get the prefixed filter kernels by learning the principal component of images’ patches using PCA, nonlinearly process the convolutional results by using simple binary hashing, and pool them using spatial pyramid pooling method. On the other hand, in our model, the output features of several stages are fed to the classifier. The purpose of combining feature representations from multiple stages is to provide multiscaled features to the classifier, since the features in the latter stage are more global and invariant than those in the early stage. Therefore, our MS-PCANet feature compactly encodes both holistic abstract information and local specific information. Extensive experimental results show our MS-PCANet model can efficiently extract high-level feature presentations and outperform state-of-the-art face/expression recognition methods on multiple modalities benchmark face-related datasets.

We propose a simple and fast algorithm called PatchLift for computing distances between patches (contiguous block of samples) extracted from a given one-dimensional signal. PatchLift is based on the observation that the patch distances can be efficiently computed from a matrix that is derived from the one-dimensional signal using lifting; importantly, the number of operations required to compute the patch distances using this approach does not scale with the patch length. We next demonstrate how PatchLift can be used for patch-based denoising of images corrupted with Gaussian noise. In particular, we propose a separable formulation of the classical nonlocal means (NLM) algorithm that can be implemented using PatchLift. We demonstrate that the PatchLift-based implementation of separable NLM is a few orders faster than standard NLM and is competitive with existing fast implementations of NLM. Moreover, its denoising performance is shown to be consistently superior to that of NLM and some of its variants, both in terms of peak signal-to-noise ratio/structural similarity index and visual quality.

Apart from the efficient compression, reducing the complexity of the view random access is one of the most important requirements that should be considered in multiview video coding. In order to obtain an efficient compression, both temporal and inter-view correlations are exploited in the multiview video coding schemes, introducing higher complexity in the temporal and view random access. We propose an inter-view prediction structure that aims to lower the cost of randomly accessing any picture at any position and instant, with respect to the multiview reference model JMVM and other recent relevant works. The proposed scheme is mainly based on the use of two base views (I-views) in the structure with selected positions instead of a single reference view as in the standard structures. This will, therefore, provide a direct inter-view prediction for all the remaining views and will ensure a low-delay view random access ability while maintaining a very competitive bit-rate performance with a similar video quality measured in peak signal-to-noise ratio. In addition to a new evaluation method of the random access ability, the obtained results show a significant improvement in the view random accessibility with respect to other reported works.

Table look-up operation plays a very important role during the decoding processing of context-based adaptive variable length decoding (CAVLD) in H.264/advanced video coding (AVC). However, frequent table look-up operation can result in big table memory access, and then lead to high table power consumption. Aiming to solve the problem of big table memory access of current methods, and then reduce high power consumption, a memory-efficient table look-up optimized algorithm is presented for CAVLD. The contribution of this paper lies that index search technology is introduced to reduce big memory access for table look-up, and then reduce high table power consumption. Specifically, in our schemes, we use index search technology to reduce memory access by reducing the searching and matching operations for code_word on the basis of taking advantage of the internal relationship among length of zero in code_prefix, value of code_suffix and code_lengh, thus saving the power consumption of table look-up. The experimental results show that our proposed table look-up algorithm based on index search can lower about 60% memory access consumption compared with table look-up by sequential search scheme, and then save much power consumption for CAVLD in H.264/AVC.

This paper presents a method for retargeting a pair of stereoscopic images. Previous works have leveraged seam carving and image warping methods for two-dimensional image editing to address this issue. However, they did not consider the full advantages of the properties of stereoscopic images. Our approach offers substantial performance improvements over the state-of-the-art; the key insights driving the approach are that the input image pair can be decomposed into different depth layers according to the disparity and image segmentation, and the depth cues allow us to address the problem in a three-dimensional (3-D) space domain for best preserving objects. We propose depth carving that extends seam carving in a single image to resize the stereo image pair with disparity consistency. Our method minimizes the shape distortion and preserves object boundaries by creating new occlusions. As a result, the retargeted image pair preserves the stereoscopic quality and protects the original 3-D scene structure. Experimental results demonstrate that our method outperforms the previous methods.

Image description and annotation is an active research topic in content-based image retrieval. How to utilize human visual perception is a key approach to intelligent image feature extraction and representation. This paper has proposed an image feature descriptor called the local structure co-occurrence pattern (LSCP). LSCP extracts the whole visual perception for an image by building a local binary structure, and it is represented by a color-shape co-occurrence matrix which explores the relationship of multivisual feature spaces according to visual attention mechanism. As a result, LSCP not only describes low-level visual features integrated with texture feature, color feature, and shape feature but also bridges high-level semantic comprehension. Extensive experimental results on an image retrieval task on the benchmark datasets, corel-10,000, MIT VisTex, and INRIA Holidays, have demonstrated the usefulness, effectiveness, and robustness of the proposed LSCP.

We propose an efficient Markov feature extraction method for color image splicing detection. The maximum value among the various directional difference values in the discrete cosine transform domain of three color channels is used to choose the Markov features. We show that the discriminability for slicing detection is increased through the maximization process from the point of view of the Kullback–Leibler divergence. In addition, we present a threshold expansion and Markov state decomposition algorithm. Threshold expansion reduces the information loss caused by the coefficient thresholding that is used to restrict the number of Markov features. To compensate the increased number of features due to the threshold expansion, we propose an even–odd Markov state decomposition algorithm. A fixed number of features, regardless of the difference directions, color channels and test datasets, are used in the proposed algorithm. We introduce three kinds of Markov feature vectors. The number of Markov features for splicing detection used in this paper is relatively small compared to the conventional methods, and our method does not require additional feature reduction algorithms. Through experimental simulations, we demonstrate that the proposed method achieves high performance in splicing detection.