最新刊期
卷 20 ， 期 5 ， 2015
- 摘要：This is the twentieth in the annual survey series of the yearly bibliographies on image engineering in China. The purpose of this survey work is mainly to capture the up-to-date development of image engineering in China, to provide a convenient means of literature searching facility for readers working in related areas, and to supply a useful recommandation for the editors of journals and potential authors of papers. Considering the wide distribution of related publications in China, 822 references on image engineering research and technique are selected carefully from 3 103 research papers published in 148 issues of a set of 15 Chinese journals. These 15 journals are considered as important journals in which papers concerning image engineering have higher quality and are relatively concentrated. Those selected references are classified first into 5 categories (image processing, image analysis, image understanding, technique application and survey), and then into 23 specialized classes according to their main contents (same as the last 9 years). Some analysis and discussions about the statistics made on the results of classifications by journal and by category are also presented, respectively. Besides, taking this opportunity the 12 126 references on image engineering research and technique, selected from 50 478 research papers published in past 20 years, are divided into four “five-year” periods, and a comparative analysis for the selection of papers concerning image engineering as well as for the numbers of selected references belonging to each category and each class is conducted. According to the analysis on the statistics in 2014, it seems that image techniques for enhancement and restoration are still obtaining many attentions, image segmentation is yet a focus in image analysis, the numbers of references on image acquisition increases a lot, and image matching and fusion as well as their applications in remote sensing and survey mapping become a hot point again. Following the comparison of all statistics in 20 years, it can find that the numbers of references for 7 classes are almost increasing every year, 3 classes are about decreasing progressively recently, while other classes are changing with ups and downs. This work shows a general and off-the-shelf picture of the various progresses, either for depth or for width, of image engineering in China in 2014. The statistics for 20 years provides even more complete and reliable information for the tendency and development.关键词：image engineering;image processing;image analysis;image understanding;technique application;literature survey;literature statistics;literature classification;bibliometrics3716|396|10更新时间：2024-05-07
- 摘要：Superpixel methods are image pre-processing technologies that have rapidly developed in recent years. These methods can segment an image into a certain amount of semantically meaningful sub-regions. Compared with the basic element pixels in the traditional image processing methods, superpixels have better abstraction of image local features and better representations of structural information. Furthermore, superpixels can dramatically reduce the complexity of the subsequent processing. Given these significant advantages, superpixels have been widely used in computer vision, particularly in image segmentation. Considering its theoretical value, this study comprehensively reviews the existing superpixel methods and their applications in image segmentation. The history of superpixel segmentation is reviewed, and the superpixel segmentation algorithms are compared in experiments using evaluation metrics to present their performance in superpixel segmentation. Then, the applications of superpixels in image segmentation are categorized and introduced. Finally, the existing limits of the superpixel segmentation algorithms are shown, and the implicit directions for future research on superpixels are concluded. The fundamental concepts, advantages, and disadvantages of the superpixel segmentation algorithms and the applications of superpixels in image segmentation are reviewed. The limits of superpixel segmentation algorithms are presented on the basis of several experiments. As effective pre-processing technologies, superpixel methods have relatively high research value. However, the limitations of superpixels require further research. These limitations include contradictions between the amount of superpixels and the segmentation quality, the superpixel segmentations of some particular objects, and so on.关键词：superpixel;image segmentation;graph theory;region merging;evaluation metrics8174|759|45更新时间：2024-05-07
- 摘要：This paper presents a novel blind steganalytic scheme of color images on the basis of RGB space to effectively prevent steganography. The proposed scheme includes intra-channel and inter-channel features. Intra-channel features are formed by features of co-occurrence matrices from the difference planes; these features effectively capture the dependency among coefficients in any color channel. Inter-channel features are extracted in second-order difference planes between channels; these features can effectively capture the dependency between channels. During classification, the costs of each sub-classifier are optimized by the genetic algorithm. Several sub-classifiers with optimal costs are selected, and the optimal decisions are synthesized through majority voting. Experimental results show that the prediction error rate of the proposed features is 4%~5% lower than that of SPAM features, whereas the prediction error rate of the selective ensemble classifier is 1%~2% lower than that of the ensemble classifier. The proposed scheme has minimal time complexity and is applicable to low-embedding color RGB images. Furthermore, the performance of the proposed scheme outperforms state-of-the-art steganalytic schemes.关键词：information hiding;steganalysis;color RGB image;channel correlation;selective ensemble3032|526|1更新时间：2024-05-07
- 摘要：Image representation has always been a basic problem in image processing. This paper studies the approach of a digital image expressed as a complex number. The new approach has been applied to image disguise, and the result is satisfying. A one-to-one correspondence established between an image and a point in a complex plane based on complex bases under positional number system notation. The relationships between different images can be converted into a plain vector operation, which helps image study using geometric method. Thus, a new image disguise method is realized. This image disguise algorithm can not only disguise one or many images, but also allows an image to be disguised many times. Experimental results show that our method is numerically stable, can accurately restore secret images in an ideal transmission environment, and can restore secret images with some satisfaction even in a noisy transmission environment.关键词：positional number system;complex number base;digital image representation;image disguise3555|543|0更新时间：2024-05-07
- 摘要：The detection of region of interest (ROI) is the a key technique in image processing. Human visual system focus on a few objects in a complicated natural environment. These objects are called region of interest. The model of region of interest detection can simulate the human visual system and accurately compute the saliency area in image processing. This model can improve the efficiency of computer processing and reduce calculation complexity. Thus, the detection of region of interest is of great significance. A bottom-top ROI detection method is proposed based on low level image cues combined with middle level cues. First, the middle level coarse saliency region is obtained via a convex hull of corner detected by boosting Harris and superpixels clustering. The original image is then transformed from RGB color space to CIELab color space, and the difference of Gaussian filter method is presented to obtain the low level coarse saliency map. Eventually, the saliency map of the initial image is obtained by fusing the two coarse saliency maps. Extensive experiments on the large data set coming from Microsoft Asian research institution show that our method performs better than state-of-the-art algorithms. For fair evaluation, the results obtained via the five methods are based on the source codes provided by the authors. Both a subjective and objective evaluations of the proposed method compared with the other five methods are presented. The subjective comparison illustrates that our method provides accurate location, well-defined boundaries, uniform highlight, and full resolution saliency map. Moreover, the objective comparison via precision-recall curve shows that our method performs well in precision. Experiments show that this method can clearly highlight the whole salient object via reduced degrees of saliency levels, significantly alleviate the influence of false positive pixels, and obtain well-defined boundaries. In conclusion, our method can be generally exploited as an image preprocessing method.关键词：region of interest;saliency map;superpixels clustering;convex hull;difference of Gaussian filter3678|431|4更新时间：2024-05-07
- 摘要：Conventional meaningful image sharing methods have defects in pixel expansion, and distributed cover images have low visual quality. To address these problems, a (, ) meaningful image sharing scheme based on adjusting difference transformation is proposed. The size of the secret image is equal to that of each cover image, and the secret and cover images are both natural. In the sharing phase, the secret image is converted into difference and location maps by adjusting difference transformation. The (, )-threshold is used to share the respective difference and location maps and then embed them into cover images. In this step, a secret key is used to produce the embedded positions of the location map in cover images and to choose different embedding methods for various difference sharing data using the location map. Third, the (, )-threshold is used to share the secret key to the sub-secret keys. Finally, each sub key with its corresponding distributed cover image is managed by a participant, and their MD5 value is published into the third trust party. In the recovering phase, the MD5 value of each participant sub key with its cover image is checked. If less than participants pass authentication, the recovering phase fails. Otherwise, the correct sub keys are used to recover the secret key before the location map is extracted and recovered by the secret key. Then, the location map is used to extract and recover the difference map. Finally, the secret image is recovered by inverse adjusting difference transformation with the location and difference maps. Experimental results show that the proposed scheme does not cause any defects in pixel expansion in comparison with conventional methods. The distributed cover images have satisfactory visual quality, and any malicious participants can be detected. The secret image in the proposed scheme has the same size as the cover image, and both secret and cover images are natural. The distributed cover images have better visual quality with a strong malicious participant authentication capability by using MD5 as an authentication code.关键词：meaningful image sharing;MD5;(;,;)-threshold scheme;adjusting difference transformation3744|237|3更新时间：2024-05-07
- 摘要：Image enhancement is very important in various visual signal processing applications. In many applications, such as photo retouching, visual inspection, and machine analysis, image enhancement methods are proposed to obtain images with better visual quality. In these cases, original images are usually not “perfect.” However, existing full-reference image quality assessment methods use “perfect” original images as reference signals to assess image quality. Therefore, existing full-reference image quality assessment methods cannot be used to evaluate the visual quality of enhanced images. In this paper, a novel visual quality assessment metric based on features of gradient, colorfulness, and luminance is proposed for enhanced chromatic images. The human visual system (HVS) is highly sensitive to gradient information, which can effectively capture both contrast and structural/texture information. Thus, in the proposed metric, a gradient enhancement map is calculated by estimating the enhancement degree of the enhanced image compared with its reference image. In addition, colorfulness is the attribute of the perceived color in certain regions appearing to be more or less chromatic. In the proposed metric, the colorfulness of an enhanced image is estimated by two factors, namely, one is the average distance from different colors to the center gray and the distance between individual colors in the image. Consequently, a colorfulness enhancement map is computed by calculating the enhancement extent of color saturation and its standard deviation. Meanwhile, luminance enhancement factor is integrated together based on the analysis that the luminance change would influence the appearance of gradient and colorfulness information. Moreover, the gradient and colorfulness features of the reference images are extracted to build the objective quality assessment metric for enhanced images. Finally, the model of the relationship between the luminance enhancement factor and the gradient/colorfulness enhancement map is built. The proposed metric is compared with the existing image quality assessment metrics, including the peak signal to noise ratio (PSNR), structural similarity (SSIM), visual information fidelity (VIF), most apparent distortion (MAD), appearance-based MAD (MADa), and augmented MADa (dxMADa). Three evaluation criteria are used for performance evaluation, namely, (1) Pearson linear correlation coefficient (PLCC), (2) Spearman's rank-order correlation coefficient (SROCC), and (3) root-mean-squared error (RMSE). Generally, a good image quality assessment (IQA) metric has high PLCC and SROCC values and a low RMSE value. In our proposed metric, compared with the best available metric for enhanced images, PLCC and SROCC improved 2.9% and 2.5%, respectively. Moreover, RMSE reduced 12.3%. In sum, the proposed metric is an obvious improvement than existing metrics when assessing enhanced images. In the proposed metric, gradient and colorfulness enhancement maps can accurately calculate the enhancement extent. By integrating the luminance enhancement estimation and features in the reference image, the proposed metric can perform better than other existing metrics. The proposed metric provides an objective score for enhanced images and solves two problems, namely, the reference images of enhanced images are not “perfect” images and similarity measure algorithms cannot be used well for enhanced images. We have conducted experiments to demonstrate the performance of the proposed metric for enhanced images. We have conducted the experiments to demons trate the performance of the proposed metric for enhanced images.关键词：image quality assessment;enhanced image;gradient enhancement map;colorfulness enhancement map;luminance enhancement factor3917|455|3更新时间：2024-05-07
- 摘要：Majority of traditional contour tracking methods only consider the overall characteristics or significant features of the moving target under a complex background, which figure out contour tracking without fully utilizing the moving target's locally feature information. When the moving target is occluded, most traditional tracking methods make these moving target easily drift, which sometimes result in the loss of the moving target. Focusing on these problems,tracking algorithm based on locally model matching of geometric active contour(LM-GAC) is proposed. Super-pixels make these similar color characteristics of pixels in the image as a class; thus, a plurality of pixels is composed of super-pixels. Super-pixels divide the moving target into a plurality of pixel blocks. The super-pixel is combined with the EMD (earth mover's distance) similarity measure to build locally feature model. Carrying on locally model matching, a noise model is then introduced to estimate the local model parameter , which can enhance the adaptiveness of the features model and the accuracy of the locally model matching. Finally, the level set segmentation method is combined with particle filter to extract the moving target contours to track moving target contours accurately. Compared with other moving target contour tracking methods, the proposed moving target tracking method maintains a higher success rate on image sequences that were under the conditions of partial occlusion, target deformation, illumination changes, and complex background. The proposed moving target tracking method, which has an average success rate reaching 79.6%, is relatively accurate and stable. Experiment results indicate that the proposed moving target tracking algorithm can modify noise model parameters and particles heavy adaptively in real timedepending on the image sequence, so the proposed moving target tracking algorithm has higher accuracy and robustness. Under complex backgrounds, the proposed moving target tracking algorithm can track the moving target contour more accurately.关键词：local model;super-pixel;EMD similarity measure;noise model;level set3353|525|3更新时间：2024-05-07
- 摘要：Most mean-shift based object tracking algorithms neglected information on the spatial distribution of dense features. This study uses dense features to enhance the reliability of tracking. Some color features gather on tracking objects, and each feature forms a region of certain size. These dense feature regions play an important role in human vision. Information regarding the spatial structures of these dense feature regions can be used in object tracking. An effective and efficient tracking object model is presented. Intensive features are found, and the areas and distances between the dense region centroids and the target object center are calculated to obtain the weight of each feature, which is applied to describe the tracked object. The intensive features in the target model are heavier than the discrete features. Simultaneously, the zero-order moment and the similarity coefficient between the target model and candidate models are used to estimate the target area. Subsequently, an area compensation method is used to compensate the object areas that are weakened by background weighting. Finally, the estimated area and the second-order center moment are used to adaptively estimate the object scale and direction. The object models are updated when the background shows significant changes. Experimental results show that the proposed method can adapt well to the object scale changes, with an average tracking accuracy of >94.6%. In addition, the proposed method has higher accuracy, efficiency, and robustness than some state-of-the-art methods. The proposed method increases the weight difference between different features in the target model. This method is efficient indistinguishing the target from the background. The area compensation method solves the problem wherein the estimated target area is less than the actual area because of the weakened target feature weight.关键词：object tracking;scale-adaptive;weighted dense features;Mean-shift3168|392|0更新时间：2024-05-07
- 摘要：Image segmentation is a critical step in image processing. Several algorithms based on statistics have been proposed, in which the statistical image model must be built under a certain assumption on the image. For example, the commonly used statistical model on pixel intensities includes normal distribution and gamma distribution (especially for SAR intensities image). Although optimal segmentation results could be obtained through most algorithms, statistical models are an approximation of pixel intensities and could not accurately describe the characteristics. Moreover, building an accurate image model, especially for remote sensing images, is difficult because of the complexity and uncertainty of spectral characteristics of objects on the earth's surface. Kolmogorov-Smimov statistic (K-S distance) defines the similarity by measuring the maximum distance of two statistical distributions. In this case, building a statistical model for an image is not necessary. By contrast, grayscale histogram could be used to describe the distribution of two classes for image segmentation tasks. K-S distance solves the difficulty in building an accurate statistic distribution model for an image. To date, K-S distance image processing is based only on pixel scale. Given that histogram is not sensitive when only a pixel changes its class, K-S distance based segmentation could not be used. In this paper, region and K-S distance based image segmentation was proposed. Voronoi tessellation was used to partition image domain into sub-regions (Voronoi polygons) corresponding to the components of homogenous regions. Each Voronoi polygon was assigned a random variable as label to indicate the homogenous region to which it belongs. All labels for the Voronoi polygons formed a label field. The intensity histogram of each homogenous region was then calculated, and the dissimilarity between two homogenous regions was determined by the K-S distance on the two histograms corresponding to the two regions. Thereafter, the potential energy function of the dissimilarity was constructed. Employing Bayesian inference, a posterior distribution was obtained using the likelihood constructed by non-constrained Gibbs expression. Finally, Metropolis-Hastings (M-H) scheme included updating labels, moving generation points, and birth and death generation points operations designed to simulate the posterior. The optimal segmentation was obtained by Maximum A Posterior (MAP) estimation. Using the proposed algorithm, segmentation was performed on simulated and synthesized images, as well as real optical and SAR images. Qualitative and quantitative accuracy evaluations were carried out to assess the effectiveness of the proposed algorithm. In addition, results from both proposed algorithm and pixel and statistic based segmentation algorithm are compared and show that the proposed algorithm performed significantly better. The analysis on the regionalized image segmentation algorithm based on K-S statistics does not need to build an image model and could be viewed as regional based algorithm to avoid the effect of image noise during segmentation. To improve the accuracy of fitting homogeneous regions with partitioned sub-regions, different geometry tessellation methods must be considered to partition the image domain. Furthermore, the proposed methodology will be developed for image segmentation with variable classes.关键词：voronoi tessellation;kolmogorov-smirnov (K-S) statistics;metropolis-hastings(M-H) algorithm;image segmentation2529|373|2更新时间：2024-05-07
- 摘要：Given the notorious semantic gap between low level features and high level concepts in image retrieval, automatic image annotation has become a crucial issue. To bridge the semantic gap, this paper proposes a hybrid generative/discriminative approach to annotate images automatically. In the generative learning stage, images are modeled by continuous probabilistic latent semantic analysis model. As a result, we can obtain the corresponding model parameters and the topic distribution of each image. If this topic distribution is taken as an intermediate representation of each image, the image auto-annotation problem could be transformed into a multi-label classification problem. In the discriminative learning stage, we construct ensembles of classifier chains by learning these intermediate representations. At the same time, the contextual information of the annotation words can be integrated into the classifier chains. Therefore, this approach could achieve higher annotation accuracy and better retrieval performance. Experiments on two baseline datasets indicate that the average precision and recall of our approach attained 0.28 and 0.32, respectively, on Corel5k dataset. In addition, these two measures of our approach attained 0.29 and 0.18, respectively, on IAPR-TC12 dataset. The experimental results proved that our approach performed better than most state-of-the-art approaches on many evaluation measures. Furthermore, the precision-recall curve showed the superior performance of our approach over several typical and representative approaches. On the basis of hybrid learning strategy, this paper presents an image auto-annotation approach, which integrates the advantages of the generative and discriminative models. As a result, the approach exhibits better, more effective, and more robust semantic image retrieval. The methods and techniques of this paper are not only usable in the fields of image retrieval and recognition, but they can play an important role in the fields of cross-media retrieval and data mining after an appropriate adaption.关键词：automatic image annotation;probabilistic latent semantic analysis;multi-label learning;classifier chain;image retrieval3328|433|5更新时间：2024-05-07
- 摘要：To overcome the crucial problem of expression, illumination, and pose variations in 2D face recognition, a fused residual algorithm based on collaborative representation is proposed. Collaborative representation classification (CRC) algorithm combines all training images to constitute a dictionary collaboratively. CRC algorithm has less complexity by using regularized least square to solve sparse coefficients, and the coefficients reconstruct a testing face. Testing faces are classified correctly based on reconstruction residuals. This approach extracts Gabor and geodesic features from 3D face depth images, and then fuses two features via collaborative representation algorithm. The fused residuals serve as the ultimate difference metric. Finally, the minimum fused residual corresponds to the correct subject. While the Gabor feature has good scale and orientation selectivity, the geodesic feature has facial intrinsic geometric structure and robust to facial expression, CRC is insensitive to occlusion. Experiments are conducted on CIS and Texas face databases, and results show that the recognitions rates of the proposed method are up to 94.545% and 99.286%, respectively. The recognitions rates outperform that of Gabor-CRC by approximately 10%. Comprehensive experiments on CIS and Texas database verify that the proposed algorithm is effective and robust.关键词：collaborative representation;gabor feature;geodesic feature;fused residual;face recognition;3D face depth image;feature selection3106|338|1更新时间：2024-05-07
- 摘要：This study aims to address the challenging problem of recovering the 3D depth of a scene from a single image. Most current approaches for depth recovery from a single defocused image model the point spread function as a 2D Gaussian function. However, these methods are influenced by noise, and a high quality of recovery is difficult to achieve. Unlike the previous depth calculations from defocus methods, we propose an approach to estimate the amount of spatially varying defocus blurs at the locations of image edges on the basis of a Cauchy distribution point-spread function model. The input defocused image is reblurred twice with two respective Cauchy distribution kernels. The amount of defocus blur at edge locations can be obtained from the ratio between the gradients of the two re-blurred images and the two scale parameters of Cauchy distribution. A full depth map is recovered by propagating the blur amount at edge locations to the entire image via matting interpolation. The original “Lenna” image and a rotated noise “Lenna” image are used, and a Gaussian noise is used to simulate the image noise and edge position error. Then, the average error of the Cauchy gradient ratio is compared with that of a Gaussian gradient ratio. Various real-scene image data are also used to compare our depth recovery results with those of existing methods. Experimental results show that the average error of the Cauchy gradient ratio is less than that of a Gaussian gradient ratio. Experimental results on several real images demonstrate the effectiveness of our method in estimating the defocus map from a single defocused image. Our method is robust to image noise, inaccurate edge location, and interference of neighboring edges. The proposed method can generate more accurate scene depth maps as compared with most existing methods based on a Gaussian model. Our results also demonstrate that a non-Gaussian model for DSF is feasible.关键词：image processing;depth recovery;depth estimation;depth from defocus;defocus blur;Gaussian gradient;Cauchy distribution3896|378|2更新时间：2024-05-07
- 摘要：As a social platform for sharing information,amicro-blog generates millions of information daily. Thus, searching for wanted information is difficult. Tosolve this problem,this paper proposes a novel user-driven visual microblog search method. The method usesfeature words to model the user's interest and to obtain the relationship between words and users based on the model. To search a relevant micro-blog, the method first locates micro-bloggers who arerelevant to the keywords and then filters micro-blogs. The method extends the query via anattention-diverting algorithm. Finally, the relevant users and feature words are visualized on the basis oftheir degree of relevance. An interactive interface is provided to enable users to search micro-blogs that are relevant to the feature words or users that they are interested in. We show through a case studythat users can efficiently find specific information by observing the visualization results and further interacting with the visualization interface. We provide an efficient microblog search method that regardsmicro bloggers as a bridge to search for relevant microblogs and narrow down the search range. The proposed method also uses the attention-diverting algorithm to extend the query before the result is finally visualized. The proposed method is efficient and useful for users to find specific microblogs.关键词：visual search;query expansion;microblog search;attention diverting2894|358|2更新时间：2024-05-07
- 摘要：The problem of easily falling into “premature” and local optimum values is encountered during endmember extraction via discrete particle swarm optimization (DPSO). Thus, this study introduced the shuffled frog leaping algorithm (SFLA) into DPSO. When SFLA-DPSO (DPSO based on SFLA) was run, the particle swarm was divided into several groups. DPSO was then used for an in-depth search of endmembers, and information was exchanged between groups. SFLA-DPSO combined the universality and parallelism of SFLA with the fast convergence of DPSO. The combined method can prevent the particle swarm from falling in the local optimum. SFLA-DPSO, DPSO, and SMACC were applied to extract the endmembers of a hyperspectral RS image in Pulang, Yunnan Province. Then, the probability of global convergence was calculatedby using SFLA-DPSO, DPSO, and N-FINDR to extract the endmembers in the respective search space of the Hyperion and AVIRIS hyperspectral images. SFLA-DPSO had higher precision than the other tested methods. The global convergence probability of SFLA-DPSO reached 100% as the number of iterations increased, whereas that ofDPSO was only between 50% and 65%. SFLA-DPSO showed a much stronger ability to overcome local convergence than DPSO.关键词：discrete particle swarm optimization;huffled frog leaping algorithm;endmember extraction;hyperspectral remote sensing;local convergence2630|264|3更新时间：2024-05-07
- Postal code：100190
- Tel：010-58887035/58887030/58887418 Email：jig@aircas.ac.cn
- Technical support is provided by Beijing Founder electronics co., LTD 京ICP备05080539号-4
京公网安备11010802024621 - It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
- Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
0