最新刊期
卷 17 ， 期 8 ， 2012
- 摘要：The purpose of the computational image aesthetics research is to endow computer with the ability to assess the aesthetics value of images as human beings do. The results can be used in many fields, for example semantic-based image retrieval, fusing the subjective perception, image aesthetics evaluation, image aesthetics retouching, photograph aesthetics prediction, art works style analysis and man-machine interaction. Computational image aesthetics is a new interdisciplinary advanced topic with good developing prospect, while it involves different subjects, including Aesthetics, Art, Cognitive Science, Computer Science, Psychology, and so on. In this paper, the latest achievements of the computation image aesthetics research are introduced at first, then a general framework of the computational image aesthetics is proposed after the analysis and summary of methods commonly applied in this field. Additionally,we point out the exiting problems and we discuss including image aesthetics measurement, extraction of aesthetics vision features, aesthetics deduction, and also the application and future developments of image aesthetics. Furthermore, some crucial solutions are pointed out to solve the exiting problems.关键词：image aesthetics;aesthetics vision features;aesthetics measure;aesthetics classification4693|102|0更新时间：2024-05-08
- 摘要：The method of threshold selection based on two-dimensional maximal Shannon or Tsallis entropy only depends on the probability information from gray histogram of an image, and does not immediately consider the uniformity of within-cluster gray scale. The segmentation effect of the Tsallis entropy method is superior to that of the Shannon entropy method. Thus, a two-dimensional Tsallis gray entropy thresholding method based on chaotic particle swarm optimization(PSO) or decomposition is proposed. First, a one-dimensional thresholding method based on Tsallis gray entropy is given and extended to the two-dimensional case. The corresponding formulae and its recursive algorithm for threshold selection based on the two-dimensional Tsallis gray entropy are derived. Then a chaotic particle swarm optimization algorithm is used to find the optimal threshold of the two-dimensional Tsallis gray entropy method. The recursive algorithm is adopted to avoid the repetitive computation of the fitness function in an iterative procedure. As a result, the computing speed is improved greatly. Finally, the computations of threshold selection method based on two-dimensional Tsallis gray entropy are converted into two one-dimensional spaces, which further reduces the computational complexity from O() to O(). The experimental results show that, compared with the two-dimensional maximal Shannon entropy method, the two-dimensional maximal Tsallis entropy method and the two-dimensional Tsallis cross entropy method, the two methods proposed in this paper can significantly improve image segmentation performance and algorithmic running speed.关键词：image segmentation;threshold selection;two-dimensional Tsallis gray entropy;chaotic particle swarm optimization;decomposition;recursive algorithm3570|69|0更新时间：2024-05-08
- 摘要：To reduce the load imbalancing problem of the slice parallel of H.264 encoding for high-definition videos with multicore processors, we firstly, predict the encoding load of each MB in the current frame based on the temporal dependence with the previous encoded frame statistics. Then the current frame is divided into slices with the predicted encoding load, so that each processor core has the relative similar computation load for each slice. Experiments using the Tile64 multicore processor platform show that it can increase the parallel encoding speedup and efficiency by about 5% with this method, when compared with the dynamic data partition algorithm based on macroblock regions.关键词：H.264;parallel encoding;multicore processor;load balance2829|52|0更新时间：2024-05-08
- 摘要：The NSP (multi-scale decomposition method) of the traditional NSCT (non-subsampled contourlet transform) algorithm has a poor detail information capturing ability and when applied to image fusion it causes a loss of image details. In this paper, we present an improved NSCT algorithm. Different from the traditional NSCT algorithm, we adopt the non-subsampled morphological wavelet decomposition, which has a better detail capture capability, to realize a multi-scale decomposition of the source image and replacing the NSP decomposition. The source images are decomposed into four parts: low-frequency, horizontal high-frequency, vertical high-frequency, and diagonal high-frequency. Afterwards, the improved NSCT decomposition on high frequencies using the NDFB (non-subsampled directional filter) for multiple directions of decomposition is realized. Our experiments show that, compared with traditional algorithms, this algorithm has a better detail capturing ability, its image fusion effect is better under the same fusion rules, and all fusion indexes are improved. Among them, the average gradient is increased by 10%. This effective image fusion algorithm can be easily realized and widely used in multi-resolution image fusion.关键词：non-subsampled contourlet transform;morphological wavelet decomposition;image fusion;edge detection3247|40|0更新时间：2024-05-08
- 摘要：Accurate estimation of Gaussian noise level is of fundamental interest in a wide variety of vision and image processing applications as it is critical to the processing techniques that follow. In this paper, a new, effective noise level estimation method is proposed based on the study of singular values of noise-corrupted images. There are two major novel aspects of this work to address the major challenges in noise estimation: 1) the use of the tail of singular values for noise estimation to alleviate the influence of the signal on the data basis for the noise estimation process; 2) the addition of known noise to estimate the content-dependent parameter, so that the proposed scheme is adaptive to visual signal and therefore it enables wider application scope of the proposed scheme. The experiments results demonstrate that the proposed algorithm can reliably infer noise levels and shows robust behavior over a wide range of visual content and noise conditions, in comparison with the relevant existing methods.关键词：noise estimation;Gaussian noise;singular value decomposition;image processing4338|64|0更新时间：2024-05-08
- 摘要：Advantages and disadvantages of some existing image diffusion denoising models are analyzed and summarized in this paper. In theory, the physical meaning of the tensor-typed diffusion model is interpreted. A new diffusivity is put forward through the analysis of local diffusion behavior of the P-M diffusion model, developing a new improved tensor-typed diffusion model is presented. It is not easy to compare the effects of various denoising models for the subjective and objective aspects, because this needs a coupling of parameters and numerical discretization methods of every model. A unified numerical implementation algorithm of diffusion models is be given, which can be employed to compare the denoising effects of every model. The results of the numerical simulation experiments confirm that, the improved diffusion model can effectively remove image noise,and simultaneously protect edge, corners, and texture as well. Furthermore, the denoised image provides a better visual impression.关键词：diffusion denoising model;diffusivity;corner;texture4033|82|0更新时间：2024-05-08
- 摘要：In order to improve the security of steganography system, an image steganography method based on Minimizing Embedding Impact and Syndrome-Trellis Codes is proposed. First, a distortion function in discrete wavelet transform(DWT) domain was designed according to human visual system and integer lifting wavelet transform, which is mainly concerned about the influences of frequency, luminance, and texture masking factor to the cover distortion. Then, the Syndrome-Trellis Codes are combined with the distortion function to propose the steganography method,to make sure that the embedding impact on the cover minimizing and centralizing is in the un-sensitivity domain. Experiments show that the proposed approach maintains a good visual quality of the stego-image and has a high security against steganalysis in space and wavelet domain. The security capacity can be about 0.4 bits/pixel.关键词：steganography;minimizing embedding lmpact;Syndrome-Trellis codes (STCs);human visual system (HVS);integer lifting wavelet2699|48|0更新时间：2024-05-08
-
Anti-compression approach to distinguishing photographic images and photorealistic computer graphics
摘要：Based on the sparse representation of images, a new approach to distinguish photographic images and photorealistic computer graphics is proposed. The proposed approach is robust enough to compression to guarantee image authenticity forensics. The Tetrolet transformation chooses the optimal tetromino partition for each 4×4 image block in terms of the minimal L-norm criterion to protect local image geometry structure and to obtain the sparsest image representation. When observing the adaptive values c, an image is represented as a normalized histogram with 117 bins corresponding to the number of occurrences of different block covering, i.e. the features of HoC (Histogram of Covering). The experimental results demonstrate the HoC features extracted from S (saturation) are able to characterize the distinct statistical properties in the local geometry between photographic images and photorealistic computer graphics. The proposed approach is applicable to image authenticity detection and auto-classification.关键词：image sparse representation;tetrolet transform;digital image forensics;anti-compression2601|45|0更新时间：2024-05-08 - 摘要：For auto mated interpretation of star sky images of low luminance and uneven contrast, it is necessary to ensure that the images are not corrupted with clouds. In this paper,we evaluate the problem of low precision and low accuracy of cloud inspection and cloud amount estimation. It is found that there is a low probability of cloud appearance around bright stars as well as in dense fields of stars following statistical analysis of a large number of star sky samples. An adaptive threshold segmentation model of the cloud is established based on a-priori knowledge after analysis of the priori probability. The thresholds applied to different images are adaptively tuned in the present model according to the local backgrounds of stars in an image. By randomly extracting one month period of star images and analyzing their backgrounds, it is verified that the variation of the adaptive thresholds are in accordance with the tendency of a sequence of images in which the gray value of the entire image background changes. Experimental results show that the accuracy of the proposed algorithm has reached up to ninety-five percent or more, a great improvement compared to the traditional algorithm. The proposed algorithm has also been put into practical use.关键词：cloud inspection;priori-knowledge;adaptive;threshold2988|73|0更新时间：2024-05-08
- 摘要：In order to play the role of co-occurrence matrix inertia in the analysis and retrieval of image texture efficiently,a new expanded order co-occurrence matrix inertia based on interval grayscale compression is studied. One part of the grayscale information of the original image is compressed and another is retained in this integrated approach. The uncompressed grayscale information is extracted by an order expansion of the matrix. The effects of the grayscale information are used randomly. Experimental results show that the algorithm differentiaties the target types better than the conventional co-occurrence matrix based algorithms.More than 82% of the objects are differentiated correctly, and the methods appropriate distinction threshold is easier to set and faster.关键词：distinction between lump coal and gangue;image texture;interval grayscale compression;expanded order co-occurrence matrix inertia2502|36|0更新时间：2024-05-08
- 摘要：In this paper, a semi-supervised fuzzy learning algorithm based on the partitioning of the outlier feature space is presented. First, a reformative fuzzy LDA algorithm using a relaxed normalized condition is proposed to achieve the distribution information of each sample represented by a fuzzy membership degree, which is incorporated into the redefinition of the scatter matrices. Moreover, we approach the problem of parameter estimation by considering the formulation of the Hopfield neural network. Using this method, the first key step of the fuzzy classification is addressed. Second, considering the negative influences from the outlier instances, we separate the outliers from the whole feature space by means of the distribution information of each sample. The strength of the technique is that it successfully uses the improved fuzzy supervised algorithm as a feature extraction tool, while quantifying those factors that exert influence ons the outlier class assignment, by means of the fuzzy semi-supervised method. Extensive experimental studies conducted on the NUST603, ORL, XM2VTS and FERET face image databases show that the effectiveness of the proposed fuzzy integrated algorithm.关键词：feature extraction;fuzzy discriminant analysis;outlier samples;semi-supervised learning;image recognition3435|42|0更新时间：2024-05-08
- 摘要：The multiphase image segmentation is modeled as a minimization problem with characteristic functions defined by level set functions, which leads to solutions of some gradient descent equations with low computation efficiency. This is improved via the dual method or Split-Bregman method using binary labeling functions and convex relaxation, thresholding techniques. In this paper, we propose a fast direct dual method (DDM) without convex relaxation and thresholding techniques. First, we design the DDM for the two-phase Chan-Vese model, which results in a binary solution of the primal variable in analytical form and a simple iterative formulation of the dual variable by using KKT (Karush-Kuhn-Tucker) conditions. Then, it is extended to the Chan-Vese model for multiphase image segmentation. The experimental results demonstrate that the proposed method has a better performance, and is more efficient than the gradient descent method, the dual method, and the Split-Bregman method.关键词：multiphase image segmentation;Chan-Vese model;binary labeling function;convex relaxation;dual method3215|52|0更新时间：2024-05-08
- 摘要：In order to meet the real-time requirements of vision detection systems, an efficient line edge extraction method using image gradient is proposed. First, by using gradient information of the image and the property that two points define a line, line segments in the image are quickly located and scanned. Then, the scanned line segments are fitted through a method minimizing the geometric distances. Finally, a collinearity judging approach based on the projection distances of line segment endpoints is applied to all the line segments, after which the collinear line segments are linked together and refitted to obtain the final line edge features. Experimental results indicate that the time consumed by the proposed method is only about a half of the fastest improved Hough transform method . Furthermore, with the given parameters, the method is highly adaptable to different images. It can satisfy the real-time and precision requirements of the line edge feature extraction processes in various vision detection systems.关键词：computer vision;line detection;gradient;line segment;line fitting3299|90|0更新时间：2024-05-08
- 摘要：The traditional wavelet and Gabor wavelet cannot show facial features well in face recognition. In this paper, we propose a 2D dual-tree complex wavelet multi-frequency uncertainty weighted fusion for face recognition. 2D dual-tree complex wavelet multi-frequency features are used to show facial features. Weights and uncertainties are calculated to get the last facial feature by multi-frequency uncertainty weighted fusion algorithm. The weighted fusion algorithm first calculates the 2D DT-CWT multi-frequency filter images of the face, and then the uncertainty weights of the multi-frequency filters are calculated. Finally, the 2D DT-CWT multi-frequency filter is integrated into the last facial feature. At the same time, the 2D-PCA method is exploited to construct the linear subspace. The Euclidean distance based classifier is adopted for classification. Using the ORL database, the experimental results indicated that compared with the use of 2D-PCA, Wavelet, and Gabor wavelet feature extraction methods, the proposed method obtains a better recognition rate.关键词：face recognition;two-dimensional dual-tree complex wavelet transform;uncertainty;two-dimensional principal component analysis2678|65|0更新时间：2024-05-08
- 摘要：Manifold learning attempts can be used to obtain the intrinsic structure of the non-linear data, which can be used in non-linea dimensionality reduction. The general regression neural network (GRNN) is a kind of artificial neural network, which can be used in non-linear regression. In this paper, the ManiNLR method, which is based on manifold learning and nonlinear regression, is proposed for head pose estimation. ManiNLR performs manifold learning on the digital image,and then uses GRNN to map the data into the linear separable space,finally using the result to estimate the head pose. Experiments show that ManiNLR can better estimate the head pose in digital images,and has the advantages of high speed and high robustness.关键词：Manifold learning;head pose estimation;nonlinear regression;artificial neural network3810|71|0更新时间：2024-05-08
- 摘要：Curve matching plays a significant role in object recognition, target tracking and fragment reassembling. An algorithm for planar curves based on corner distance matrix and concentric circles is presented. The algorithm includes two steps, namely rough matching and exact matching. The curves are represented using corner distance matrics in the rough matching stage,and then they are matched with a sub-matrix. As for exact matching,first,the representation of the curve uses concentric circles and then measures their similarity through two curve representation sets of concentric circles. The algorithm is robust to translation,rotation,and scaling. It can be used to match block objects and reassemble the graphic. The experiment results show the effectiveness and feasibility of algorithm.关键词：curve representation;curve matching;corner distance matrix;sub-matrix;concentric circles3499|58|0更新时间：2024-05-08
- 摘要：It is important for shape processing and analysis to have a noise-robust and detail-preserving shape representation. In this paper, we propose an Elastic Quadratic Patch (EQP) model, which is extended from the basic idea of Elastic Quadratic Wire (EQW), for robustly representing three-dimensional shapes. In the model, energy function quantifying 0 and 1 discontinuity is constructed based on overlapping quadratic patches for each point and its neighborhood on the surface. This function is in quadratic form and can be easily minimized through a specific vector of quadratic surface parameters. The EQP representation, which is stable and geometry preserving, can be then obtained through a point-wise iteration. In experiments, we mainly take facial depth image as experimental data to evaluate EQP’s performance on smoothing and detail preserving. Model parameters are first analyzed under different noise levels (=1,5,10). Global and local comparisons with splines and wavelets are then presented, which demonstrate, under relatively large noise, the superiority of EQP both on quantitative SNR and qualitative visual effects.关键词：shape representation;elastic quadratic patch;geometric feature;shape smoothing;feature preserving2678|49|0更新时间：2024-05-08
- 摘要：In this paper we present a two-stage method to segment white blood cell imags by a pixel classification model that is trained online using an extreme learning machine (ELM). During the training stage, we first locate leukocyte nucleus by mean-shift algorithm in the RGB color space. Then we dilate the leukocyte nucleus until the maximum ratio of entropy and area of the nucleus region occurs. The region including the nucleus could be regarded as positive candidate region for sampling. While the other regions excluding the positive one, are regarded as negative candidate regions. A two-class ELM could be trained with the training set via learning by sampling. Different training sets produce multiple models of ELM. In the test stage, multiple models of the ELM can be integrated to classify pixels in order to extract leukocytes. The proposed algorithm does not need to change any parameter during run-time. It is very robust to various staining and to the illumination in cell imaging. Experimental results demonstrate the effectiveness of the method.关键词：white blood cell;image segmentation;Mean-shift;extreme learning machine(ELM);classifiers ensemble3005|85|0更新时间：2024-05-08
- 摘要：For remote sensing image restoration with a variety of degradation factors,we propose a Bregman iteration based image restoration algorithm for remote sensing images to eliminate the irregular sampling effect,debluring and denoising. Moreover,based on this algorithm, combined with nonlocal regularization,we propose a method to determine the nonlocal filter parameter adaptively. Using alternating minimization, we split the complex original problem into two sub problems that are easier to solve. Our experimental results show that the proposed algorithm has a faster convergence speed and better restoration results compared to other total variation and Bregman iteration based algorithms, and By adding the nonlocal regularization, it can keep the detail information better.关键词：image restoration;total variation;Bregman iteration;nonlocal regularization3275|73|0更新时间：2024-05-08
- 摘要：Automatic building extraction from remotely sensed images is affected by the mixed pixel problem that lowers the accuracy of the extracted buildings. Sub-pixel mapping is a procedure to predict the land cover maps at the sub-pixel scale, and hence reduceing the influence of the mixed pixel problem. However, the sub-pixel mapping models adopt isotropic neighborhood to calculate land cover spatial dependence for simplicity, instead of using prior spatial information of buildings, making the shapes of the resultant building inaccurate. In this paper, a novel anisotropic Markov random field based sub-pixel mapping (AMSPM)approach, which manages the spectral information of the remotely sensed image and the a priori information of buildings simultaneously, is used for extracting the buildings at the sub-pixel scale. In the proposed model, an anisotropic neighborhood that only encourages the land cover dependence that both, parallel and perpendicular to the principal axis orientation of the target building, is adoed as the prior information of a building. A QuickBird multi-spectral image and an Airborne Visible Infra-Red Imaging Spectrometer (AVIRIS)hyperspectral image are applied and our results shows the propose method can not only enhance the spatial resolution of the extracted buildings, but also preserves the edge and the corner shape of the extracted buildings. The proposed model is effective for extracting buildings at the sub-pixel scale.关键词：building extraction;sub-pixel mapping;Markov random field;anisotropic neighborhood3767|75|0更新时间：2024-05-08
- 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