最新刊期
卷 17 ， 期 9 ， 2012
- 摘要：For the constraints validity problem of the Huber-Markov random field prior model in MAP super-resolution, an adaptive threshold HMRF model is proposed in this paper to make an adaptive choice of the threshold of the Huber edge penalty function. In iteration, by analyzing the relationship between the Huber penalty function and the distribution character of the image spatial gray level variation, statistical properties of the image spatial gray level variation is used as the basis for dynamically determining the threshold. Experimental results show that the proposed method can choose the threshold of Huber penalty function adaptively for images with different edge characteristics, restrain the noise, and preserve the edges details effectively.关键词：super-resolution reconstruction;huber-Markov random field;adaptive threshold;Gaussian fitting3274|225|0更新时间：2024-05-08
- 摘要：The problem of color constancy is one of the important research directions in computer vision, aiming at recognizing the color of the target without the influence of the illumination variances in the scene. Many color constancy algorithms have been proposed in recent years. Most of them use the traditional Von Kries diagonal transformation matrix, which has bad effects to the color images taken under low luminance and high luminance.The Luminance-compensation based transformation matrix proposed in this article is based on the image formation model and optics theory. This new method compensates the image intensity according to the luminance variant of the image pixel. Results of the experiment demonstrate that the proposed method can achieve a good correction for lightness, contrast and color fidelity.关键词：color constancy;diagonal transformation matrix;luminance-compensation;color correct3899|152|0更新时间：2024-05-08
- 摘要：In order to improve the noise handling of-SVD strong method, we propose a new image denoising method based on a sparse and redundant representations model in the wavelet domain called Single Scale Low-frequency Wavelet K-SVD (SLWK-SVD). The basic idea is to follow three steps: first, use the wavelet transform on the noisy image, then employ the K-SVD algorithm on the low-frequency wavelet coefficients, and finally, replac the high-frequency wavelet coefficients by zeros. The experimental results show that compared to the K-SVD method, the proposed method is more robust to strong noise. At the given strong noise level (variance from 50 to 100), the PSNR of the denoised image improved about 0.5—1.5 dB. Meanwhile, the proposed method can overcome the problem of fluctuation of the denoised image when using the K-SVD, and improve the visual effect of the recoverde image.关键词：image denoising;wavelet transform;sparse and redundant representations;K-SVD3018|208|0更新时间：2024-05-08
- 摘要：UMHexagonS is a fast integer-pixelmotion estimation algorithm, which was adopted by the H.264 video coding standard, however, in the application of real-time scenes, there still existing some shortcomings, such as the excessive use of search points and the slow search speed. Therefore, further improvement and optimization is needed for this algorithm. Based on the UMHexagonS algorithm, a new fast motion algorithm based on motion information adaptation is proposed in this paper. Dynamical search windows are used for different-size blocks adaptively. Motion-type adaptation search strategies are adopted by the motion intensity of the current block. In addition, directional cross-shaped search and adaptive rectangle-diamond search are used according to the bias characteristics of actual motion sequences in the horizontal and vertical direction. Then an adaptive octagon search is adopted by using the direction information of motion vectors, and to carry on an adaptive hexagon search for different-size blocks. Experimental results show that the motion estimation time of the novel algorithm is reduced 13%32% and the average PSNR is increased by 0.0125 dB. In addition, both the scenes adaptive ability and real-time performance are strongly strengthened.关键词：video coding;H.264;motion estimation;motion information adaptation2587|140|0更新时间：2024-05-08
- 摘要：It proposes texture feature fusion-based two-stage clustering segmentation algorithm. First, we choose texture feature, the average of difference and color component as feature vector for segmentation. Then,at the stage of segmentation, aim to the disadvantages of Fuzzy c-means, it computes the clustering center and the number of clustering center based on histogram. Finally, we use feature vector to cluster through Fuzzy c-means. Compared with some well-known methods, the proposed method has a better segmental result.关键词：local binary pattern;fuzzy C-means;clustering segmentation;histogram2560|141|0更新时间：2024-05-08
- 摘要：In the image inpainting process, the Criminisi algorithm easily confuses texture parts for edge parts, resulting in inpainting sequence deviation which finally influences the inpaint effect. By making full use of the characteristics of image textures and edges, we introduce a differential-factor to improve the ability of the algorithm to distinguish edge parts. At the same time, the improved algorithm uses a new search method to prevent the appearance of flaws during the image inpainting process. Experiments confirm that the improved algorithm can overcome the insufficiencies of the original algorithm such as texture extension and can better maintain linear structures. The Power Signal-to-Noise Ratio of the result compared to the original algorithm has been improved about 2 dB to 3 dB, and the results have a better visual appearance.关键词：image inpainting;differential-factor;priority model;linear structure2627|144|0更新时间：2024-05-08
- 摘要：For the seam feature measurement of the tight butt joint without misalignment in laser welding, current vision sensor cannot obtain the seam gap width and seam position stable and reliable enough. In this paper, a seam measurement method based on the micro depth of field is proposed. A hight optical magnification lens, a CCD(Charge Coupled Device)/CMOS(Complementary Metal-Oxide-Semiconductor Transistor) camera and an external LED(Light-Emitting Diode) lighting unit are used to constitute the non-contact measurement system. The measurement system can be used for the tight butt joint in laser welding. The angle between the seam surface and the optical axis of the camera is about 45°. The installation method solves the problem that the depth of field and the field of view are too small to measure the tight butt joint with the multi-time magnification image measurement system. Experiments indicate that the measurement system can obtain the seam gap width and seam position stable and correctly. With this measure method, the seam surface normal vector also can be measured. The measurement accuracy for the seam gap width and seam position is 5 μm and 6 μm, respectively. Furthermore, the narrowest welding seam gap that can be measured is 0.02 mm, which fully satisfies the demand of the seam measurement in laser welding.关键词：laser welding;butt seam;vision measurement;micro depth of field;definition3099|107|0更新时间：2024-05-08
- 摘要：An early smoke detection algorithm based on codebook model and multiple features is presented in this paper. First, the foreground is obtained by using the codebook algorithm. Second, the model of color distribution and the model of shape feathers of smoke are applied to detect the suspected smoke area in the foreground. Finally, the false alarm rate is reduced effectively by using dynamic features in the diffusion process of smoke. Comparing with other algorithms using ROC curve,the experimental results show that our algorithm has good detection performance and achieves real-time requirement which is very important for many applications.关键词：smoke detection;codebook model;multiple features;ROC curve2558|134|0更新时间：2024-05-08
- 摘要：A new active contour model based on global and local image information is proposed for image segmentation. The energy functional for the proposed model consists of three terms, i.e.,the global intensity fitting term,the local intensity fitting term, and the adaptive parameter term. The global intensity fitting term incorporates global image information and the local intensity fitting term uses local contextual information. The weighting factor between the global and local intensity fitting term is adaptive by the image content.By incorporating the local and global image information into the proposed model, the images can be efficiently segmented. In addition, convex optimization is added to the new model to get the global minima. Finally, the Split-Bregman method can effectively improve the segmentation speed. Experimental results demonstrate that the proposed algorithm is robust to the choice of initialization values, can get the more accurate segmentation result,and especially is about 1.5 to 2 times faster than the C-V (Chan & Vese) model.关键词：image segmentation;C-V model;convex optimization;Split-Bregman method2550|168|0更新时间：2024-05-08
- 摘要：Since the RGB color space does not closely match the human visual perception and has no ability to describe the spatial structures, the Gaussian color model, which uses the spatial and color information in an integrated model, is used to obtain more complete image features. A color-texture approach based on the Gaussian color model and a multi-scale filter bank is introduced to classify the porcelain shard images. First, the RGB color space of the image is transformed into the Gaussian color model and then the normalized multi-scale LM filter bank is used to construct the filtered images on three channels. Afterwards, the primary feature images are found by using principal components analysis and the maximum responses of the Laplacian of Gaussian filters and Gaussian filters are separately selected. These images compose a feature image set, in which the feature parameters are extracted. Finally, a support vector machine is used to learning and classification. From experimental results, the proposed method is better than gray-based method, RGB-based method and RGB_bior 4.4 wavelet based method. It can achieve a classification accuracy of 96.7% on Outex texture database and a classification accuracy of 94.2% on porcelain shard images. This method can be used in other color texture classification tasks.关键词：image classification;porcelain shard images classification;Gaussian color model;multi-scale filter bank3176|144|0更新时间：2024-05-08
- 摘要：The wheel set is the major running component of a train. Online measurement of wheel set wear parameters is important for train safety. Acquiring a completely structured light projection image is important to wheel set profile dynamic inspection method.We develop a new image segmentation algorithm based on region growing. According to the characteristics of wheel set images acquired in different conditions and the location of their base points, a new growth standard of the algorithm is proposed. The wheel set profile images are extracted effectively through proper seed selection and growth process. By processing a series of images with severe noise and external disturbances, the overlap area ratio is more than 80%, and the error segmentation area ratio is less than 0.02%. The wheel set profile curves can be extracted effectively based on the new algorithm.关键词：wheel set measurement;image segmentation;region growing;single connected region growing2839|201|0更新时间：2024-05-08
- 摘要：To resolve the problems of large-scale data and partial overlapping in image clustering, a novel sliding window based multiple-label propagation clustering algorithm is proposed. An undirected graph is constructed in which the vertex is denoted by the image and the edge represents the relation between images weighted by the similarity computed according to the image distance. Then, community detection is performed by a multiple-label propagation based sliding window. Because a sliding window can store multiple labels, each image may obtain one or more labels. Experiments carried out on public networks and images returned by search engines show that our method can find explicit clusters with partial overlapping.关键词：image clustering;label propagation;sliding window;image distance3158|107|0更新时间：2024-05-08
- 摘要：Image matching plays an important role in the field of image processing. Based on the measure of medium truth degree, we establish the medium similarity measure and present a new grey-based image matching method that is governed by the medium similarity measure. The experiments show that the results from this method match a template better than that of the classic image matching algorithms, and it exhibits good anti-noise and anti-distortion behavior with a high matching precision and speed.关键词：image matching;medium truth degree;medium similarity;measure;grey2866|107|0更新时间：2024-05-08
- 摘要：A novel multi-focus image fusion algorithm based on image sparse representation theory and on the Human Visual System is introduced in this paper. For the new algorithm, the related images are partitioned into image patches. The learning for an adaptive dictionary is implemented by an iterated processing. All image patches are decomposed by the orthogonal matching pursuit (OMP) algorithm, and the coefficients with prominent properties are selected for image patch reconstruction. The reconstructed patches are realigned according to their partition order and the overlapped patches are averaged to get the fused image.The experiment results demonstrate that the proposed algorithm maintains the property of the state-of-the-art typical algorithm. In addition, with a noisy input, the proposed method can generate noiseless results and the objective criteria outperform the typical algorithm when noise variance increasing.关键词：image fusion;multi-focused image;sparse representation;orthogonal matching pursuit3140|123|0更新时间：2024-05-08
- 摘要：Locating people has been one of the most popular subjects in computer vision, which is the basis for human-computer interaction. Locating people and motion tracking is the inverse problem of multi-camera imaging, which is gaining videos from one or more cameras, to compute the location or motion of people in videos. In this paper, we propose two types of geometrical constraints for multi-camera approaches of locating people. The accuracy of people locating algorithms relies on the exact location in the world coordinate system during experiments, which has been a great difficulty in computer vision experiments. In this paper, we present a new people locating experimental method based on virtual data, from which we can more easily get the accurate location of people in the world coordinate system. Because of the completeness of the geometric constraints, we do not need to assume the cameras’ placement and restrict the location (occlusion) or motion of the people in the scene. Using virtual and real data in the experiments, it can be more accurate for verifying the accuracy and adaptability of the algorithm. The experimental results show that the algorithm works stable and accurate. Addition, the trajectory of the people can be acquired based on the results.关键词：geometric constraint;homography mapping;EM algorithm;virtual experiments2738|114|0更新时间：2024-05-08
- 摘要：For progressive mesh models, a method of dynamic evaluation and dynamic modification is proposed to design a reasonable model structure for deformations. The evaluation specifies the levels of detail and mesh regularity and the modification changes of the topological and geometric mesh structure. The global levels of detail for all nodes of the active vertex tree are decided by the view frustum, the surface normal, the screen-space geometric error, and the deformation contribution. Topological regularity is evaluated by geometric properties. A series of modifying operations are used to correct the topological and the geometric mesh structure, where topological modifications include a global pre-construction of a mesh model and real-time adjustments on the local mesh connectivity and geometric modifications using Laplacian smoothing to change the vertex position and to eliminate deformation distortion. Guided by the evaluation results, proper modification is selected for a specified submesh, which guarantees the real-time efficiency and adaptivity. This proposed approach has been applied on progressive mesh models to prove the performance.关键词：progressive meshes;dynamic evaluation;dynamic modification;Laplacian smoothing2610|99|0更新时间：2024-05-08
- 摘要：In image based computer aided diagnosis, higher accuracy can be achieved if the objects’ information of shape and texture can be used simultaneously. In spine MRI, intervertebral disc degeneration can be characterized by shape and texture variances. Therefore we propose a framework of computer aided diagnosis on disc degeneration as follows. First, we describe the disc degeneration by modeling non-rigid shape deformation and texture variance respectively. The deformation between the two shapes is measured by the geodesic length in the shape space. Similarly, the texture difference is measured by the Bhattacharyya distance between intensity distributions of two disc regions. Secondly, these two kinds of distances are added by weights as inputs of an unsupervised clustering machine. The clustering result is aimed at assisting doctors in distinguishing normal and degenerated intervertebral discs effectively. In experiments, 108 lumbar discs from 18 examinees are selected as experimental data. The highest accuracy of our proposed method is 92.1%, higher than the accuracy when shape or texture information is being used separately.关键词：shape;texture;computer aided diagnosis;magnetic resonance imaging;intervertebral disc degeneration2698|129|0更新时间：2024-05-08
- 摘要：Medical image analysis is helpful for doctors to diagnose diseases. However, the images usually have noise and intensity inhomogeneities, which makes it hard to obtain satisfactory results using the traditional image segmentation methods. To solve these problems, we propose a coupled model based on local image information, which can segment images while restoring the bias field. In order to obtain global optimal results accurately and quickly, we improved the coupled model to be a convex function and solved it based on the Split-Bregman method. The experimental results show that our method can reduce the effect of the noise and intensity inhomogeneities, and obtain more accurate segmentation results while estimating the bias field efficiently.2935|122|0更新时间：2024-05-08
- 摘要：Local cloud and fog can cause low contrast and poor visibility in optical remote sensing images of certain ocean regions, which hinders ship surveillance. To overcome this, a multi-scale phase spectrum is used to reconstruct the low observable image to form a saliency map in the first step. Then, a global threshold is used to extract the regions of interest (ROI), which has higher saliency. The order statistic of mean intensities from the sub regions of a circular window around each ROI is used to estimate the local threshold for target pixel segmentation. The mean saliency, shape complexity, and spatial extent are extracted from the target pixels to form a feature vector. Then a minimum distance classifier on the extracted feature vector is trained to discard the false alarms. Results on many cloudy SPOT-4 panchromatic images show the effectiveness of the proposed algorithm.关键词：ship detection;saliency map;optical remote sensing;feature extraction;minimum distance classifier3007|140|0更新时间：2024-05-08
- 摘要：SAR and optical images have large differences in imaging-mechanism and spectral characteristics. Moreover, SAR images are always severe contaminated by speckle noise. Consequently, it is very difficult to obtain satisfying results while fusing SAR and optical images. Considering the advantage of non-sampled Contourlet transform(NSCT) comparing with other multiscale decomposition methods, a method of image fusion based on pulse coupled neural networks(PCNN) and NSCT is proposed. The source images are first decomposed in the NSCT domain. Energy of log in the NSCT domain is the input to motivate PCNN and coefficients in NSCT domain with high firing frequency are selected as coefficients of the fused image. Then the final fused image is obtained by NSCT reconstruction. Experimental results demonstrate that the proposed algorithm outperforms many other algorithms in both objective criteria and visual appearance.关键词：image fusion;NSCT;PCNN;synthetic aperture radar(SAR)image;multispectral image2971|174|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