最新刊期
卷 9 ， 期 9 ， 2004
- 摘要：Image segmentation is a key process from image processing to image analysis, which is also a basic technique in Computer Vision. In this paper the authors first introduce the theory of the active contour. The active contour is something different from the common segmentation method. During the processing, the active contour finds the optimal value for every pixel in a small domain but also considers the relationship between different pixels as well. And as a result, the active contour can give out a smooth and continuous contour of the aim object. The basic idea of the active contour is to make the contour move to the destination with the internal and external energy. When the contour moves to the target point the total energy of the contour becomes minimum. Traditionally the external force of the active contour is given by the gradient potential energy, which has some insurmountable shortcomings, thus it cannot direct the contour to move to the destination correctly. Due to the character of the edge, the diffuse method is applied to the gradient of the edge i.e. the gradient vector flow (GVF). The GVF field maintains the merit of the gradient in the range nearby the edge but also diffuses the energy field to the slowly changed range as well, where the traditional gradient energy is very little. The GVF not only expands the effective range of the energy field but also enhances the ability to deal with the concave surface as well. The experimental results show the effectivity of the method.2834|223|0更新时间：2024-05-07
- 摘要：For implementing image segmentation fast and accurately, and by analyzing the existing watershed transformation algorithms as well as considering the idea of chain code this paper presents a chain code based watershed transformation algorithm. In this paper, the traditional concept of chain code is first expanded into point out chain code and point in chain code. And then two characters, which depict how to make watershed transformation based on chain code are proposed and discussed. In the end, the detailed description of the chain code based watershed transformation algorithm is presented, and its complexity is analyzed in detail. The new algorithm is different from the traditional watershed algorithms, and it has two main steps: the first step is to generate point out and point in chain codes of every pixel by simulating raining, and the second step is to mark the labels of regions by simulating flooding. Experiments show that the new algorithm's time and space complexity is very low. Further more, such transformation result is more helpful for the following image understanding.2495|224|0更新时间：2024-05-07
- 摘要：This paper, beginning with analyzing the characteristics of Topology adaptive Snake model, points out its weak points: sensitivity of initial position, tendency of the stability in a place, and failure of automatic Topology change. Therefore, this paper proposes a new model: the adaptive active contour model based on distance equalization. This new model, through the average distance between the vertex and the line of its neighborhood dots, improves the internal energy term in the Snake model, thus much more stabilizing the contour in the process of movement. Meanwhile, the characteristics of the contour itself determine the movement of the contour, so that the adaptability of such model has been comparatively improved. By constructing and applying the inflation force, this model could be of wide use in relatively bigger areas to capture the images. This paper also proves that the experiments of segmenting synthesis and real images, conducted under the guidance of the model, have achieved effective results.2545|203|0更新时间：2024-05-07
- 摘要：Fingerprint segmentation is an important step in fingerprint identification processing. For effective fingerprint segmentation, the properties of the imaginary of Gabor Filter are analyzed and introduced into the fingerprint segmentation processing as a new feature of the fingerprint area. By using this new feature and other two assistance features, namely the local mean value and the local variance, good segmentation results are gotten with the weighted average segmentation method, which shows the result of the imaginary of Gabor Filter is an effective feature of fingerprint images. In order to get a faster algorithm, the Gabor filter algorithm is also accelerated in special conditions, which reduces the complexity of the proposed algorithm efficiently.2596|215|0更新时间：2024-05-07
- 摘要：An algorithm of paper counting based on texture analysis is proposed in the paper, in which machine vision method is employed, for the reason that paper counting machines used widely have some defects, such as taking up a lot of room, consuming much energy, being hulking, inconveniencing the maintenance work etc. So two complementary methods for paper counting based on Image Processing technology are proposed in this paper. One is paper counting by 2D Gabor filter, the other is paper counting by 1D line by line frequency analysis, in which periodicity of distribution of intensity caused by periodic placement of paper is considered as stripe texture. The information for counting and the paper number is gained through filtering the image and retaining the stripe texture using Texture Analysis knowledge because the orderly placement of paper in the image causes the interrelated resource in frequency domain. The experiments show that paper counting by 2D Gabor filter is affected deeply by the quality of the paper image and has limited application range in despite of running faster, and paper counting by 1D line by line frequency analysis can works well and gain the accurate paper number despite low speed. Generally speaking, algorithm of paper counting by digital image processing and texture analysis is infrequent in China, so some attempts are made in the paper. The result indicates that the algorithm of paper counting has fine future in application.2495|203|0更新时间：2024-05-07
- 摘要：In aerial images, vertical line is of significant use for building extraction, which provides not only the key and reliable evidence for building existence, but also the height of buildings. In this paper, in order to extract vertical lines, a simple method for determining the vanishing point of vertical lines is first analyzed based on the knowledge about the aerial photogrammetry. It is demonstrated that, suppose the imaging plane in aerophotograph is approximately parallel with the horizontal plane, the central point of an aerial image is approximate to the vanishing point of vertical lines. This information provides a geometrical constraint for vertical line extraction in object space and makes the automatic extraction of vertical lines very straightforward and simple. Second, in order to improve the efficiency and precision of line detection and to acquire the complete description of vertical lines (e.g. line parameters, pixel positions, and length etc.), the adaptive fuzzy Hough transform, which combines the geometrical constraint in object space and gradient orientation information in image space adequately to achieve an excellent performance, is adopted to extract vertical lines automatically. Lastly, the Experimental results are given to indicate that the presented algorithm is effective and efficient for real|time applications, and provides a key evidence and clue for automatic building extraction from aerial images.2567|214|0更新时间：2024-05-07
- 摘要：The TDNLMS (two dimensional normalized least mean squared error) adaptive filter can be used as a pre whitening filter in small object detection application in digital image. The performance of the digital implementation of the adaptive filter is influenced by the finite precision effect. In this paper, the finite-precision effect of TDNLMS adaptive predictor used for small objects detection is analyzed, and a method of determining computation word length of the digital adaptive filter by experiments is presented. The relation between the word length of the adaptive predictor and the features of the environment, including the step size parameter, input data word length, statistic characteristic of image processed and the support region of the predictor, is discussed. Simulation results are consistent with the analysis. Through the method presented in this paper, a finite precision TDNLMS adaptive predictor is designed, and compared with an infinite precision predictor witch is implemented by double precision floating point numbers. Simulation results shown that the MSE (mean squared error) and the mean SNR (signal to noise ratio) gain produced by the finite precision TDNLMS adaptive predictor is very close to the infinite precision TDNLMS adaptive predictor.2336|203|0更新时间：2024-05-07
- 摘要：This paper proposes a new procedure to erase noises on raster map. Basing taxon and clustering of color map pixels, and by establishing a polar coordinate system with its origin on any noisy pixel and searching along all eight directions till road pixel or area pixel is found, of course, sometimes neither road pixel nor area pixel is found because pixels from origin to edge of map along some direction are all noisy pixels, the number of road pixels, the number of area pixels and the number of edges at all eight directions can be abtained. Letr denotes the number of road pixels,adenotes the number of area pixels andedenotes the number of edges.r,a andeare called the character parameters of noisy pixels. Then by comparingrwitha, the character of any noisy pixel can be extracted. With“equity”clustering rule, most of noisy pixels can be clustered and grouped into either road or area. Then with“partial”clustering rule, the limitrophe noisy pixels, which can not be grouped into either road or area, can also be grouped into either road or area. At last,all noisey pixels can be grouped into either road or area. This new method can completely erase the noises on raster map, so as to obtain more effective identifying with all roads and areas on a map.2094|191|0更新时间：2024-05-07
- 摘要：Chip analysis is very important for chip control. In order to get chip parameters such as chip flowing direction and chip sideward-curl radius as well, charge coupled device (CCD) is often used to detect chip images. During machine operation, there are a lot of additive noises in the chip images which are detected by CCD. In order to de-noise and get edges of chip, a method based on dual-tree complex wavelet transform threshold is described in this paper. Because 2D dual-tree complex wavelet transform produces six sub bands at each scale, each of which are strongly oriented at distinct angles, it has significant advantages over real wavelet transforms for certain image processing problems. It has improved directionality and reduced shift sensitivity, which is important in chip image processing. In addition, the original chip image isn't known generally. Wavelet threshold estimation by generalized cross validation(GCV) for chip images denoising is described in the paper. Some examples are given at end of this paper, proving that the methods can improve the ability of denoising and get better edge of chip.2643|198|0更新时间：2024-05-07
- 摘要：This algorithm, which is based on both statistical characteristics of complex wavelet energy and SVM, is proposed in order to effectively detect and track targets in image, which may cause changes, such as translation, scaling and rotation. So, the problem of image matching is transformed as that of classification. The transformation of complex wavelet that has properties of scale, shift invariant and directional selectivity effectively extract the statistical characteristics of image, such as mean, standard deviation and skew. The statistical characteristics of sample templates are input into SVM to train support vectors of SVM. Then, those statistical characteristics of any sub-image from original image are input into SVM in order to match target. This is a two-stage algorithm of coarse-to-fine. Firstly, the set of candidates is sifted by SVM. Secondly, a new optimal rule, which is nonlinear distance function, is proposed to decide the optimal matching from the candidate set. Those experimental results show that this algorithm addresses the problem of confidence level, which generally exists in traditional matching methods. This algorithm's performance is superior to those of both learning method of neural network based on RBF and gray-level correlation matching method, which compares with them. Finally, a good matching result is obtained.3044|180|0更新时间：2024-05-07
- 摘要：The problem of Point-sets matching frequently arises in the domains of computer vision and medical imaging. In this paper, a new deterministic-annealing-based point-matching algorithm is presented, in which a quadratic form is used as barrier term. The existing deterministic-annealing-based algorithms usually add barrier term of the form∑pi,jlogpi,jinto their objective functions in order to control matching matrixP' s entropy. The shortcoming is thatP' s two-way constraint conditions need to be satisfied by either using Sinkhorn method or other kinds of methods. These methods are generally iterative approximate, and in practice, have the risk of lowering point-matching' s accuracy. To tackle the problem, we propose to use the form∑pi,j(pi,j-1) as barrier term. In doing so, the objective function becomesP' s positive quadrative form, and the resultingPwhich minimizes the objective function in each step of the deterministic annealing can be exactly solved for. Simulation results demonstrate the effectiveness and robustness of the algorithm.2460|219|0更新时间：2024-05-07
- 摘要：Most of current algorithms of image fusion are based on wavelet transform and the visual characteristics of eyes of human being are not sufficiently taken into account, thus failing to get perfect fusion effect in some circumstances. In this paper, we propose a new wavelet image fusion algorithm based on human visual system. In our algorithm we fully considers the characteristics of human visual system and calculate the relative terms of neighbor region's texture and edge luminance of each pixel at each level to adaptively acquire the weighted fusing coefficients. By using the method of multi|level iterative fusion, improved fused images are thus generated. Experiments show that our algorithm can achieve quite better fusion effect and is advantageous for human eye to detect and recognize targets.2466|222|0更新时间：2024-05-07
- 摘要：There are various applications of clustering analysis techniques in the field of image retrieval. For the lack of valuable prior knowledge in the image retrieval process, unsupervised clustering algorithms should be applied. This paper proposes a new unsupervised clustering method: clustering algorithms will automatically stop according to the outlier information. This method also complements the shortages of current clustering algorithms in outlier detection and using. To show its feasibility, the paper proposes several improvements on two classical clustering algorithms, CURE and ROCK. The empirical results show that by using new method, these two algorithms can stop automatically and also achieve better performance.2493|204|0更新时间：2024-05-07
- 摘要：This paper introduce a rapid RCS prediction system emX for complex targets using physical optics,geometry optics,physical theory of diffraction and graphic hardware. This paper focused on the RCS computation of electrically large,complex and perfectly targets. The faceting approach is used to model the target's geometry. This paper uses the close form solution of physical optics and the graphic hardware to accelerate RCS computation. The results show that this method is an efficient RCS rapid prediction method.2629|176|0更新时间：2024-05-07
- 摘要：The implicit polynomial curves have a lot of merits, such as the capability to describe irregularly shaped objects, object recognition and insensitivity to noise, and are used widely in CAGD and computer graphics. A new method for closed curve construction is introduced which is based on the combination of BP neural network and principle of implicit curve construction. The algorithm, first constructs the input and output of the BP neural network from the constraint points and changes the implicit function that represents object boundary into explicit function, then uses BP neural network to fit the curve of the explicit function, and finally obtains the fitting curves that represent the object boundary from the simulation surface. The algorithm has more advancements than the method of fitting the curves of explicit functions by BP network, which can not fit the closed curves. It has good numerical stability and robustness in dealing with noisy or missing data. The Experimental results are given to verify the effectiveness of recovering incomplete images and object boundary reconstruction.2905|410|0更新时间：2024-05-07
- 摘要：In the methods of two part texture mapping and environment mapping, a sphere is usually used as the intermediary surface, where a kind of sphere mapping algorithm is contained as their important component. Furthermore, the algorithm of sphere mapping may also be applied in recognition of surfaces and generation of landscapes, etc. In order to decrease the texture distortion caused by the algorithm of sphere mapping, the paper discusses the traditional algorithms of sphere mapping and its shortcomings at first according to the non|distortion criterions of sphere texture mapping. And then, a new algorithm of texture mapping is put forward, which is applicable to partial area of a sphere. Because the ratio of areas between a texture region on the texture plane and the corresponding mapping region on the partial sphere is used as a constraint, the quality of texture mapping is obviously improved. Finally, some of the experimental results are presented, which show that the new algorithm is quite effective and has certain advantage compared to the traditional algorithms of sphere mapping.2542|174|0更新时间：2024-05-07
- 摘要：To design a securer and easier constructed watermark, the biorthogonal integer wavelets with parameter based on lifting scheme is constructed, and a Hash function is built by transmutative Rijndael code. A digital watermark is designed by integer wavelets with parameter combined with RH arithmetic operators. This watermark algorithm posed is a kind of weak algorithm. Without original image and standard watermark, it is shown to be most sensitive to distortion, that is even 1 bit of change to the picture can be sensed. And furthermore the region distorted can be easily oriented in this algorithm. Compared with the traditional digital watermark based on integer wavelets, the integer wavelets with parameter can reduce operation and increase security by changing the parameter. The integer wavelets with parameter not only simplify the calculation but also improve the security of the watermark. And the transmutative Rijndael code increases the availability ratio of code and circuit, which reduces the cost of hardware realization greatly. Lots of simulation results show that the weak watermark is sensitive to the distortion of the image. When the parameter in the wavelet is wrong, the watermark cannot be detected correctly. The weak watermark method presented in this paper has wide applications.2407|191|0更新时间：2024-05-07
- 摘要：When compressed video stream is transmitted over internet or wireless network, it is sensitive to packet loss and error occurrence due to the limited bandwidth and the instability of the channel, especially for motion compensation-based video coding standard. An error in one frame not only debases the reconstructed quality of the current frame, but propagates to the succeeding frames which use current frame as their reference frame. As a result, error concealment algorithms are developed for hiding erroneous blocks by exploiting high correlation in video sequence. In this paper, a new temporal error concealment algorithm which suits for H.264, the latest international video coding standard, is proposed. In this algorithm, an erroneous macroblock is firstly divided into four 8×8 blocks, which work as the concealment unit instead of the macroblock. Then, a modified boundary matching algorithm is exploited to recover the lost motion vectors with different boundary pixels for different blocks. Finally, the best-matched blocks are obtained using one-quarter pixel diamond search method. Experimental results show that the proposed algorithm makes use of the information in H.264-compressed bit stream efficiently, and has a better reconstruction quality for erroneous blocks than conventional temporal error concealment algorithms.2406|205|0更新时间：2024-05-07
- 摘要：Fractal image coding is a very promising compression technique, in which an image is encoded by a contractive transformation whose fixed point is close to the original image, and then is decoded by using the iteration procedure stemmed from Banach fixed-point theorem. However, it has not been widely used because of its long encoding time and high computational complexity. A fast fractal encoding algorithm is thus proposed in this paper. The proposed algorithm uses an inequality linking the root-mean-square (RMS) and mean intensity deviation to convert the range-domain block matching problem to the nearest neighbors search problem in the sense of mean deviations. In detail, after the codebook blocks are sorted according to their mean deviations of intensities, the encoder uses the bisection search method to find out the best matched codebook block regarding to mean deviations of a given range block. Because the closeness of mean intensity deviations of two blocks cannot ensure their good approximations in the RMS sense, the encoder utilizes the inequality to again search for the best-matched block (in the RMS sense) in the nearest k-neighbor of the best-matched block (in the sense of mean deviations) to a given range block in order to further improve the image quality. The experimental results demonstrate that the encoding procedure is much faster than that of the baseline fractal algorithm, while it gives an insignificant degradation in the subjective quality.2660|186|0更新时间：2024-05-07
- 摘要：A geo-spatial data model of feature-oriented object tree (FOOT) is presented. Spatial partitioning, spatial clustering and spatial indexing were emphatically discussed. Database services, application interface and application services in Four Tier Architecture are analyzed. By the spatial data organization and systematic architecture, function modules with massive spatial database management for the system are introduced. And also detailed object logic relation for the modules were explained. At last, on the base of design for the FOOT model and key technique of systematic implementation, an example of massive spatial database platform running in Linux and Oracle was given. The testing with 30 GB spatial data shows that the systematic main aims implement in feature-object spatial data organizing, integrative massive spatial data management and distributed spatial data computing. By the uploading national land use map and national county-administration map of 1∶400, the display is rapid. the efficiency of map windowing, spatial overlaying and buffer analyzing are very high, which validates the correctness, validity and rationality of FOOT.2477|201|0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