最新刊期
卷 8 ， 期 2 ， 2003
- 摘要：Nowadays volumes of image and video data are stored and transferred in the form of compressed data structure. The Discrete Cosine Transform(DCT) is one of the most widely used digital compression techniques for multimedia. Data processing such as video editing or feature extraction directly in the DCT domain other than in the pixel domain will save the time needed for the compressed data to be coded or decoded as well as the memory for buffering by processing less data. Such techniques are desirable for the in time massive data processing applications as Internet multimedia retrieval?video editing, therefore have been hotly discussed during recent years. In this paper, we present a review of the data procession techniques in the DCT domain proposed in the literatures, and try to figure out the advance direction in the future.关键词：Computer image proceesing;Summarize;DCT;Image coding;Compressed domain processing;Video editing;feature extraction3105|201|0更新时间：2024-05-08
- 摘要：In order to eliminate impulse noise and preserve the fine structure features of the image, a method of the image denoising based on monotony set operator is presented. This method, firstly, uses the threshold decomposition principle to decompose the image into a bank of level sets, and then apply a special set operator to filter the level sets, finally restructure a image with the level sets filtered. To compare with traditional medial filter and Gaussian filter, the image denoising method based on set operator has the characteristics of shape preserving and contrast invariant. In addition, Fast algorithm based on Heap data structure is presented in order to increase the efficient of the filter algorithm. According to connection, the algorithm can find maximum and minimum regions, and if the area of the regions is less than a special value, then they will be cut. So the small peaks and small vales in the image can be eliminated. Experimental results show that the efficient of the algorithm is obvious over the method used one by one threshold, and set operator has remarkable advantages in terms of eliminating impulse noise, shape preserving and contrast invariant.2208|183|0更新时间：2024-05-08
- 摘要：This paper presents a novel method for constructing a panorama from sequences of images with light changing. First, the two images, which will be composited, are decomposed into level sets. A morphological distance is defined here to match the level sets in the overlapping parts of the images. Each level set in a part is matched with the level sets in another part, so the difference in contrast is found and a transform function is obtained. The contrast of an image is adjusted to fit the contrast of the other image using the transform function and then the two images are stitched together by minimizing the intensity error between two images. Every two images are composited together using the same way and a global alignment is implemented to eliminate the visible gap or overlap between the first and last images in a sequence. This work has relaxed the constraint, which needs the same ambient light conditions for all images in the sequence, in traditional panorama mosaic and has a wide variety of applications. Finally, the performance of our method is demonstrated with an experiment.关键词：Computer image processing;panorama;level set;Image mosaic;Ambient light;Morphological distance2656|212|0更新时间：2024-05-08
- 摘要：Texture identification and segmentation of images are very important issues in image analysis. In this paper we discuss the mechanism of using Fourier transform on texture identification and then suggest a texture image segmentation algorithm based on Fourier transform. Even though the edges of texture images may have various orientations and their locations in the image may be random, for the magnitude of Fourier transform of the image, the contribution of all edges with the same orientation will be stacked up in the orientation being perpendicular to the edges. This special phenomenon is called as auto registration of the magnitude spectra. The auto registration also means a re distribution of contributions of all patterns of the texture, according to the orientations of edges and the frequency locations of patterns rather than spatial locations of them. In this paper we illustrate and theoretically analyse the auto registration property of the magnitude spectra and then propose a method to exploit this property on texture identification and image segmentation. Experimental results show several advantages of this method. It is demonstrated that Fourier transform based method is capable and promising on texture identification and segmentation and a deeper research on this subject is worthwhile doing.关键词：Computer image proceesing;Image analysis;Image segmentation;Texture identification2488|253|0更新时间：2024-05-08
- 摘要：It is very important and difficult to extract geological altered information using remote sensing data. The method of Spectral Match Filter(SMF) in this paper was developed according to Ref. . The altered mineral spectra selected from standard spectral database were thought of as reference spectral vector in 0 mean domain. The filtering detector is established owing to the match degree between very one pixel in multiple band images and reference spectral vector in 0 mean domains. The altered information are segmented using both SMF and suitable threshold processing. Both a gold deposit at Dongtianshan of Xinjiang and a porphyry copper belt at Gangdese of Xizang are selected as test areas. Some altered minerals are extracted from ETM + data using SMF. The result shows that the locations of some altered minerals or altered minerals' composition are consistent with mineral deposit or ore spot. And the result also suggests the method be in good generality.2782|207|0更新时间：2024-05-08
- 摘要：This paper mainly discussed a high resolution land cover Image classification algorithm based on Fuzzy ARTMAP Neural Network, experiment and it's evaluation. We firstly introduced the fundamental theory of Fuzzy ARTMAP Neural Network classifier. Followed is a land cover classification experiment on SPOT XS high resolution image. Three algorithms were tested: the Maximum likelihood Classification (MLC), the Back Propagation (BP) Neural Network, and the Fuzzy ARTMAP Neural Network. Individual classification result was presented. We compared these different classification results and evaluated their accuracy through manually interpreting five hundred of randomly selected sample points. Our assessment shows that Fuzzy ARTMAP has a comparably better result, with overall classification accuracy higher 17 41%, 7 32% than MLC and BP. We also analyzed some misclassification between tillage and forest classes by different classification methodologies and gave some explanations. Finally, a superiority of the Fuzzy ARTMAP Neural Network classifier on high resolution land cover classification is concluded.3239|250|0更新时间：2024-05-08
- 摘要：The main objective of the technique of image data fusion is combining or amalgamating information from multiple sensors such that the new images are more suitable for the purpose of the computer processing tasks such as segmentation, feature extraction, and object recognition. This paper proposes a multisource image fusion algorithm based on hierarchical markov random field model. The algorithm defines the fused image as the hidden labels, and the multisource image can be defined as observation of the hidden labels. The hidden labels can be extended in the quadtree manner. In order to reduce the computational burden, this paper uses a hybrid structure which combines a spatial grid of a reduced size at the coarsest level with sub tree below it, down to the finest level. The hidden labels can be estimated by a noniterative inference on sub tree with ICM algorithm defined on the top of the hybrid structure. In order to circumvent the drawback of classical MAP criterion on the hierarchical graph structures, this algorithm combines nonlinear causal Markov model defined on hierarchical graph structures with bayesian SMAP criterion. This algorithm can applied to multi source remote sensed image fusion, and can contribute to the correctness of image pixels labeling and the reduction of the multi source image volume. The contrast experiments on synthetic and satellite images indicated the advantage of this algorithm relative to the classic algorithm.关键词：Computer image processing;Multisource image fusion;Hierarchical MRF model;Sequential maximum a posteriori;Image segmentation2455|191|0更新时间：2024-05-08
- 摘要：The shot change detection is a basic and important technology in video retrieval. There are some basic algorithms, including the template matching algorithm, the histogram algorithm, etc., used to detect abrupt shot change in digital video, but they can't always attain high precision in practical detecting. A new algorithm used to eliminate the influence of the video movement on the shot change detection, named twi difference algorithm, and a new algorithm used to calculate the frame to frame difference of two images, called the pixel matching algorithm, are proposed in this paper. It is proved that the twi difference algorithms obviously improve the detecting precision in comparison with the direct frame to frame difference algorithms, and the pixel matching algorithm attains higher detecting precision than the basic frame to frame difference algorithms in being, and the pixel matching twi difference algorithm, which can realize the no missing detection for the abrupt shot change on the condition of no or few false detection, gains the best detecting precision.关键词：Computer image processing;Abrupt shot change detection;Twi difference algorithm;Pixel matching algorithm;Pixel matching twi difference algorithm3244|227|0更新时间：2024-05-08
- 摘要：In order to support content based functions in the visual standard of MPEG 4, each frame of video sequences should be represented in terms of video object planes(VOP's).Segmentation and tracking of video moving objects are used for guiding the extraction of video object plane from video sequences. Interactive video objects segmentation is a good trade off between efficiency and robustness. The proposed method comprises user interaction and automatic objects tracking of semantic video objects. First, the semantic video objects boundary can be found by using a combination of human assistance and a morphological segmentation tool, Accurate of video object boundary is effectively increased by updating contour tracking method . Second, the semantic video objects in the remaining frames are obtained using affine motion estimation and compensation of the previous semantic video objects.six parameter motion vector is estimated on the basis of estimation of two parameter translational motion vector. Experimental results demonstrate that this method is effective to segment and track video semantic objects.关键词：Computer image processing;video segmentation;Video objects;Automatic objects tracking;Interaction2845|193|0更新时间：2024-05-08
- 摘要：A novel hardware architecture of zerotree coding is presented for MPEG 4 texture coding. Under the architecture, two bottlenecks in zerotree coding are handled. The recursive scans of parent and children coefficients are avoided, and the skips of the significant coefficients and their descendents are fulfilled easily. The label coefficient is implemented in one scan by exploiting the features of MPEG 4 zerotree symbol alphabet. A ZTR address buffer is designed to simplify skipping processing of significant coefficients, and to fasten the search for the descendent coefficients of ZTR/VZTR nodes. A preprocessing unit of significant coefficients is also proposed with bit or and bit not and logic circuits, which is essential for independent coding of individual bitplane. The architecture is tested in a platform with FPGA chips. With the application of MPEA 4, the design can be applied to various equipments as an independent IP core. A parallel structure needs to implement for applications with stricter time requirement.关键词：Computer image processing;MPEG 4;Zerotree coding;Hardware architecture2376|175|0更新时间：2024-05-08
- 摘要：Shape from shading is an important research domain in the computer vision, this algorithm can be applied to fabric surface reconstruction, which is the base to evaluate fabric wrinkle degree objectively. A new algorithm of shape from shading is proposed in detail in this paper, including realization process and calculation method. According to reflection map equation, restrict qualification should be added to real image because of noise affection, the whole error expression can be gotten by using restrict qualification, for increasing iteration convergence, reflection map equation is linearization and boundary condition is taken initial value, the 3D surface shape can be gotten through iteration calculation. In detail, firstly, the synthetic image is selected to do iterate calculation, which can acquire the reconstructed image exactly, and then the real replica is imported to test this method, the precision of result is very high; this algorithm can be applied to reconstruct the shape of real fabric, the wrinkle feature parameters are proposed in this paper, which can be extracted from the 3D contour data of surface shape. The result shows that these feature values can reflect the fabric wrinkle degree from different sides; it has closely linear correlation between feature values and fabric wrinkle degree, which can be used to assess fabric wrinkle degree objectively.关键词：Computer image processing;shape from shading;Iteration calculation;Surface height2559|185|0更新时间：2024-05-08
- 摘要：This paper discusses aδfunction based algorithm for image edge detection. This paper firstly proposed regularized Shannonδfunction, which is a low-pass filter and is infinitely differentiable in frequency domain, and overcomes the shortness of Shannonδfunction, that Shannonδfunction is an ideal low-pass IIR filter and its Fourier transform is not differentiable. This paper gives formulas of regularized Shannonδfunction and its first or- der derivative both in time domain and in frequency domain, and studies the relations between regularized Shannon δfunction and its first order derivative and the parameters ofsandt. Then this paper provides two kinds of edge detection algorithm based on regularized Shannonδfunction and its first order derivative. One is D algorithm for detecting image edge in detail, the other is C algorithm for detecting image edge from noised image. D algorithm uses the first order derivative of regularized Shannonδfunction for edge detection. C algorithm uses regularized Shannonδfunction for smoothing noise and uses the first order derivative of regularized Shannonδfunction for edge detection. Finally this paper does two simulation experiments. Simulation experiments of D algorithm show that, the property of this algorithm is related to its parameters and the edge detection ability of this algorithm is better than that of Sobel algorithm and Prewitt algorithm. Simulation experiments of C algorithm show that, this algorithm is better than Sobel algorithm and Prewitt algorithm and the edge detection ability of this algorithm is the same as that of Canny algorithm. In a word, the method of this paper is an efficient edge detecting algorithm for detecting details form clean image and detecting edges from noised image.关键词：Computer image processing;Shannonδfunction;regularization;Image edge detection3287|189|0更新时间：2024-05-08
- 摘要：In order to accelerate the matching and locating speed of multiple circuit cell templates which are all similar to each other in integrate circuit (IC) micro images, a fast multiple template matching algorithm based on the uniform grid structure of IC image is proposed. First, the binary topological structure of the original image and multiple templates is decimated base on their uniform grid structure, and the decimated binary topological structure is the coarse resolution representation of the original image and templates. Second, a synthesis strategy is designed to construct a binary tree model of multiple templates' topological structure. Third, the tree model is applied to search in the coarse resolution images, and the binary tree decision is used to recognize multiple templates during the search. Finally, the matched targets in the coarse resolution image are guided to a small region of the original image. The corresponding original template is matched on that region for the true target and exact position. This algorithm is tested with IC micro images database. It shows that the proposed multiple templates binary tree model and the decimation of topological structure can highly increase the matching speed and the efficiency of the cells matching and location system. Compare to searching multiple templates sequentially on the original images, the speed up factor of binary tree model is very high.关键词：Computer image processing;Multiple template matching;Decimation of topological structure;Synthesis;Binary tree decision2767|181|0更新时间：2024-05-08
- 摘要：To distinguish different object areas, run length connectivity algorithm as well as pixel labeled method of connecting area are usually adopted in binary pictures in the image identification. The present paper points out some limitations of the pixel labeled algorithm and presents the corresponding improved one. It adopted the pixel labeled algorithm to reduce the numbers of scanned picture and adopt RLE coding method to improve efficiency of the area 's combination. This improved algorithm, which greatly optimizes the original one can be applied to most kinds of shapes of objects and label all pixels with one scanning. Furthermore, this paper indicates that some problems will come out when the improved method is applied to handling images with up fork area. The author proposes solutions to these problems. Finally it shows that the improved algorithm can be used to sum up fork points in the up fork pictures.关键词：Computer imgae processing;Connected area;Pixel label;Binary picture;Label merge;Label rectified3425|207|0更新时间：2024-05-08
- 摘要：Because the fingerprint matching algorithm based on point pattern matching is not quick enough, a new fingerprint matching algorithm is presented here. Lines matching technique is used in the algorithm for seeking pairs of reference minutiae. First the similar degree of two lines is calculated to choose a pair of reference minutiae. Then the translation and rotation parameters of two fingerprint images(the template image and the input image) are figured out basing on the coordinates of the pair of reference minutiae. Afterwards the posture of input image is adjusted relative to the template image. And then the number of pairs of matching minutiae is counted with the method of coordinate based matching. At last the fingerprint matching result is given. Through experiments, the algorithm is proved accurate, quick, and invariable to translation and rotation of images. According to fingerprint images of moderate sizes, the needs of online application can be met. The algorithm is expected to be a convenient and effective fingerprint matching technique.关键词：Computer image processing;fingerprint matching;Lines matching;Discrete line sampling;Reference minutiae3242|192|0更新时间：2024-05-08
- 摘要：Medical image elastic registration is an important subject in medical image processing. Previous work has concentrated on selecting the corresponding landmarks manually and then using thin plate spline interpolating to gain the elastic transformation. However, the landmarks extraction is always prone to error, which will influence the registration results. Localizing the landmarks manually is also difficult and time consuming. In order to solve these problems, a novel method is proposed in this paper. By smoothing the thin plate spline interpolation functions, the influence of the landmarks error can be decreased effectively. And basing on the process, a semi automatic method is used to extract the landmarks, which can simplify the selection for points. Combining these two steps, an exact, fast and robust registration approach is obtained. The approach is composed of two steps. First, it searches the contours and makes them discrete to gain the corresponding landmarks. Then by interpolating images with the smooth thin plate spline, the registration images are obtained. To validate the effect, series experiments are implemented. The experiments show that the novel method can reduce the influence of the landmarks error and gain the satisfactory registration results. In this approach, the smooth parameter is an important factor, which must be selected carefully. The principle for determining the parameter is presented in the end of this paper.关键词：Medical image;elastic registration;Thin plate spline;Smoothing process;Semi automatic points selection;robust3872|196|0更新时间：2024-05-08
- 摘要：Automatic analysis based on image content is a hotspot with bright future of medical image diagnosis technology research. Analysis of the want of skull can help doctor to diagnose. In this paper, a new method is proposed to automatic detect the want of skull based on CT image content. Region growing method, which seeds and growing rules are chosen by k means clustering dynamically, is applied for image automatic segmentation. The segmented region boundary is found by boundary tracing. The shape of the boundary is analyzed, and the circularity is taken as description parameter. Then, the rules for computer automatic diagnosis of the want of skull are reasoned by entropy function. This method is used to analyze the images from the third ventricles below layer to cerebral cortex top layer. Experimental result shows that the recognition rate is 100% for the 100 images, those are chosen from medical image database randomly and are not included in the training examples. This method integrates gray and shape feature, and isn't affected by image size and position. This research achieves high recognition rate and sets a basis for automatic analysis of brain image.2362|182|0更新时间：2024-05-08
- 摘要：A run time visibility algorithm based on major occluder for complicated scene with high density is presented. First, some classic visibility algorithm such as Beam Tree, Potential Visible Set(PVS) technique and other ideas based on shading objects are summarized in this paper. Then an improved algorithm based on real time application is presented. By creating an occluding tree dynamically through primary shading objects predefined in the scene user designed, the new algorithm can eliminate all objects and scenes in the shading area of the shading tree. The algorithm performs very high efficiency when the scene user designed is organized according to Binary Space Partitioning(BSP) tree and rendered from front to back. The algorithm apply simplified shading proxy different form shading objects to primary shading objects, and present an efficient and practical simplified method for shading objects with box type. The new improved algorithm has been applied to RTG 3D Real time Graphics Toolkits and the ability of obviously increasing rendering speed for complicated scene has been verified by many practical instances in developing procedure.关键词：Computer graphics;Visibility;Potential Visible Set(PVS);Beam tree;Binary space partitioning;Occluding tree;Bounding box2650|183|0更新时间：2024-05-08
- 摘要：Intelligent vehicle can automatic drive, so the drive fatigue can be avoided totally and drive safety can be improved markedly. The research of intelligent vehicle is important aspect of intelligent transportation system. In order to ensure reliable navigating, the navigation mark should keep clean and clear. When the navigation mark becomes blur and smudges, the correct rate of the mark recognition descends, and the navigation reliability of intelligent vehicle also descends. In order to settle the problem, the method of recognizing blur and smudge navigation lane is studied by using fuzzy neural network for JLUIV-2 vision navigation intelligent vehicle. Two fuzzy neural network models are developed. One model is made up of 5 layers, its fuzzification function is a normal distribution probability function, another model has 6 layers, and its fuzzification function isπfunction. The modified quick BP algorithm is used to train the two fuzzy neural networks. Practical recognizing experiments are made by using image of blur and smudge stripe navigation mark. The results show the two fuzzy neural networks can effectively recognize the blur and smudge lane of JLUIV-2 intelligent vehicle. In order to satisfy the real-time requirement, a 10×300 interesting area abstracted form 222×300 image is processed in navigation.关键词：Pattern recognition;fuzzy neural network;intelligent vehicle;Vision navigation2306|224|0更新时间：2024-05-08
- 摘要：It is one important aspect of RS applications in mining areas to extract mining subsiding land from RS images. The precision and efficiency of traditional methods based on spectral features is low. In order to extract mining subsiding land from RS images with high precision and efficiency, new methods and algorithms should be proposed. The integration of RS and GIS can be used to this filed, and GIS can support the RS image processing and information extraction. In this paper, according to the spectral and spatial properties of subsiding land in studying area and related knowledge, some new methods and models used to RS image processing and subsiding land extraction with the support of GIS are proposed, including layered classification supported by GIS, classification of changeable region identified by GIS, classified image post processing based on GIS and domain knowledge and direct extraction models based on GIS and domain knowledge. In those new methods, both GIS data and some spatial analysis functions are used to RS image classification and information extraction. GIS can serve as the auxiliary proof or direct information of classification, and provides check and comparison basis for the results, also it can be used as image processing platform. It proved that the methods supported by GIS could make full use of spectral features, Geo information and properties, domain and expert knowledge and other statistics data. Those new methods are more precise and effective than traditional methods, and the best precision can reach about 89%. So RS images can be used to monitor land subsidence situation effectively and dynamically.关键词：Computer information system;Geographical information system(GIS);Mining subsiding land;RS image;Domain knowledge2176|204|0更新时间：2024-05-08
- 摘要：Electrical drawing is an important engineering language to describe the structure, principle, function and etc. of electrical system and equipment. This paper proposes an XML based language, XvgML, which is designed to represent the electrical drawings. The XvgML definition includes three steps: establishing the object model, defining and naming the elements and properties, and compiling the DTD document. Two kinds of method can be applied to optimize XvgML. One is to remove the property equal to the default value which is set in DTD document, the other uses LZW algorithm to compress the XvgML document. These two methods make an obvious effect to reduce its storage and promote application performance. XvgML is composed of geometry elements and electrical elements so that it is able to model electrical systems and equipments. It is more powerful than other vector graphics languages with only geometry elements. With inherited from XML, XvgML is well formed, simple and extensible, and can be well integrated with knowledge based system. An example of electrical circuit layout Intelligent CAD has been build, it benefits from using XvgML as the intermediate language between the CAD engine and expert system. In short, XvgML will be widely used as an electrical graphic format in Web based Intelligent CAD system, Web based quoting system, or as an intersystem data exchanging medium.关键词：Programe design language;XML;Intelligent CAD;Internet;Knowledge based system;Electrical equipment2694|187|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