最新刊期
卷 6 ， 期 10 ， 2001
- 摘要：The goal of content-based image retrieval(CBIR) is uppermost to reduce the semantic gap between the simple visual features and the abundant semantics delivered by an image, and a critical point in the advancement of content-based retrieval is the image semantic modeling and extraction. This paper reviews the state of the art of image retrieval using semantics mainly focusing upon two aspects: image semantic modeling and image semantic extraction. In the paper, image semantic model is generalized firstly as three main components: image semantic knowledge, image semantic hierarchical model and semantic extraction hierarchical model. Then some typical methods of semantics extraction of are analyzed in detail by classifying into five classes: semantic by user interaction, user query as visual semantic template, objects and their layout recognition, scene and event semantics extraction, emotion semantics extraction, and some limitation of them are pointed out. Finally, three critical problems including object modeling and recognition, semantic knowledge base and user retrieval model faced in image semantic processing are explained, and some resolved strategies are presented elementarily.2715|280|0更新时间：2024-05-08
- 摘要：Shape from shading(SFS) is one of the critical techniques to shape recovery in computer vision,which obtains 3-D shape of the visible surface of an object from only one image of it using the shading knowledge in the given picture. In order to give an outline of over 30 years' research work on SFS problems and try to make sense to beginners of advantages and disadvantages of varions methods to solve such problems, this paper adopted the common classification of all SFS methods presented up to now, namely, minimizaiton methods, propagation methods,localization methods, and linearization methods,to each of which some typical algorithms were analyzed both from principles and experiments point of view. Comparisons between and evaluations of these methods together with their corresponding algorithms were also given in several aspects,such as the uniqueness of the recovered surface,the approximation ability to the true surface,the effectiveness and applicability of the algorithm,etc.Through the discussion, we agreed that there is no method applicable to all kinds of SFS problems, and each method has its own range of applicability. In the end,the paper concluded in the unresolved problems to SFS as well as some indications to future work.7347|324|0更新时间：2024-05-08
- 摘要：Since the agent is autonomous, cooperative and distributed, and it is based on the BDI model of the agent, we present a multi-agent frame system(MAFS) based on Internet/Intranet for the distributed processing of the object recognition. The content includes the architecture of each intelligent agent and cooperative agent, the knowledge language for MAFS, the cooperation and communication between agents, the knowledge expression and the agent calcultion. In the same while, put forward a method of image drawing segmentation for bridge object recognition and the distributed processing thought of image recognition under the network environment. We also present the model of problem solving for bridge recognition on MAFS, revise the model according to the believable degree, and realize the bridge object recognition finally. The result shows that the method of the distributed object recognition based on MAFS is efficient and useful, and it can be easily used and realized in many other distributed processing fields, thus we can achieve the distributed object processing in the high-speed network system.3831|499|0更新时间：2024-05-08
- 摘要：With increased computing power, electronic storage capacity, and the rapid development of the Internet, the potential for large digital image libraries is now growing at an astonishing pace. Providing efficient access to images, however, is not an easy task. To overcome this difficulty, content-based image retrieval(CBIR) was thus proposed as a solution. In CBIR, images would be indexed by their own visual content, instead of being manually annotated by text-based keyword s. In this framework, many research studies have been performed, and many commercial and academic CBIR systems, widely applied in multimedia databases, digital libraries, medical image management, public security department and satellite image management, have been developed in the past few years. Most CBIR systems, however, answer users' query by similarity match based on multi-dimensional physical image features. Because of human subjectivity, different persons or the same person under different circumstances may perceive the same visual content differently. To address the difficulties arising from human subjectivity, we propose a content-based interactive emotional image retrieval approach. Through interactive evolution computation, human intuition and emotion is integrated into the evolution process to realize on-line retrieval by human-computer interaction. To deal with the problem that the user may tend to be tired arising from that the user has to evaluate a large number of individuals when the evolution time is too long, neural networks are used for off-line learning to alleviate human fatigue. Based on image content, an emotional image retrieval system has been realized. The experimental results demonstrate the effectiveness of our approach.2829|294|0更新时间：2024-05-08
- 摘要：Edge is an important feature of images. There are many ways to detect the edge of animage. In this paper, the cellular neural network is proposed for edge detection. Cellular neural network is a large scale nonlinear analog circuit suitable for real-time signal and image processing. The key problem is to find a set of parameters for the network. The high-pass filter is utilized to design the parameters of cellular neural network for detecting the binary images. A gray-scale image can be divided into 2 binary planes with different gray level. The edge of gray-scale images then can be detected through synthesizing the edge of each binary plane. Finally, the edge detection result of CNN is compared with that of Sobel and Log algorithms It can be seen from the simulation results that the proposed method is effective. Besides, because the cellular neural networks can use high-speed parallel computation and is easy to be implemented in hardware, therefore it has more potential in real-time image processing.2338|281|0更新时间：2024-05-08
- 摘要：Wool fibers play a very important role in the clothing industry. Wool fibers mainly include two types:wool and cashmere. Due to different property, they have widely different prices. However,it is always a challenging task to differentiate and recognize wool and cashmere. This paper presents an automatic recognition scheme for the wool fiber images by the electron microscope. At first the wool fibers are segmented from the background by a global thresholding method. Using the dynamic clustering method, the boundary lines of each wool fiber in the image are detected. According to these lines, different wool fibers are divided apart. Then we use Canny's algorithm to detect the edges of each wool fiber and do the post-processing. Using the character of the scales on the surface of the wool fiber, the features of the wool fiber such as the fineness and the length of the scale on the edge images are extracted. Owing to these feature parameters, we finally recognize whether a wool fiber is wool or cashmere in terms of the Bayes DecisionRule. Experiments demonstrate that the system works quickly and effectively, and has remarkable advantages in comparison with the previous systems.2642|283|0更新时间：2024-05-08
- 摘要：Using OCR tools to transform scanned document images into editable text files is a important way in printed documents processing, such as those in text retrieving applications and digital library projects. Nevertheless, the OCR systems that we generally employed can not work correctly and efficiently with document images having deflections. Trying to manipulate this deflection correction procedure automatically, We study the properties of the image' s schema histogram and it' s correlation series. The result shows that under a small angle of deflection (less than 8°),the horizontal correlation series varies negative exponentially with the angle of deflection. For this we construct a scheme that can adjust the deflection automatically depend on the image' s histogram pattern.To do this, we first choose a non-deflected sample image from the image set to find its correlation series which is in turn used to construct the negative exponential function. This experiential function can be used to determine the deflection angles of the whole set of document image. Practically, this method has shown very good performance in automatic deflection correction.2792|265|0更新时间：2024-05-08
- 摘要：Based on analysis for computer digital imaging processing, a technique described, for calibrating the digital imaging system and reducing the error created by the difference between the system model and true system. In our calibration all kind of factors involved in the processing are considered, including intrinsic factors and extrinsic factors, known factors and unknown factors. In the calibration list there are the linear geometric project parameters and nonlinear distortion parameters of the optical imaging processing, the parameters of the optics-electric transfer processing, the parameters of the image signal D/A and A/D transfer, etc. An effective method of error compensation for 3D data accompanied by calibration processing is also proposed in this paper. The error compensation model and its parameters are created in the calibration processing and applied in the 3D computer vision processing. Both above techniques have been used in 3D computer vision system and achieved high accuracy of 3D data.2598|269|0更新时间：2024-05-08
- 摘要：Texture analysis plays an important role in image analysis and pattern recognition, and texture description is a basic research topic of texture analysis. Though texture has not a uniform definition, it is acknowledged that neighborhood properties represent the dominant of a texture. Otherwise the distribution and spatial dependencies of local gray tones are periodic to homogeneous texture. So this local information can be used to describe a texture primitive, and it will work well for texture discrimination. A new texture descriptor is proposed based on the 8-neighbour gray tone spatial dependencies by using 8-neighbour Fourier series. The 8-neighbour of the point are treated as a periodic series, and their Fourier series are computed. Then the local Fourier coefficients map is generated from these local Fourier series of the whole image. A histogram of local Fourier series of the texture image was extracted by quantizing these Fourier coefficients as a texture descriptor. Because the quantization uses only the magnitudes of Fourier coefficients, this descriptor is shift-invariant and rotation-invariant. Experiments on 13 samples of Brodatz textures demonstrate the efficiency and effect of texture classification.2646|310|0更新时间：2024-05-08
- 摘要：Image interpolation methods used to improve medical image quality on display device or in the field of lossy image compression wherein some pixels. Therefore, image interpolation methods have occupied a peculiar position in medical image processing and many interpolation methods have been presented. But the image processed by the conventional interpolation methods (such as linear interpolation, double-linear interpolation etc.) have lost the image texture and caused smoothness effect. To avoid these effects, many scholars look for other interpolation methods. Through survey the greylevel of the natural image,paper has proved the greylevel of the natural image meet the isotropic fractional brownian of random(FBR). In this paper, we present processing head CT by fractal interpolation. And we have given the method to compute the correlative parameter. To evaluate the processed image quality, we introduce fuzzy degree and fuzzy entropy from fuzzy mathematics. According to the property of fuzzy degree and fuzzy entropy, the high quality image has less fuzzy degree and fuzzy entropy than the low quality image. Simulation results show that this new method has better performance than the other interpolation methods.2785|321|0更新时间：2024-05-08
- 摘要：Determination of degree of osteoporosis from spectrum feature analysis of bone radiographs will provide a screening tool to identify individuals with high fracture risks, and also an objective way to monitor treatment effectiveness reliably. This article shows that since the anisotropic trabecular appear weakly in bone X-ray images and contribute a small portion to the power spectrum, and cause that the signal(trabecular) power spectrum is masked by the noise and the MTF(modulation transfer function) of the imaging system, finding spectral characteristics of trabecular patterns may be considered as an inverse problem. It is assumed that the MTF of the imaging system and the noise introduced by film digitizer are isotropic. The isotropic parts are summarized as the synthetic MTF. An isotropic function have the same projection across all angles. The power spectrum of trabecular patterns can be approximated by reconstructing the 2-D power spectrum from the projection power spectra divided by the synthetic MTF based on Filtered Backprojection Algorithm and Fourier Slice Theorem. Experiments on the bone X-ray images and the synthetic images demonstrate that the anisotropic trabecular pattern have distinguished peaks in the reconstructed power spectrum. Once the signal power spectrum is obtained, it is staightfoward to find maximum peak to estimate the trabecular spacing and the trabecular angle and other texture features to describe the degree of anisotropy. These features are directly related to bone structure, therefore, can be used to produce a summary index to predict the fracture risk.2496|274|0更新时间：2024-05-08
- 摘要：In order to achieve fast printed characters recognition under low price hardware implementation, a fast algorithm based on multi-stage classification for Machine Printed Character Recognition is proposed in this paper. From preprocessing, feature extraction to pattern classification, the proposed method has reasonably improved according to traditional ones. Firstly, 36×36 matrix is applied for character normalization, rather than the traditional 48×48 matrix, to reduce the computation complexity for feature matching and spatial requirements for dictionary storage; Secondly, an improved coarse periphery feature with overlapping division is introduced to strengthen the stability of recognition; Thirdly, different judging criteria, including absolute distance, Euclidian distance and similarity matching, are adopted to match different requirements on recognition speed or accuracy. For 7000 samples of first class Chinese characters, the first and first five correct recognition ratio turns out to be 95 percent and 98 percent respectively in our experiments. The investigation has built a solid theory foundation to the research and development of our“Electronic Reading Pen”project.3774|309|0更新时间：2024-05-08
- 摘要：To solve the problem of acquiring the information of view outline, some experts have proposed many methods. But there are too much calculation and too complex determinant in prevail methods, which has become an important problem. Based on topological mapping, this paper puts forward automatically acquiring the information of view outline. At first, based on topological mapping principle, the method maps points of the candidate path onto the projection line. And then, using the topological sequence compatibility of points on the projection line and the candidate path, the problem of determining the trend of the next branch is converted to decision of the trend of projection lines. Finally, according to the sort order of mapping points, the path can be determined. Since the projection line is horizontal, it is easy and convenient to obtain the mapping points. This algorithm also saliently improves the computation efficiency by refraining from the complex angular calculations. Moreover, the algorithm will play an important role in automatic partitioning of engineering drawings, assembly diagram processing and dimension arrangement.2820|270|0更新时间：2024-05-08
-
Management and Usage of Metadata and Method Metadata in Resources and Environment Information System
摘要：To share knowledge in an resources and environment information system, it is necessary to find an approach to manage data, which includes spatial data, remotely sensed data, attribute and knowledge, and models, which consists of spatial analysis models and spatial decision support models. Our approach of solving the problems is to create an integrated frame for metadata of resources and environment information and method metadata of environmental modelling, and integrate data and models in the resources and environment information system based on the frame. In context of the resources and environment information system, data usually include attribute data, vector data, raster data, and knowledge data on narrow sense. Its metadata defines the specifications, qualify control and information sharing for the narrow sense data. And some standards for the metadata had been established. This article considers that it is essential to normatively describe ways and means of environment modeling to share knowledge about modeling, so puts forward the concept of method metadata to standardize descriptive information about environment modeling for the first time. We think that the integrated metadata frame should consist of the metadata about the data and the method metadata about models. In development and operation of resources and environment information system, we can integrate database, remote sensing, geographic information system and environment modeling based on the integrated metadata frame, due to these two kind of metadata adequate describing characters of data and data processing in the system. The integrated metadata frame will be adopted to integrate and share data and models by some projects, such as the resources, environment &district economy information system, and the research on quantitative remote sensing. It is positive that the integrated metadata frame will do a good job.2451|235|0更新时间：2024-05-08 - 摘要：In the close range photogrammetry , the quality of the generated digital surface model is controlled by some course parameters . This paper discuss these parameters from technique method , band choice and fusion , image correlation parameter , analysis their influence to the quality of output DEM , and present the experienced way of the choice of these parameters. Using two different technique: Block or Stereo, the number of correlation points is near, but the quality of the result is different , the result of Stereo is better. There are many bands can be used, but only one is needed in the image correlation. To the different color band , Red and Green seem to have good result, the Red band is better. The fusion band , such as intensity band or the first principle band , can increase the correlation points. At the image correlation course, the correlation template size is sensible to In some range, increasing the template size can increase the number of correlation point. The larger template range , the larger number of correlation points. But at the same time, the error area appears seriously. To resolve this problem, using rejection factor, maximum parallax and skip factor can get a better result image.2403|257|0更新时间：2024-05-08
- 摘要：Connected component detection is an elementary and time-consuming operation in digital image processing. It is widely applied in pattern recognition, characteristic extraction, image understanding and so on. For the purpose of increasing efficiency of connected component detection, based on claaical data structure "queue", This paper develops a new data structure called "rectangular queue". A fast algorithm for connected component detection of small rectangular areas of a binary image by using the new data structure is also proposed. The new algorithm can be used to detect connected components of pixels in a small not only square area but also rectangular area. It can also be used to detect both horizontal lines and vertical lines in an image. Based on some experiments, the conclusion is obtained that the speed of the new algorithm increases by over one times compared with the currently available ones. The algorithm has been applied to extract point symbols, linear features and further linear features with different width from binary image.2727|264|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