We propose an extension of RBF networks which includes a mechanism for optimizing the complexity of the network. The approach involves two procedures: adaptation (training) and selection. The first procedure adaptivel...
详细信息
We propose an extension of RBF networks which includes a mechanism for optimizing the complexity of the network. The approach involves two procedures: adaptation (training) and selection. The first procedure adaptively changes the locations and the width of the centers of the basis functions and trains the linear weights. The selection procedure performs the elimination of some of the basis functions using an objective function. By iteratively combining these two procedures we achieve a controlled way of training and modifying RBF networks, which balances accuracy, learning time, and complexity of the resulting network.
Stereo computation is one of the vision problems where the presence of outliers cannot be neglected. Most standard algorithms make unrealistic assumptions about noise distributions, which leads to erroneous results th...
详细信息
A caesium magnetometer attached to a handcart enables accurate magnetic prospection of large areas within short times. The roughness of the earth's surface, minor differences in handling the handcart and a lot of ...
详细信息
Recently, much attention has been devoted to image-based scene representations. They allow to construct an arbitrary view of a 3D scene by the interpolation (transfer) from a sparse set of real 2-D (reference) images,...
详细信息
The problem addressed in this paper is related to displaying a real 3-D scene from any viewpoint. To display a scene, a relatively sparse set of 2-D reference views is stored. The images that are in between the refere...
详细信息
Stereo computation is one of the vision problems where the presence of outliers cannot be neglected. Most standard algorithms make unrealistic assumptions about noise distributions, which leads to erroneous results th...
详细信息
Stereo computation is one of the vision problems where the presence of outliers cannot be neglected. Most standard algorithms make unrealistic assumptions about noise distributions, which leads to erroneous results that cannot be corrected in subsequent postprocessing stages. In this paper we present a modification of the standard area-based correlation approach so that it can tolerate a significant number of outliers. The approach exhibits a robust behavior not only in the presence of mismatches but also in the case of depth discontinuities. The confidence measure of the correlation and the number of outliers provide two complementary sources of information which, when implemented in a multiresolution framework, result in a robust and efficient method. We present the results of this approach on a number of synthetic and real images.
The basic limitations of the current appearance-based matching methods using eigenimages are non-robust estimation of coefficients and inability to cope with problems related to occlusions and segmentation. In this pa...
详细信息
The basic limitations of the current appearance-based matching methods using eigenimages are non-robust estimation of coefficients and inability to cope with problems related to occlusions and segmentation. In this paper we present a new approach which successfully solves these problems. The major novelty of our approach lies in the way how the coefficients of the eigenimages are determined. Instead of computing the coefficients by a projection of the data onto the eigenimages. we extract them by a hvpothesize-and-test paradigm using subsets of image points. Competing hypotheses arc then subject to a selection procedure based on the Minimum Description Length principle. The approach enables us not only lo reject outliers and to deal with occlusions but also to simultaneously use multiple classes of eigenimages.
An IFS is a set of affine and contractive *** union(so-called collage) of the subimages generated by transforming the whole image produces the image again. In the edge maxima matching algorithm the IFS-Codes for a sin...
详细信息
An IFS is a set of affine and contractive *** union(so-called collage) of the subimages generated by transforming the whole image produces the image again. In the edge maxima matching algorithm the IFS-Codes for a single image are calculated by affinely mapping maxima of the edge of the entire image to the correspondmg maxima of the edge of each of the *** are the following three principle ideas of this algorithm to be proofed. The decomposition of the image in a minimal set of subimages is found by two adjacent edge points(so-called touching points of a subimage).These touching points are recognized by being the only points which neighborhood have no affine and expanded equivalent in the image A unique representation for each of the subimages is found by choosing most expanded affine mappings of four edge extremes to characterise each of these image parts. An affine mapping(the IFS-Code) between the image and the subimage is calculated by using an affine mvariant representation thru such quadruples of edge extremes
作者：
Aoki, YIijima, TMemberJapan Advanced Institute of Science and Technology
Hokuriku Ishikawa Japan 923 Yasubim Aoki:graduated from the University of Kanazawa Faculty of Engineering
Department of Electrical Engineering and Information Science in 1992 and received his Master's degree in 1994 from the Japan Advanced Institute of Science and Technology Hokuriku Department of Information Science where he is currently in the doctoral program engaged in research on pattern recognition. Honorary MemberTaizo lijima:graduatedwith a B.E. degree from the Tokyo Institute of Technology in 1948 and received his Ph.D. degree later. He joined the Electro-Technical Laboratories in 1948 and
since then has been involved in research on electromagnetic theory and pattern recognition theory image processing voice recognition and research and development of OCR. After serving as the head of Iijima Special Research Laboratories he became Professor at the Tokyo Institute of Technology in 1986. He because Professor Emeritus of that institute in 1986. He became Professor at Tokyo Engineering University and Professor Emeritus there in 1991. In 1991 he became Professor at the Japan Advanced Institute of Science and Technology and was appointed Vice President of the institute in 1992. He served successively as editor examiner secretary of general affairs inspector counsel committee-head of pattern recognition and specialized studies head of Tokyo chapter and vice president of this institute. In 1991 he became Honorary Member. He received the Paper Award from the Institute four times and the Literary Works Award once. He received the Achievement Award and the Distinguished Service Award in 1976 and 1989 respectively. He received the Purple Ribbon Medal from the Emperor in 1989.
Theoretical studies on Iijima's ''Theory of patternrecognition,'' which is known as a method of multiple similarity, have shown clearly that there are situations when it is difficult to determine ...
详细信息
Theoretical studies on Iijima's ''Theory of patternrecognition,'' which is known as a method of multiple similarity, have shown clearly that there are situations when it is difficult to determine the viewpoint and visual field, important parameters to introduce topological aspects into the pattern space, observing a figure itself. To cope with this problem, this paper proposes transforming the observed image from a figure f(r) to a second-order differential image rho(r), called ''figure source.'' The aim is to reconstruct the theory in such a way as to enable the application of similarity evaluation methods as in the conventional case, without compromising the current theory, and all the necessary proofs are given. In this paper, the figure source equation as well as rules representing its properties are developed, and several related physical quantities are redefined.
A theory of natural observation methods was originally formulated for two complementary systems: neighboring type and equilibrium type. This paper introduces a new system, normal-type system, that is different from th...
详细信息
A theory of natural observation methods was originally formulated for two complementary systems: neighboring type and equilibrium type. This paper introduces a new system, normal-type system, that is different from the forementioned two types. One characteristic feature of the normal-type natural observation method is its ability to reconstruct an original waveform from the sum of a finite number of observation values. The natural observation theory of a normal type evolves naturally under the condition that a waveform and its power component can be expressed in the sum form. These meanings are discussed by comparing a frequency response of the normal-type natural observation transform with that of a conventional filter bank. Furthermore, a criterion for determining observation parameters in order to realize the natural observation of normal type under a given tolerance of error is described. It is believed that the proposed theory offers new insights into systematization of the natural observation theory.
暂无评论