This paper Describes a method of selecting proper view direction for recognition of polyhedral objects. Our method is based on range maps and have two simple criteria: (a) the number of visible planes, (b) the uniformity of areas of image planes, to evaluate images taken from some view directions. Gradient distribution which represents the histogram of gradients have some information about feature of planes, so that we found that evaluation based the criteria correspond to entropy of gradient distribution. Several experiments, using real range maps, are presented and the efficiency of this entropy measure is discussed.
When assembling mechanical parts or restoring broken pieces we are required to superpose two congruent shapes. Automatic superposing method however, has not been proposed yet. This paper, therefore, provides a method applicable to any shape in a two-dimensional space. In this method, one shape is translated and rotated so as to nullify the area of a region lying between the two shapes.
Composite video signal being digitized, it is always affected by voltage gain and offset of the accompanying analog circuitries. Further, because of non-ideal low pass and band Pass filters., signal coupling exists between luminance and chrominance signals. Therefore, the data obtained from A-to-D converters should be compensated by some means. This paper deals with four incremental models of the digitizing process. They are tested by experiments. The degree of precision is expressed by the three kinds of color differences, including a measure of the just noticeable color difference based on the MacAdam' s ellipses, in order to evaluate the models and to get the most reasonable one.
An extraction method of pairs of feature points from binary image is presented. The chain codes are used to detect indentations, called turning points, in boundary. The two pairs of vectors are used to extract the pairs from turning points. A pair of feature points mean two points at terminal or bending, turning position of a figure, and a line of segmentation can be drawn between them. Because the length of vectors is variable according to the distance of two turning points, this method can be applied to line like figures and blobs. The extraction procedure is applied to the boundaries of damages TAKUHON characters. It is then possible to pick out convexs and positions where character links with background noise. Some experiments are given.
An approach for interactive motion picture editing, which involves a direct manipulation of the space-time structure, is proposed. The subject of this approach is to offer an easy, enjoyable and creative environment by a support of image understanding technique. Three experiments are made: a separation of a still object in a room scene, a separation of a moving object in a outdoor scene, and a visualization of the space-time structure of this outdoor scene. These results show appropriateness of our approach.
In this paper, we discuss a method which enable us to retrieve trademarks similar to an inputted trademark from various trademarks including logotypes and realistic marks and etc. We aim at making retrieyed trademarks similar to these corresponding to human subiectivity. To do this, it is necessary to reflect the human subjectivity on the system as much as possible. Here, a method is proposed, in which human subjectivity quantitatively reflects on the system by using the result of questionnaires set to many people 8nd the result of experiments by oup uethod is shown.
In this paper, we describe our similarity retrieval method for logical schematic diagrams, especially for plants's piping and instrumentation diagrams. In such a diagram, the connection between symbols and lines are essential, and a way to store such data efficiently is presented. And the use of knowledge-base for interpreting diagram structures and retrieving conditions are discussed, to retrieve diagrams suited to a variety of similarity measures. Therefore, we can perform retrieval of diagram data according to its contents, with upgraded man-machine interface functions.
With the evolution toward a broadband digital network, nationwide image-signal transmission systems will be constructed in order to offer a high-quality and reliable service to customers. The largest users of image transmission systems will be broadcasting stations. For contribution of image signals inside a studio and between studios or broadcasting stations, the capability for multiple encoding and decoding of the signals without visible degradation is required. The best would be if there are no quantization-noises at all. This paper presents a subband coding which has the possibility of errorfree encoding. Ah architecture, of short-kernel subband filters which can reduce the entropy of subband signals is proposed.