This paper presents an effective method of selecting candidates for correspondence in segment-based stereo. Boundaries of objects are extracted from stereo images and segmented at feature points. The resulting line or curve segments are used as units for correspondence. The initial candidates for correspondence are obtained based on the epipolar constraint. And, more reliable candidates are selected using the features of each segment such as intensity, direction, and shape. When there is another camera in addition to the basic two cameras, the image can be used to further reduce the candidates.
In order to set up a decorative pattern database of old ceramic products, it is necessary to develop the decorative pattern in a three dimensional shape. This paper describes the recovery of the surface of revolution from a monocular image with (unknown camera) parameters, and development of the decorative pattern. Using genetic algorithm (GA) and four direction fields, the camera parameters and the shape parameters are obtained. Then, the surface of revolution is reconstructed and developed into a plane with the pattern. As a result of using a monocular image of an old ceramic product digitized by scanner, the surface of revolution is reconstructed and developed into the pattern image.
In this paper, a method for the extraction and recognition of road markings (lines, numerals, characters, symbols, arrows, etc.) from a road scene image is proposed. At first, the input image is segmented to regions using outlines. After considering the conditions of the road structure, those regions are partially rejoined, and then the road region can be extracted. After the road region is transformed into a plane image by perspective transformation, the road markings can be extracted. Using projection profile pattern matching between the extracted road markings and the standard patterns is practiced and the road markings are recognized. As the experimental result, practical recognition performance can be obtained.
In this paper, we present a control method of data cache for a multithreaded processor and its evaluation. A multithreaded processor is effective for 3D-CG, however the increase of the working set size is unavoidable, and this limits the effectiveness of the data cache. Usually, the size and/or the associativity of the cache are increased in order to achieve a higher cache hit rate. This causes the chip size to increase, but the performance remains limited. An inter-thread non-blocking cache control method is proposed for reducing cache miss penalties. This control method achieves higher performance than the blocking cache method and also requires much less hardware cost than a traditional non-blocking cache method. In the case of the proposed cache control method, the performance degradation decreases to half and the performance ratio achieves 80-90% of an ideal cache case.
We have developed a 100-inch slim liquid-crystal rear projection display. Cross-dichromatic mirrors are used to combine three colors. Projection lenses were designed to decrease the color shading induced by these mirrors. In order to make the set slim, the projection lenses were also designed to have a short projection distance. To obtain good contrast ratio characteristics, the central axis of the projection lenses is inclined to that of the LCD panels. Transversal magnification color aberrations, which induce convergence error of the three colors on the screen are few. By using these technologies, we have developed a slim set with a depth of 1.1 m.
A facial expression is defined by a time sequence of facial images. Various kinds of facial expressions can be generated from a basic facial image sequence by modifying the time generation functions. This paper describes a 3-D image generation method for facial expressions that uses three kinds of time generation functions. The facial expressions generated by these functions are compared and a method that breaks observed facial expressions down into these time generation functions is presented.