In this paper, I introduce the current information about CG(Computer Graphics) and CV(Computer vision) technology at many universities in The United State such as MIT, New York State University, Southern California University, during my staying there in 1992. The current themes in these fields are multi-media oriented system, designing of hardware system and new applications of virtual reality etc.
An image database is projected collecting textile pictures to aid designers from a graphical aspect. Classification of images uses eight scales ranging each from +3 to -3 based on eight pairs of adjectives. Our first study about the classification was carried out by questionnaires of some fifteen people aganist about fifty pictures. Our study is to classify thousands of textile pictures, automatically. Noting on colors in pictures, prominent colors are chosen as image-colors which can represent the first impression of a target picture. The image-colors give the scales corresponding a pair of adjectives, warm and cool.
A color image can be defined by a function(vector filed) F:V2→V3 where V2 is a 2-dimensional pixel coordinate space and V3 is a 3-dimensional color space with r,g,b color signals transformed to CIE L*a*b* uniform color space. The derivative of the vector filed F'(x) is characterized by the Jacobian matrix. The largest singular value of Jacobian matrix shows the vector gradient magnitude. In this paper we show that the detected edges obtained from the largest singular value are more accurate and finer than the edges obtained by the differential operator method. Where a color and its complement color collide on a boundary line. the direction of the projected vector to the (a*.b*) space reverses and the source or sink arises. Where two colors collide, only the luminance component of color signals grows larger. This means that the image is enhanced at that area.
We developed a lips animation prototype system. The lips animation in this system is the motion of lips synchronized with synthesized speech sound in Japanese. Our system consists of three sub-systems, such as speech synthesizer, motion generator and the real-time texture mapper. Speech synthesizer generates sound data and pronouncing timing data of speech from a input text. Motion generator makes lips motion data from pronouncing timing data. The real-time texture mapper synthesizes lips animation images from lips motion data and four prepared basic lip-shapes. We confirmed by experiment that this system can make smooth lips motion automatically.
This paper introduces the research acivities in Tokyo Denki University in the field of Image processing and human interfaces. In the concrete, research thema which are eagerly carried out in 3 Iaboratories in Information and Communication Department are shown. Some of them are as follows:(1) as image processing techniques, Face recognition, Lip reading, Generation of facial expressions and so on, (2) as human interface techniques, Functional design and Development of pen-computer, (3) as fundamental techniques, Analysis and Synthesis of free-form curves and surfaces are described.