This report presents some personal impressions of the 10th International Conference on Pattern Recognition (ICPR) which was held in Atlantic City, New Jersey, during 16-21 June, 1990. This is the latest in a series of bi-yearly conferences held in the field of pattern recognition including computer vision, image processing, signal processing; and etc. The scope of the conference covers a broad range of these topics; from algorithms to architectures. This year, the conference was organized as an umbrella event composed of four parallel specialist subconferences. We attempt to present the atmosphere of the 10th ICPR, focusing on the COMPUTER VISION track.
The back propagation network,(BPN) is applied to human-face identification. A mosaic pattern with 12 X 12 blocks, transformed from central part of human-face image, is put into BPN. This combination yields hundreds of person's identification with robustness to deformation such as deforcused image or different face expression. Hidden units of the BPN extract peculiar and delicate features of human-face, which can not be obtained from existing statistical methods. A few of hidden units can especially select only men or women.
When there exist many elements, such as dots bars and so on In the visual field we tend to organige them perceptually to form some groups. We call this phenomenon "grouping". There are three principal factors of grouping: factor of proximity, factor of similarity, and factor of good continuity. We conducted related psychological experiments, then propose a model for grouping using these factors incorporating our experimental results, and show simulation results for some simple patterns.
We measured binocular eye movement when a static random-dot stereogram (RDS) and a dynamic RDS is shown to a subject, and tried three dimensional display of the view points of eyes from obtaind result in order to see which place of the RDS did the subject gaze. In the result, we found vergent eye movement (vergence) can be observed after stereoscopic perception rather than when searching a corresponding point.
This paper proposes a new point of view for investigating small involuntary movement as a deterministic system. We measured small involuntary movement by a limbus reflection method and described the effective degrees of freedom by a fractal dimension analysis. The correlation dimension was around one for each coordinate. This suggests that the small involuntary eye movement is not pure random noise but a chaotic phenomena. Consequently, strong correlation among eye muscles is expected.