It has been nearly thirty years since the study on eye movement was first undertaken in the field of engineering. However, many reports on this subject are still being published. The characteristics or control mechanisms of saccade, smooth pursuit movement, vergence movement and optokinetic nystagmus, etc., many problems are taken as subjects for study. Especially, the behavior of smooth pursuit system is complicated relayted to the function of central nervous system or vision. So, the number of reports on this subject is distinguished. In this paper, recent studies on eye movement are surveyed giving priority to smooth pursuit movement.

抄録全体を表示

Signal to noise ratio (S/N) which is often used can not completely express picture quality. The objective estimation which mesures picture quality accuratelier than S/N is wanted. This paper presents the estimation method which uses visual model. This method apply correspondingly to subjective estimation, because the model simulates the image in brain. A new estimation measure, EPR (Error signal Power of Retinal response), which uses the retinal characters, is also proposed to realize the method for gray still pictures. EPR is very useful for the variety of noise.

抄録全体を表示

人間の視覚系は特定の狭い空間周波数領域だけに応答する多くのチャンネルの並列的配置により構成され、それにより2次元入力パターンの空間周波数分析を行っていると考えられる。また、各々のチャンネルには弁別しきいが存在し、その値未満のコントラストのグレイティングは知覚することがてきない。これは、1枚の画像を空間周波数がある一定の帯域幅を特つ何枚かの画像に分け、その各々について視覚の空間周波数特性を考えられることを意味している。このよすな視点からシミュレーションを行った結果、視距離をブラウン管の高さの5倍とした場合には、周期が4[pel]以上のグレイティングのほとんどが弁別しきい以上となることが分かった。そこで本報告では、原画像を帯域幅が4[pel]以上と4[pel]未満の2枚の画像に分け、各々をコサイン変換しグレイティングの振幅をコントラストに変換した後に量子化を行い、帯域幅が4[pel]未満の画像については、コントラストが弁別しきい以上となるグレイティングの分布によってブロックサイズをアダブティに変化させ符号化を行う方法を提案する。また、ブロックサイズによる符号化特性を調ベアダブティブな符号化法について検討をしていく。画質評価は、他の手法と比較するためにS/N比で行った。その結果、量子化器の最適化を行っていないにもかかわらず高いS/N比が得られた。

抄録全体を表示

A high quality color priting image compression method is presented. This contains dividing image into nxn blocks, vector-quantizing each block independently. In this stage, quantizing level is decided by means of range statistics. As the result of this method, about 1/10 of compression ratio and about 30dB of S/N raio is achieved.

抄録全体を表示

In this study, the relation between the number of pixels and the informations distinguishable in facial images by human vision was examined through a psychological experiment. The results showed that the discrimination limit in determining faces is 7 ppf (pixels/face height), that in determining people is 13 ppf, that in determining large facial expressions is 15 ppf and that in determining small facial expressions is 38 p/f.

抄録全体を表示

This paper describes the visual condition (resolution, tone levels) about chrominance signal in video communication using small screen. It has been shown that MOS value is 3,in the luminance signal (128×128 pixels, 32 tone levels), in the chrominance signal (16×16 pixels, 8 tone levels), and in the scrren size less than 4 inches. The relation between visual distance (d/H) and S/N has been shown to get MOS 3 value.

抄録全体を表示

Lu and Fender, and Gegory found that depth could not be perceived in iso-luminant color random-dot streograms (RDS). The aim of this paper is to confirm whether or not the "chrominance channels"are responsible for streopsis even under iso-luminant conditions. Thus, we conducted subjective evaluation tests using a positive and negative iso-luminant red-green/yellow-blue RDS to study the role of color in human depth perception. We found that the horizontal and vertical disparity limits of binocular fusion for the iso-luminant color RDS were the maximum at about 25 minutes and 3〜5 minutes respectively, 0.5-0.7 times lower than the disparity limits for black/white RDS. The temporal frequency limits of binocular disparity for the iso-luminant color RDS were about 3 (Hz). This finding suggests that the cells of the disparity detection area in the visual cortex are responsive to both black/white and iso-luminant color stimuli.

抄録全体を表示