In binocular vision, regions which cannot be binocularly paired have considerable effect on human stereo perception. In this paper, we point out the function of stereo perception we found through experiment concerning perception of volume, which is considered as the effect of unpaired region on stereo perception. And then we propose some strategies to realize such function upon computer vision systems. We have done experiments on the effects of one-dimentional change of regions between paired, unpaired and unpairable regions on perception of volume. From the results we propose flow of process and the way to process information from the unpaired region. And we also pointed out the conditions data models of perceptional image must fulfill to enable such processing.

View full abstract

Understanding human intention can offer an appropriate user interface using virtual workspace. Grasping hand motion is taken up in this paper, and it is considered that each motion reflects human intention. As one trial to prediction of target and human intention understanding, the experiment was performed to investigate human hand-eye motion when we try to grasp some objects formed in virtual space. The results show that the angle between center gravity direction of object and that of hand is almost constant, and this doesn't depend on object's shape, size, orientation and distance. That leads to the possibility of target prediction and manipulation aid.

View full abstract

New Types of visual effects has been found in the case of the occluded objects and/or occluding objects had motion. The illusory object can be perceived clearer than when they are stational. Several new findings in the visual effect, perception of size change, shape change (distortion) and eccentric motion of the illusory object, were made ; and they cannot be expected from the stational case. In connection with these phenomena, several famous geometrical illusory figures have been investigated in the case when their inducing elements were dynamically moved, then the interesting visual effects have been found. In addition, 3-D perception with moving objects have been investigated and interesting phenomena have been observed.

View full abstract

Binocular fusional limits are measured under two conditions. One is 'fixating' condition that a subject fixates a center point and estimates fusional limit for the whole screen, the other is 'free-observing' condition that a subject changes his gaze point between far and near target alternately and estimates fusional limit at each fixating point. In free-observing condition, fusional limit is in proportion to viewing angle and has weak dependence on viewing distance. There are some cases that fusional limit of far point is smaller than the disparity when the lines of sight are parallel. In fixating condition, the fusional limit is ±40 [arc min] as maximum and it is larger than Panum's fusional limit. Moreover, a method to design stereoscopic image parameters to display all objects within a subject's fusional limit, is examined using the measured fusional limits.

View full abstract

We proposed a model for motion integration based on neuronal properties in monkey's middle temporal(MT) area. It has been reported that neurons in area MT, which might play a crucial role in motion processing, should be organized in a columnar fashion. Little is clear, however, about how the responses are integrated by intercolumnar interaction. There are two dominant candidates such as vector summation or Winner-Take-All(WTA). We demontrated in our model that the direction would be determined by WTA, and that vector summation could be regarded as a specific case of WTA mechanism.

View full abstract

This report describes reduction of computational complexity or digital image data coding. Requirments of high performance of image data have been increasing the number of the pixels of the data to be processed. Therefore it costs a huge amount of computational load. This problem can be solved by representing value of multiplier coefficients (DCT coefficients) as short as possible. However, the condition that the bit reduction does not increase quantization noise of decoded image must be considered. This report showes that the bit reduction under the condition is feasible and drives the minimum bit of examples such as MPEG or JPEG algorithms. The bit reduction enable us to reduce hardware complexity of the algorithm without increace of the coding noise.

View full abstract

It is presented that pattern recognition model based on induction field exists in human visual perception system. The theory of the field of induction on the retina is a useful method that explains many visual perceptions mathematically. Many research activities make efforts to recognize patterns by using the theory of the field of induction. As a result, it is supposed that patter recognition model based on induction field exists in human visual perception system. In this paper, a pattern recognition model is proposed from many psychological researches. First, practicableness and nature of proposed model are examined by computer simulation. Finally, existence of the pattern recognition model based on induction field is discussed.

View full abstract

Filling-in phenomena, which occur at stabilized retinal images as well as at retinal scotoma, is suggested to be a fundamental processing for visual perception. In our visual system, information such as brightness and color around edges of retinal images is presumed to expand somewhere in the cortex within 100ms to construct our two dimensional visual image perception. If such expansion is true, we can visualize the process by presenting two frames of different images on the display fast and alternatively, since each image may be perceived incompletely because of insufficient expansion. Using this technique, we really observed the visual perception of insufficient expansion which proved the process of filling-in.

View full abstract