[in Japanese]

View full abstract

We have measured the binocular eye movement to investigate the sight-line control mechanism in binocular stereopsis, so that the mechanism can be applied to computer vision. The experiment uses a cylinder, a prism and a plane. The effect of limb occlusion is determined by using a t-test to evaluate fixation points while gazing at a textured object and while gazing at a non-textured object. For the cylinder, the binocular fixation point tends to be located more outward when the subject gazes at the limb of a textured object than that of a non-textured object. For the prism and the plane, the fixation points do not shift. These results show that differences in the visual distance between an object and the background do not seriously influence the fixation point. However, when limb occlusion is perceived, humans can not gaze at the limb with both eyes, but shift the fixation point to reduce the incongruity between the left and right retinal images.

View full abstract

Experimental results are reported for how people accurately recall the location of a point from their short-term memory.
People identify a specified picture by using key words on the objects in pictures and the spatial relationships of those objects to distinguish one picture from another. The layout information in pictures can be used to retrieve them from a database.
However, the layout information given by people is often vague. Therefore, the wider the search area, the longer the retrieval process. On the other hand, the narrower the search area, the greater the risk of missing the target picture.
This study evaluates the accuracy of recalling locations in a picture within the time-span of short-term memory. It is found that the true position is estimated based on an area that represents only 5% of the picture and which is centered on the position specified.

View full abstract

[in Japanese]

View full abstract