Directional-edge-space processing for multiple-resolution image perception

In this paper, we propose logic processing to carry out in the directional edge space in order to establish multiple-resolution image perception in our VLSI perception system. In the con-ventional system, the resolution of input images is lowered by shrinking their sizes before edge ex-traction. However, in the human visual system, scaling of images for perception is carried out after the oriented edge detection in the primary visual cortex. Being inspired by such a physiological fact, the present scheme has been proposed. In the proposed scheme, we can lower the resolution of input images by converting the full-size edge maps into 1/2-size, 1/4-size, 1/8-size edge maps. It was con-firmed by experiments that these edge maps can preserve the edge features in the original full size image at reduced scales much better than those obtained by the conventional image scaling & edge detection scheme.