| A new technique to achieve rapid recognition of object edges and vertices within a visual scene using the information of the regional gray level changes, viz. regional feature detection, instead of local edge detection using the gray level changes between adjacent pixels, has been developed, programmed and tested on a variety of images. The hexagonal grid, with less directional sensitivity, is a better approximation of the human vision grid, is more suited to natural scenes than the square grid and avoids the 4-neighbor/8-neighbor problem. A computer code has been developed to detect edges of digital images of real objects taken with a CCD camera within a hexagonal grid. The digital images obtained from the improved CRS digital image processing system are a good approximation to the images which would be obtained with a real physical hexagonal grid.;A program which uses the differences of the regional gray level changes was developed in the programming language C++. This scheme of regional feature detection is simpler, easier to implement, and more robust than the local edge detection scheme employed by R. W. Stimets and H. Tang, and its theoretically lower resolution may not be a handicap in many cases.;The two techniques of local edge detection and regional feature detection have been quantitatively compared and the regional method is superior. (Abstract shortened by UMI.). |
