Application of image contours to three aspects of image processing:compression, shape recognition and stereopsis

摘要

Proposes the use of contours of constant intensity as imageprimitives for the purpose of image compression, object recognition andstereo matching. Compression ratios of 10-15% have been achieved, whichapproaches those of conventional coding techniques. The advantages ofstoring the images in this form is that explicit shape informationcontained in the image is available directly from the coded form. Thecontour set is partitioned into two parts, the long closed contourswhich tend to represent object boundaries and the shorter unclosedsegments representing background detail. The former may be used as inputto the object recognition and stereo matching schemes which have beendemonstrated in this paper. Although further work is required before thecompression ratios achieved supercede existing techniques, it isbelieved that the slight loss in compression ratio is a small price toconcede for obtaining a compressed form which may be used directly