Efficient contour extraction in color images

Abstract

An extension of second-order differential edge detection methods to color images was previously proposed. This work shows how, by casting the problem into a subpixel resolution oriented framework, an efficient algorithm for edge detection and contour traversal can be devised for both graylevel and color images. The results from an implementation on a standard PC show that the computational cost is largely dominated by convolution-type operations, whose impact can be substantially reduced by using specialized image processing devices.