Bitstream Based Edge Detection Techniques

摘要

We consider here edge detection as a problem of identifying points in an image where the image brightness changes sharply or has discontinuities. Specifically, we investigate several well-known algorithms for edge detection and modify them for performance directly on the bitstream of a JPEG compressed image. By performing the processing in the compressed domain, we dramatically reduce its complexity. The low complexity combined with the fact that it does not require JPEG decoding makes this approach well suited for distributed wireless sensor networks. To extract information from JPEG compressed images, we can use the number of bits allocated to each of the 8x8 subblocks. Using this preponderance number, we are able to extract information regarding the frequency content of the image. In this paper we will present a comparison of firstorder, differential, and phase congruency approaches to edge detection and determine how these algorithms perform when applied directly on the bitstream domain. The performance of the edge detectors will be determined by using the Figure of Merit (FOM) and Closest Distance Metric (CDM).