The Parallel Detection of Interesting Points in Distance Transform for Image Matching

This paper presents a parallel method to detect interesting points as image feature pixels in distance transform for image matching. Unlike the traditional methods in which edge detection is regarded as the pre-processing step to obtain a binary edge image for distance transform operation, in our approach the interesting points are detected as feature pixels by means of a dynamic thresholding procedure. The concept of remote procedure call(RPC) in distributed systems is introduced for the parallel implementation to achieve the speedup without specific software and hardware requirement. Thus a guided image matching system can be established by the hierarchical detection of interesting points in distance transform.