A fast parallel image matching algorithm based on MPP computing model

Abstract

Image matching is a branch of image processing. Its application is widely the precision guidance field. And the image correlation processor is one of the key technologies in image matching system. In this paper, the embedded MPP computing model is used as the image correlation processor. It mostly accomplishes image matching processing and image preprocessing. The embedded MPP system is designed as 16-bit fixed array microprocessor and composed of 64 × 64 processing elements (PEs) array. The PEs array is connected as 2-D mesh of circle connected computer. All of PEs perform the same operation according to the mode of single instruction stream multiple data (SIMD) stream. According to traditional idea of the Normalization product correlation algorithm, a fast parallel image matching algorithm based on MPP model is proposed. The result shows that, the algorithm complication is lower and computation speed is faster than the traditional algorithm, and the matching precision is the same as last one.