Parallel image normalization on a mesh connected array processor

摘要

Image normalization is a basic operation in various image processing tasks. A parallel algorithm for fast binary image normalization is proposed for a mesh connected array processor. The principal operation in this algorithm is pixel mapping. The basic idea of parallel pixel mapping is to utilize a store and forward mechanism which routes pixels from their source locations to destinations in parallel along the paths of minimum length. The routing is based on a simple yet powerful concept of flow control patterns. This can form the basis for designing other parallel algorithms for low level image processing. The normalization process is decomposed into three procedures: translation, rotation and scaling. In each procedure, a mapping algorithm is employed to route the object pixels from source locations to destinations. Simulation results for the parallel image normalization on generated images are provided.