Parallel Implementation of Features Extraction Using Morphological Filter

This paper describes parallelism of image data 11sing a parallel process and developed the parallel features extraction using a morphological filter on the new generation multi-transpt~ter's network system. This system solved the limitation of communication problem in conventional transpnter networks using Inmos T9000's virtiial channnel processor and Inmos C104 packet routing switch. A morphological filter is often applied to features extraction and pre-processing in the fields of medical image processing. However, the large amount of the calculations of morphological filters have been an obstacle to the research of this method. Therefore, a fast calculation technique is necessary. In order to achieve faster calculations, parallel processing is an effective technique. The performance and efficiency of the system is examined in regard to the recognition of shapes, parallelism and network topology. The processing software was implemented in "occain2". Experimental results show high performance and efficiency in parallelism. Massively parallel processing systems could be implemented using this system in the future.