摘要
通过分析低层图像处理算法特点，指出标准ＳＩＭＤ结构不能完成某些低层图像处理算法的数据综合任务，提出的改进型ＳＩＭＤ结构增加了多指令控制、共享存储器单元，能执行更多种类图像处理算法，讨论了图像在ＳＩＭＤ结构计算机上的一维、二维分块方法，结果表明一维图像分块在互连网络硬件量、算法执行速度两方面更具优越性．
Based on an analysis of the specific features of a low level image processing algorithm, it is suggested by the authors that the standard SIMD is incapable of accomplishing the data synthesizing tasks required. An improved SIMD architecture with multiple instruction control and a sharing memory is presented, which will be able to implement many kinds of image processing algorithms. The 1 D and 2 D image blocking schemes are discussed. The results show that 1 D image blocking is better in terms of the amount of hard ware in the interconnection network and the implementation speed for some algorithms. with the LUT technique, a flexible and simple arbitrating circuit for the sharing memory is designed. The speed up ratio and programmable ability of the architecture proposed for the image processing algorithms are briefly discussed.
出处
《华中理工大学学报》
CSCD
北大核心
1997年第A01期30-33,共4页
Journal of Huazhong University of Science and Technology
