灰度共生矩阵的快速实现和优化方法研究  

Study on fast implementation and optimal methodof gray level co-occurrence matrix

作　　者：郑红[1] 李钊[1] 李俊[1] 

机构地区：[1]北京航空航天大学自动化科学与电气工程学院,北京100191

出　　处：《仪器仪表学报》2012年第11期2509-2515,共7页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(60543006);博士点基金(201003259);重点实验室基金(9140C150105100C1502)资助项目

摘　　要：灰度共生矩阵计算具有计算量大、计算延时长等缺点,严重影响计算的实时性。针对此问题在充分分析灰度共生矩阵计算并行性的基础上,提出了快速计算灰度共生矩阵的并行算法和并行体系结构,利用图像中像素之间计算的可并行性,设计多个Processing Elements(PEs)对多个像素进行并行计算,可有效地实现灰度共生矩阵的快速计算。由于嵌入式可重配置硬件平台资源有限,对灰度共生矩阵计算性能进行优化的同时还对资源消耗进行优化。通过对计算延时(latency)和资源消耗(area)的分析,建立新的优化目标函数,实现了性能和资源折衷的优化设计。Gray level co-occurrence matrix （GLCM） has the defect of great computation complexity and large computation latency, which seriously affects the real-time property of the calculation. Aiming at this problem, based on the deep analysis of the parallelism of calculating gray level co-occurrence matrix this paper proposes a parallel algorithm and parallel system architecture for calculating GLCM. Utilizing the parallel calculation possibility of image pixels, multiple Processing Elements （PEs） are designed to carry out parallel calculation for multiple pixels, which can achieve rapid calculation of GLCM. Because of the limited hardware resources of the embedded reconfigurable system, besides the optimization of GLCM calculation performance, this paper also achieves the optimization of the resource consumption. A new optimal objective function is proposed through analyzing the computation latency and resource consumption, and the optimal design that achieves the trade-off between calculation performance and resource consumption is realized.

关 键 词：灰度共生矩阵 并行计算 优化设计 计算性能 资源消耗 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]