DSP图像处理的程序优化  

Code optimization of DSP image processing

作　　者：黄德天[1] 陈建华[2] 

机构地区：[1]中国科学院长春光学精密机械与物理研究所,吉林长春130033 [2]中国科学院研究生院,北京100039

出　　处：《中国光学与应用光学》2009年第5期452-459,共8页Chinese Optics and Applied Optics Abstracts

摘　　要：中值滤波在滤除噪声的同时,能够很好地保护图像边缘,是最常用的图像预处理方法之一。本文针对中值滤波算法排序量多、处理速度慢的缺点,提出了对中值滤波进行优化的方法。介绍了中值滤波的原理,使用基于TMS320C6000系列DSP的C程序优化方法对中值滤波代码进行优化,包括使用编译器选项、内联函数、软件流水和循环展开等。介绍了算法优化的基本步骤和具体方法,在CCS2软件平台下选择C64DSP做仿真,实验结果表明,优化后代码的运行时间从优化前的8.37ms减少到2.91ms,提高了执行效率,能够满足当今实时图像处理的需要。The median filter is one of the image preprocessing methods most in use, for it can not only filter the noise, but can also protect the image edge. For a lot sorting processing and low processing speed existing in the algorithm, the optimized algorithms for the median filter were presented in the paper. The principle of median filter was introduced and the C code optimization based on the TMS320C6000 Series DSP, including compiler options, intrinsics, software pipelining, loop-opened, etc. , were used to optimize the code of the median filter. Experimental results indicate that the running time of the optimized code has decreased from 8.37 ms to 2.91 ms as compared with that of the original coding method. It is concluded that the proposed algorithms can enhance the efficiency of the code execution, reduce the processing time of median filter and can meet the requirement of real-time image processing further.

关 键 词：中值滤波 程序优化 内联函数 软件流水 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]