基于DSP的实时去雾优化与实现  

Optimization and implementation ofreal-time dehazing based on DSP

作　　者：朱学俊[1] 凌强[1] 李博伦[1] 徐骏[1] 陈仁爱[1] 王传根 邱换春 王庚中 张乾坤 

机构地区：[1]中国科学技术大学自动化系,安徽合肥230027 [2]安徽超远信息技术有限公司,安徽合肥230021

出　　处：《微型机与应用》2015年第22期96-98,共3页Microcomputer & Its Applications

基　　金：国家自然科学基金(61273112)

摘　　要：提出一种针对基于暗原色先验去雾算法的优化方法,分为算法级优化和基于DSP的优化。算法优化中主要采用降采样处理,以及针对YUV格式转化的优化;基于DSP优化中采用针对TI C674x系列的优化方法 ,如编译器优化、浮点转定点、查表法、DMA(直接内存存取)的使用以及线性汇编等。最终针对1080P的YUV格式的交通视频可以达到实时去雾处理,并且有较好的去雾效果。This paper presents a method to optimizethe performance of a dehazing algorithm based on dark channel prior intensity. There are two levels of optimizations, including the algorithm-level and the DSP-level one. The algorithm-level optimization is mainly accomplished by down-sampling and some efficient format conversion techniques of YUV. The DSP-based optimization targets for the optimization on TI C674x, such as the compiler optimization, conversion from floating-point to fixed-point multiplication, realizing division by a look-up table, Direct Memory Access （DMA） and linear assembly programming. Finally traffic videos of 1080P YUV formats can be real-time processed and yield good dehazing performance.

关 键 词：实时 去雾 DSP 优化 

分 类 号：TP368[自动化与计算机技术—计算机系统结构]