基于DM642融合系统的A Trous小波实时图像融合算法  

A Trous wavelet real-time image fusion algorithm based on DM642 fusion system

作　　者：许廷发[1] 秦庆旺[1] 倪国强[1] 

机构地区：[1]北京理工大学信息科学技术学院光电工程系,北京100081

出　　处：《光学精密工程》2008年第10期2045-2050,共6页Optics and Precision Engineering

基　　金：国家自然科学基金资助项目(No.60672082);武器装备预研基金项目(No.51476040305BQ0101)

摘　　要：给出了一种基于DM642融合系统的A Trous小波实时图像融合算法。选用适合视频图像处理的TI最新高性能定点数字信号处理器DM642作融合系统核心处理器,利用DM642的600 MHz高速运算能力,EDMA和EMIF的高速数据传输能力,以及它的视频编/解码单元无缝连接的灵活可配置的视频端口,构建了双波段图像融合系统平台。在此系统上对可见光图像进行A Trous小波变换,提取出不同分解层的特征信息,将特征信息与红外图像进行融合。融合后的图像保持了可见光图像的线条及边缘特征,还融合了红外图像的目标特征。仿真实验结果表明,在融合系统平台上实现双波段图像实时融合算法只需39.46 ms左右,满足25 frame/s的工程需求。A Trous wavelet real-time image fusion algorithm based on DM642 fusion system is presented in this paper. By taking the DM642,a high performance fixed-point DSP for video and image processing, as the kernal processor in this fusion system, a dual-band image fusion system platform with the advantages of the high operation speed of 600 MHz, the high data transfer rates with EDMA and EMIF, and the configurable video ports capable of glueless I/F to video decoder/encoder is constructed. A Trous wavelet transform is used to perform the decomposition of visible image at several scales in the system,from which the image details are extracted to fuse into the infrared image. The fused image shows that the discriminating lines and feature edges of visible image are preserved, also the point target features of infrared image are fused. The experimental simulation results indicate that the dual-band image fusion algorithm expends time of 39.46 ms, which meets the engineering requirement of 25 frame/s.

关 键 词：图像融合 A TROUS小波 实时处理 DM642融合系统 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]