兴趣区域优先的多尺度压缩感知渐进编码算法  

ROI progressive coding algorithm based on multiscale compressed sensing

作　　者：杜梅[1] 赵怀慈[2] 赵春阳[2,3] 

机构地区：[1]沈阳师范大学软件学院,辽宁沈阳110034 [2]中国科学院沈阳自动化研究所,光电信息处理重点实验室,辽宁沈阳110016 [3]中国科学院大学,北京100049

出　　处：《光电子．激光》2015年第10期2016-2022,共7页Journal of Optoelectronics·Laser

基　　金：中科院光电信息处理重点实验室基金(OEIP-0-201102)资助项目

摘　　要：针对面向目标探测识别的无线图像传输应用,为了解决探测识别任务对图像质量的高要求和无线信道带宽约束之间的冲突,提出一种感兴趣区域(ROI)优先的多尺度压缩感知(CS)渐进编码算法。首先,选取多尺度CS测量技术进行图像压缩,在相同压缩比情况下,可以保持更多的图像的细节信息;其次,针对目标探测识别对重构图像纹理、边缘信息完整保持的需求,选取基于小波的轮廓变换(WBCT)作为CS的稀疏表示方法,可更多地提取方向信息,从而可以较完整地保持边缘轮廓信息;最后,针对重构图像ROI与背景区过度划分、ROI周边相关信息损失过多的问题,给出ROI边界平滑处理的方法,可有效改善重构图像的主观视觉质量。实验结果表明:在ROI压缩率高达0.4的情况下,总的压缩率降至0.18,既保护了ROI信息又显著提高了压缩比,基本满足目标探测识别任务的需求。Aiming at the application of the wireless image transmission for target detection and identifica- tion, in order to solve the conflict between the requirement for high image quality in detection and recog- nition task and the constraint of wireless channel bandwidth, a region of interest （ROD progressive cod- ing algorithm based on multiscale compressed sensing is proposed. First, the technology of multiscale compressed sensing is selected to compress an image in order to preserve more details of the image in the condition of the same compression ratio. Then,wavelet based contourlet transform （WBCT） is chosen as the sparse representation to extract more direction information for the demand of completely maintaining the texture and boundaries of the reconstructed image. Finally, an ROI boundary smoothing method is provided to effectively improve the subjective visual quality of the reconstructed image, for overcoming the deficiency of over differentiating between ROI and background of the reconstructed image and loosing excessive ROI-related information. Experimental results indicate that the total compression ratio Jails to 0. 18 in the case that the ROI compression rate is as high as 0. 4,which both protects the ROI and signif- icantly improves the overall compression ratio. The algorithm can enhance the efficiency of wireless im- age transmission,supplying subsequent detection and recognition task with high quality images.

关 键 词：感兴趣区域(ROI) 渐进编码 多尺度压缩感知(CS) 基于小波的轮廓变换(WBCT) 边 界平滑 

分 类 号：TN911.73[电子电信—通信与信息系统]