压缩感知在感兴趣区域编码中的应用  

Application of compressed sensing on region of interest coding

作　　者：杜梅[1] 赵怀慈[2] 周浩[1] 

机构地区：[1]沈阳师范大学科信软件学院,沈阳110034 [2]中国科学院沈阳自动化研究所,沈阳110016

出　　处：《沈阳师范大学学报（自然科学版）》2016年第4期479-486,共8页Journal of Shenyang Normal University:Natural Science Edition

基　　金：辽宁省科技厅自然科学基金资助项目(2013010420-401)

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与无线信道带宽约束之间的冲突,提出一种基于压缩感知的编码方法。鉴于压缩感知优秀的抗干扰特性,利用其进行图像压缩,并将位平面提升技术引入压缩感知。首先,对图像进行分块压缩感知,然后,对获得的信号进行量化、位平面分解;然后提升感兴趣区域位平面,并分别给出3种不同的感兴趣区域位平面编码方案;最后,在解码端通过解码、阈值迭代法重构,得到感兴趣区域质量优于背景的重构图像。实验结果表明,在相同码率下,重构图像的感兴趣区域PSNR(Peak Signal to Noise Ratio)高于通常的压缩感知编码方法,验证了方法的可行性和有效性。因此,基于压缩感知的感兴趣区域编码方法能够提高无线图像传输效率,从而更好地满足目标探测识别的需求。For the applications of target detection and identification in the wireless image transmission,in order to solve the conflict between the requirement for high image quality for detection and recognition task and the constraint of wireless channel bandwidth,a coding algorithm based on compressed sensing is proposed.Compressed sensing is employed to compress images due to its excellent anti-interference capability,and the bit plane scaling technology is introduced into it,too.First,the blocked compressed sensing signals are quantified and decomposed to bit planes.Then,the bit planes of the region of interest are shifted,and three different coding schemes of the region of interest are put forward,respectively.Finally,after decoding,the thresholding iterative algorithm is used to reconstruct the image whose region of interest is much clearer than the background.The experimental result indicate that under the same bit rate,the peak signal to noise ratio（PSNR）of the reconstructed region of interest encoded by bit plane shifting algorithm is higher than that by the usual compressed sensing coding technology,through which the feasibility and effectiveness of our method is verified.The result is indicated that the region of interest coding algorithm based on compressed sensing can improve the efficiency of wireless image transmission system,and can meet the requirement of target detection and identification better.

关 键 词：感兴趣区域 压缩感知 位平面提升 面向探测识别 

分 类 号：TN919.81[电子电信—通信与信息系统]