基于JPEG2000的遥感图像感兴趣区域编码新算法及其VLSI设计  

A New Region of Interest Coding Algorithm Based on JPEG2000 for Remote Sensing Images and Its VLSI Design

作　　者：王建新[1] 朱恩[1] 刘文松[1] 祁友杰[1] 

机构地区：[1]东南大学信息科学与工程学院,南京211100

出　　处：《电子与信息学报》2016年第4期958-963,共6页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(61271058)~~

摘　　要：为解决空间遥感图像数据量及信道带宽之间的矛盾,该文提出一种基于JPEG2000的感兴趣区域(Region Of Interest,ROI)编码算法。主流的JPEG2000 ROI编码算法难以兼顾ROI质量和系统计算量,且在低码率编码时有完全丢失背景的隐患。该算法通过精确控制各子带中背景系数的精度,使ROI分配到更多码流。并引入了人眼视觉特性,使较少的背景码流产生尽量好的视觉效果。另外,根据该算法提出了针对矩形ROI的超大规模集成电路(VLSI)设计,此设计经过简单调整,亦可适用于主流的ROI编码算法。测试结果表明,该算法在ROI质量和重建图像视觉效果上均表现优异,且支持任意形状ROI编码,兼容JPEG2000协议。该VLSI设计仅使JPEG2000系统运行时间增加一个周期,具有极高的吞吐率,可满足实时处理要求。To resolve the conflict between the amount of image data and the channel bandwidth, a Region Of Interest（ROI） coding method based on JPEG2000 is proposed in this paper. The mainstream methods for JPEG2000 ROI coding have a difficulty in balancing ROI quality and system computation, and have a hidden trouble in losing background completely at a low bit rate. The proposed method achieves ROI coding by precisely controlling the accuracy of background coefficients in each subband. The human visual property is introduced to obtain better visual effect with limited background codestream. Also, the Very Large Scale Integration（VLSI） design of the proposed method for the rectangle ROI is proposed. This design can be suitable for the mainstream ROI methods for JPEG2000 by some necessary adjustments. Experimental results show that the proposed method obtains an excellent performance in ROI quality and visual effect of the reconstructed images, supports arbitrary ROI shape coding, and is compatible with the standard. The VLSI design only adds one clock cycle on the coding time of JPEG2000 system, has a high throughput, and satisfies real-time applications.

关 键 词：遥感图像 感兴趣区域编码 JPEG2000 人眼视觉特性 超大规模集成电路 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置] TN492[自动化与计算机技术—控制科学与工程]