摘要
本文提出一种基于JPEG2000标准的Bayer图像高性能RBCR(Remove Bayer Component Relation,RBCR)压缩算法。在RBCR压缩算法中,根据Bayer图像相关性较高的特点,对Bayer图像进行颜色分量分离,得到处理单元子图,对各子图进行1×4整型离散余弦变换,降低Bayer图像中各颜色分量空间域内的相关性;对变换后的各分量DCT系数使用标准JPEG2000算法独立完成小波变换、Tier1编码、MQ编码和率失真斜率计算等,再基于率失真斜率联合截取方法完成各个分量的码流截取,即使用相同的率失真斜率门限值,按照率失真斜率值由高到低的顺序依次完成所有分量编码码块的码流截取,最后各个分量的截取结果再进行独立的码流组织输出。在RBCR算法中通过加入DCT变换降低Bayer图像相关性和对各个分量码流的率失真斜率联合截取,提高恢复图像质量且精确控制了码率。实验结果表明,RBCR算法与各个分量独立压缩方法相比,恢复图像质量得到提升,尤其在4倍的压缩倍数下效果最佳,峰值信噪比平均提高1.814dB,复杂峰值信噪比平均提高2.414dB。可以满足深空探测低复杂度和高质量图像的要求。
Based on the JPEG2000 compression framework,an RBCR compression algorithm for Bayer remote sensing images with high quality was proposed in this study.In this algorithm,the color components of the Bayer image were separated,and the processing unit subgraphs were obtained.Each subgraph undergoes a 1×4 integer discrete cosine transform,so the correlation of the four color component space domain in Bayer images was reduced.The JPEG2000 algorithm was used to perform the wavelet transform,Tier1 coding,MQ coding,and rate-distortion slope calculation on the transformed DCT coefficients;subsequently,the code stream interception of each component was completed based on the rate-distortion slope joint interception method.In accordance with the rate-distortion slope values arranged in descending order to complete the code stream interception of all component encoding code blocks,the interception result of each component was finally subjected to the independent stream organization output.In the RBCR algorithm,by adding the DCT transform to reduce the Bayer imagecorrelation and intercepting the rate-distortion slope of each component stream,the recovery image quality was improved and the code rate is accurately controlled.The experimental results show that,compared with each component independent compression method,the RBCR algorithm has improved image quality.At 4 times the compression ratio,the peak signal-to-noise ratio increased by an average of 1.814 dB,and the signal-to-noise ratio increased by an average of 2.414 dB,which meets the requirements of low computational complexity and high image quality for deep space detection.
作者
雷杰
于露露
罗晓红
李云松
LEI Jie;YU Lu-lu;LUO Xiao-hong;LI Yun-song(State Key Laboratory of Integrated Services Networks,Xidian University,Xi′an 710071,China)
出处
《光学精密工程》
EI
CAS
CSCD
北大核心
2019年第1期191-200,共10页
Optics and Precision Engineering
基金
国家自然科学基金资助项目(No.61502367
No.61501346
No.61701360
No.61571345
No.91538101)
高等学校学科创新引智基地项目资助(No.B08038)
陕西省自然科学基础研究计划项目资助(No.2016JQ6023
No.2016JQ6018)
