基于仿射不变的分布式多时相图像压缩

Distributed multitemporal image compression based on affine invariance

针对星载图像压缩存在的编码端资源有限、计算和存储能力有限等问题,提出了一种基于仿射不变的分布式多时相图像压缩方案。首先将图像分为K块和WZ块,K块直接进行JPEG2000编解码,将解码后的K块在解码端进行最相似图像检索找到图像大致位置,并将最相似图像与K块做图像配准,再进行误配点剔除得到其单应性矩阵,利用相邻块的相对位置的一致性,将所得单应性矩阵应用到相邻块上,并将其进行色差校正以得到边信息;WZ块采用分布式编码方案,利用K块变换后所得的边信息图像辅助WZ块解码。实验表明,本文所提方法的边信息质量有明显提升,无损压缩性能优于传统编码方案,同时本文在有损压缩方面的性能也要优于目前主流方法。

Aiming at the problems of limited resources at the encoder,and limited computing power in the compression of spaceborne images,a distributed compression scheme based on affine invariant features for multitemporal images is proposed.First of all,the image is divided into K block and WZ block.The K block is directly compressed by JPEG2000,and the most similar image is queried at the decoder,the most similar image is registrated with the K block to find the approximate position of the image,and then delete the redundant points to obtain its homography matrix.After that,the homography matrix is applied to the relative block because of the consistency of the relative position,the registered image is color compensated to get better side information;meanwhile,the WZ block uses a distributed coding scheme,the side information image is used to help the WZ block decoding.This research not only tests the lossless compression effect,but also tests the lossy compression effect.Experimental results show that the proposed scheme can generate higher side information quality,and has better loseless compression performance than tranditional methods.Furtherly,it is demonstrated that the performance of lossy compression is better than that of current popular methods.