基于全景图的图像集压缩用于云存储

当前的图像集压缩方案通常只适用于相似图像,对云中图像集内部的差异性十分敏感。为了解决这个问题,提出了一种基于全景图的图像集压缩方案,该方案主要利用全景图和图像集之间的相关性来减少冗余。首先,在全景图上选取一些关键点,利用直线投影算法由全景图得到各个关键点处的视口图像;然后,对于图像集中的每张图像,选取最匹配的视口图像作为参考;最后,使用基于块的运动补偿来执行图像间的预测编码。实验结果表明,提出的编码方案比JPEG平均节省46.3%的比特,并且实现了与IEEE1857.4帧内编码可比的性能。对于大型图像集,图像具有不同视点和对象但却拍摄于同一场景,方案压缩效果良好,值得深入研究。

Panorama-based Image Set Compression for Cloud Storage

This paper proposes a panorama-based compression scheme which mainly exploits the correlations between panorama and image set.First,it selects some key points on the panorama and obtain the viewport images at each key point using rectilinear projection algorithm.Then,for each image in the set,it searchs the best matched viewport image as the reference.Finally,inter-image prediction coding is performed.Experimental results show that our scheme achieves up to 46.3% bitrate saving over JPEG and acquires a comparable performance compared with the intra coding of IEEE 1857.4.The proposed scheme provides a promising approach to the storage of massive images different in viewpoints.