基于张量Tucker分解的彩色图像压缩

Compression of color images based on tucker-tensor

传统图像压缩方法基于向量数据进行压缩处理,破坏了高维数据的空间本征结构.为了克服传统方法的不足,提出了将彩色图像表示成高维数据张量(A^(I_1×I_2×I_3))形式,利用张量Tucker分解,取分解后的最大子张量及其对应的特征向量,并将其进行量化编码实现图像压缩.大量实验结果表明,在相同压缩比下,本文压缩后重构图像的峰值信噪比(PSNR)与传统JPEG压缩方法相比具有较大优势,并且在视觉上本文方法重构图像的颜色信息损失量较小.

Traditional image compression methods handle vectored data to compress, but the process undermines the spacial intrinsic structures of high dimensional data. In order to overcome the shortcomings of traditional methods, the authors presented a novel method of color image compression. In this paper, the color images were encoded into tensors （ AI1×I2×I3 ）. The authors did the tucker decomposition of tensor to get the largest Kn sub-tensors and their eigenvectors, and then used Huffman coding to compress the color images. Experimental results show that at the same compression ratio, the Peak Signal to Noise Ratio （PSNR） of the reconstructed images of the authors＇ method are much better than the JPEG compression, and the authors lose less color information visually.