基于DCT域采样和重建的低码率图像压缩算法

本文构建了一种基于DCT域采样和超分辨率(Super Resolution, SR)重建的低码率图像压缩编码算法。在编码端对原始图像进行分块离散余弦变换（DCT），并提取每个DCT系数块的低频系数，然后再反变换到空间域，从而得到在DCT域下采样的低分辨率（Low Resolution，LR）图像块。用JPEG标准对下采样图像块编解码后,采用基于学习的方法恢复DCT域高频系数，重建出高分辨率（High Resolution, HR）的图像。实验结果表明，在码率较低的情况下，本文算法比JPEG编码标准具有更好的率失真性能；同时，在相同码率下，本文算法重建的解码图像视觉效果更好。

Low Bit Rate Image Compression via Down-sampling and Reconstruction in DCT

In this paper, an image compression algorithm is presented via combining the Discrete Cosine Transformation (DCT) domain down-sampling and super-resolution reconstruction.Firstly, the image is first divided into non-overlapping blocks, and each block is transformed by DCT. Secondly, low frequency coefficients of each block are extracted and inverse-transformed to generate the down-sampled low resolution image. After the down-sampled image is compressed and decompressed by the standard JPEG ,the high frequency coefficients are recovered by the learning-based method in the spatial domain. Experimental results show that the compression algorithm presented in this paper has better?rate-distortion performance than JPEG at low bit rate, and the visual quality of decompressed image by our method is also better at the same coding bit rate.