中文摘要：

针对JPEG的中低码率压缩图像即高压缩率图像存在较严重的块效应以及量化噪声,提出了一种对JPEG标准压缩图像进行优化的重建-采样方法.该方法对JPEG压缩图像采用三维块匹配算法（BM3D）进行去噪,去除图像中存在的块效应和量化噪声,进而提高超分辨率重建的映射准确性,再使用外部库对去噪后图像进行基于稀疏表示的超分辨率重建,补充一定的高频信息,最后对重建后的高分辨率图进行双三次下采样,得到与原始图像大小一致的图像作为最终优化图像.实验结果表明,该方法在中低码率情况下能够有效地提高JPEG压缩图像的质量,对高码率压缩图像也有一定效果.

英文摘要：

Considering the low bit-rates compression images of JPEG present serious block artifacts and quantization noise, a method of reconstruction down-sampling which can optimize JPEG and improve the JPEG compress image quality is proposed. Firstly, BM3D is used to denoise the JPEG compressed image to remove block artifacts and quantization noise, in order to improve the accu- racy of super-resolution reconstruction mapping. Secondly, the sparse representation based super-resolution image reconstruction is applied to reconstruct the denoised image to add some high-frequency information, the dictionary used in reconstruction is trained in advance. Finally, down-sampling the reconstructed image, obtain the final decoding image of the same size as original image. Experimental results show that when the bit rate is low, the proposed method achieves significantly PSNR increase compared with JPEG at the same bit rate, and also gains some PSNR increase for high bit-rate compress image.