基于快速方向重叠变换的图像压缩

传统二维DCT无法稀疏表示除水平或垂直方向外的边缘,而具有强方向表示能力的方向预测离散余弦变换(DPDCT)计算复杂度过高.针对这些问题,快速方向重叠变换(FDLT)沿给定的方向模式进行变换,避免了DPDCT中的插值运算,并进一步集中分散在变换块间的能量,可以快速、稀疏地表示图像中各向异性边缘信息.此外,FDLT通过设计块边界提升,保证了算法完全重构.实验表明,FDLT计算复杂度不超过DCT的2倍.采用同样的编码方法,基于FDLT的压缩图像与基于DCT的压缩图像相比,峰值信噪比可提高0.5dB以上,而且图像边缘细节更加清晰、完整.

Fast Directional Lapped Transform Based Image Compression

Traditional discrete cosine transform (DCT) can only sparsely represent the horizontal and vertical edges in images, and the computation complexity of directional prediction DCT (DPDCr), which is of ability to represent more directions, is much higher. To overcome these shortcomings, the fast directional lapped transforms (FDLT) is proposed, in which the transform is performed on the predefined direction mode and the energy in edges lying across blocks is compacted further. In Comparison with DPDCr, FDLT needs no interpolation. So FDLT can sparsely represent the anisotropic edges in images much faster. Furthermore, special lifting algorithm is designed to ensure the perfect reconstruction. The computation of FDLT is no more than 2 times of DCT's. Coding with the same set partition method, PSNR of images compressed with FDLT is 0.5dB higher than that with Dcr. FDLT based compression also achieves clearer edges and details in the reconstructed images.