Dual x-tree wavelet image coding
Li Li, Canhui Cai · 2010
The lack of directional selectivity has harmed the performance of traditional discrete wavelet transform based image coding, especially when the original image includes multidirectional textures. To offer a better compression ability for multidirectional textural images, a novel wavelet image coding scheme, called dual x-tree wavelet image coding is proposed in this paper. First, the 2-D dual-tree discrete wavelet transform (DDWT) is performed on the input image. Then the noise shaping procedure is exerted on the transform coefficients to get their sparse representation. Finally, an improved x-tree image coding algorithm is applied to encode the coefficients. Different from the common used noise shaping, we adopt a pruning phase to the procedure to make the coefficients better fit our dual x-tree encoder. To make good use of the strong correlation between two wavelet trees produced from DDWT, the dual wavelet trees are jointly encoded to improve the coding performance. Simulation results have demonstrated that the proposed algorithm achieves about 0.5dB gain over state-of-the-arts for multidirectional textural image at low bitrate.