【摘要】 由量化框架扩展系数所产生的多描述图像编码,在差错信道下具有良好的图像恢复性能。该文提出一种基于框架扩展的高效多描述图像编码。首先构造一种在量化噪声下具有均方误差意义上最优框架重建的均匀紧支框架。然后以小波零树作为框架扩展的基本运算单元,在小波域对不同子带信息采用不同维度的框架进行扩展。通过不同维度的框架扩展实现对不同子带信息的不同保护。框架扩展后的数据经量化编码形成多描述码流。实验结果证明,相较于其他几种基于框架扩展的方法,该文方法在同样的信道条件下具有更好的图像差错恢复性能。更多还原

【Abstract】 Multiple description image coding based on quantized frame expansion performs very well over error-prone channels. In this paper, an efficient multiple description image coding based on frame expansion is proposed. Firstly, an optical uniform tight frame is designed, which can minimize the mean squared error with quantization. Then, basic unit of frame expansion is constructed from wavelet zerotrees and different frames are applied to coefficients in different sub-bands. In that way, different sub-bands gain different protections during the transmission. After quantization and entropy coding, multiple description bit streams are generated from frame expansion. The experiment results prove that compared with other methods based on frame expansion, the scheme can obtain better recovery images under the same channel condition.更多还原