适用于丢包信道的多描述图像量化方法

摘要

为提高不可靠网络条件下传输图像的重建质量，提出一种结合遗传算法和小波变换的多描述图像编码方法。利用离散小波变换（DWT）将源图像分解为4个重要性不同的子带，通过自适应遗传算法优化的量化步长对子带系数进行均匀量化，增大引入相关后各个图像描述中的低频信息比重。3组测试图像的仿真分析结果表明，在丢失2或3数据包时，相校基于固定步长的MDTC． DWT／UQ方法，该方法具有更好的鲁棒性，用假设检验的方法进一步验证了该方法的有效性。%To improve the quality of reconstructed image under unreliable network condition,a multiple description transform coding for image that combined the genetic algorithm and wavelet transform was presented.Discrete wavelet transform (DWT) converted a still image into four descriptions in which essentialities were different.Subband was uniformly quantified using adap-tive genetic algorithms (GA)so that the low frequency information of image was protected.Experimental results show when compressing to the target bit rate,compared with traditional uniform quantization with a fixed step size,the proposed method has more improvements both in the obj ective evaluation of PSNR and subj ective visual effect in case of packet lose.