基于形态学膨胀操作的小波图像比率可分级编码研究

Rate Scalable Wavelet Image Coding Based on Morphological Dilation

图像渐进传输、图像数据库浏览等多分辨率环境下的多媒体应用导致了图像比率可分级性编码算法的产生 ,比如嵌入式零树小波图像编码方法 (EZW) .Servett等人给出了一种基于形态学方法的图像编码方法(MRWD) ,该方法根据图像小波分解后各子带中重要系数的聚类特性 ,利用数学形态学中的膨胀算子直接对各子带中的重要系数进行检测、提取和编码 ,取得了优于 EZW的编码效果 .然而 ,该方法还存在着一些值得改进的地方 .该文在分析和改进 MRWD方法的基础上 ,提出一种基于形态学方法的小波图像比率可分级编码的新方法 ,该方法的主要思想如下 :(1)针对小波图像分解的不同层次子带 ,采用了不同的结构元素来实现各子带重要系数的膨胀操作 ;(2 )对剩余空间的重要系数 ,根据其在不同分解层次上的分布情况 ,采用了直接提取和膨胀操作相结合的编码方法 ;(3)根据小波图像分解最低频子带的特殊性 ,对其进行了单独处理 ;(4)采用了逐次逼近的量化模式 ,使生成的嵌入式码流具有比率可分级的特性 .该方法除了具有 EZW方法的优点外 ,在一定程度上克服了 MRWD方法的不足 ,取得了较好的效果 .

The multimedia applications in a multiresolution environment, such as progressive transmission of image, image database browsing, etc., require rate scalable image coding algorithms, such as EZW (Embedded Zerotree Wavelet). Servetto et al . introduced a method of morphological representation of wavelet data (MRWD) in 1999, which is based on the observation that, in the wavelet transform domain, significant coefficients within each subband are more likely to be clustered and the distributions of significant coefficients in the subbands of the same orientation are similar. It used the morphological dilation operator to predict, extract and encode the significant coefficients within each subband and achieved better performance than the encoding in EZW. However, there are still some deficiencies in MRWD. In this paper, on the basis of discussing and improving MRWD method, a novel approach to rate scalable wavelet image coding based on a morphological representation is proposed. Its main idea is as follows:(1) using different structuring element for morphological dilation to extract and encode the clustered significant coefficients in the different subbands from low frequency to high frequency;(2) using the two kinds of method, encoding only the probable scattered significant coefficients and their positional information and using morphological dilation, to encode the significant coefficients in the each remaining space;(3) using lossless or less loss coding algorithm, such as DPCM, to encode the lowest frequency subband image separately from other highpass subbands so as to preserve its most information because it concentrates most energy of original image;and (4) using the successive approximation quantization to produce an embedded bit stream in order to make the encoder offer scalability in rate. The proposed method not only has some good properties such as EZW encoder, but also improves some deficiencies in MRWD. Simulation results show that it is superior to the EZW and MRWD algorithms in PSNR.