一种基于SPIHT和匹配追踪的图像编码方法——SPMP算法

摘要

为了更好地利用图像的结构特征,提高图像重建的质量,提出了一种基于多级树集合划分(SPIHT)和匹配追踪(MP)的分层图像编码方法--(SPMP)算法.该方法首先采用拉普拉斯金字塔(Laplacian Pyramid)算法将原始图像分解成低频平滑层和高频细节层,然后使用离散小波变换和SPIHT算法编码图像的低频成分,使用基于克隆选择的匹配追踪算法编码图像的高频细节层.实验结果表明,该方法能够产生渐进:PSNR的位流,图像重建质量要明显高于小波图像编码算法.%To make full use of the structural characteristics of original images to improve reconstructed images' quality this paper proposes a novel hierarchical image encoding method based on set partitioning in hierarchical trees (SPIHT) and matching pursuit (MP) algorithm, named the SPMP algorithm. This new method divides the original images into a smooth layer in low frequency and a detail layer in high frequency by using the Laplacian Pyramid algorithm. In the smooth layer, it uses the discrete wavelet to transform the images from the space domain to the frequency domain, and then it encodes the coefficients in the frequency domain by using the SPIHT algorithm. The method adopts the matching pursuit (MP) algorithm to encode the detail layer based on the clone selection algorithm. The experimental results demonstrate that when using the SPMP algorithm, the output bitstream is embedded with the progressive peak signal-to-noise ratio (PSNR), and the reconstructed images quality is significantly better than that obtained using the wavelet transform encoding, even more obviously under the high compression ratio.