不同的小波分解法對影像壓縮效果的影響

Abstract

小波（Wavelet）轉換已經成為近年來影像壓縮的主流，除了最有名的JPEG2000採用DWT取代傳統的DCT，也有許多以DWT為基礎的影像壓縮演算法，如EZW，SPIHT，SLCCA，MRMD，等等，都可以達到相當不錯的效果。本篇論文採取類似小波包（wavelet packet）的分解方法，將不同類型的影像搭配不同的wavelet family做轉換，使影像被分解成數個子頻帶(sub-band)，再依每個子頻帶的特性，搭配不同的小波係數做進一步的分解。我們也分析與比較了不同的分解層級數，濾波器階數，壓縮率，和不同的影像內容之間的關係，藉由探討這些性質，我們可以決定是否進行子頻帶分解以得到更進一步的壓縮效果。

In recent years, wavelet transform had become the main stream of image compression. Most famed compression standard JPEG2000 use DWT instead of traditional DCT. Except for JPEG200, there are also many image compression algorithms which are based on DWT, like EZW, SPIHT, SLCCA, and MRMD. They also obtain good performance and results. This paper uses a decomposition method which is similar to “wavelet packet”. To decompose different images with different wavelets, and further divide high frequency subbands of the original image by the characters of subbands. We also analysis and compare the relationships between the numbers of decomposition, filter orders, compression ratios, and different image contents. By investigating these properties, we can decide whether a subband will be decomposed or not in order to get improved performance.