A fast algorithm for low-memory embedded wavelet-based image coding without list
Poonlap Lamsrichan · 2011
This paper proposes a new image coding technique with low-memory usage and fast encoding/decoding time. In this embedded coder, the wavelet coefficients are tested with adaptive threshold and encoded into bit-stream with wavelet difference reduction (WDR) technique. However, the original WDR is a list-based coder and needs quite a lot of memory during the encoding/decoding process. Our new technique is non-list coder together with subband selective algorithm. Non-list has reduced the memory requirement to only 1/4 times of the original image (from the 4-8 times of the list-based WDR). The subband selective technique reduces the number of coefficients to be tested in each bit-plane of significant pass thus computational time can be saved considerably, especially in high threshold bit-plane. In order to have the coder with fast algorithm, simple raster scan is used in this coder. Without any prediction scheme, the PSNR performance of the proposed coder is slightly better than the list-based WDR. The simple prediction scheme increases PSNR to the same level as SPIHT. Nevertheless, the favor of this coder is low-memory and fast computation speed. The comparison of computational time between this non-list and list-based WDR is shown in the paper.