Image compression algorithms based on the discrete wavelet transform are nowadays very popular for their power and other properties. However, classical decomposition of the image with one static mother wavelet is generally far from optimal. This paper introduces a lossless still image compression method based on spatially adaptive wavelet transform, where the mother wavelet is varying along the signal to achieve better energy conservation. The decision-making
process, which is responsible for mother wavelet selection, works in dependence on local pixel values and internal statistics. New mother wavelet is determined for every coded pixel (deccorelation step). The preliminary results of this method on several images of different types show, that the compression ratio of proposed method is comparable to the JPEG2000 algorithm with static 5/3 wavelet filter. Our intention is to apply this new adaptive method to the compression
of astronomical images and optimize carefully the decision-making process according to this type of images. Since these images have very specific properties, we expect that our method will overcome JPEG2000 and also recent compression algorithms used in astronomy (HCompress, PMT or FITPRESS).
"Spatially adaptive DWT for image compression", Proc. SPIE 6180, Photonics, Devices, and Systems III, 618021 (18 April 2006); doi: 10.1117/12.675850; https://doi.org/10.1117/12.675850