Wavelet Lossy and Lossless Image Compression Techniques - Use of the Lifting Scheme

Lossless compression schemes with progressive transmission capability play a key role in critical data storage, retrieval and transmission, when a large volume of data must be handled fast, without distortions, over large networks with limited bandwidth. The standard wavelet compression techniques, even if loss-less in principle, do not reconstruct exactly the original image because of the rounding of the floating-point wavelet coefficients to integers caused by the coding. The use of the lifting scheme allows to generate truly loss-less non-linear integer-to-integer wavelet transforms. This paper presents two such wavelet based compression techniques able to operate in the lossless mode. Two methods more effective than the classical zerotree coding are proposed for the coding of the significance maps. Experimental results obtained on natural and multimodal medical images show that the proposed compression techniques provide better rate-distortion performances than the embedded zerotree coder combined with the integer wavelet transform. Both schemes produce a losslessly compressed embedded data stream, with the advantage of supporting progressive refinement of the decompressed images. Therefore they are good candidates for telematics applications that require fast tele-browsing of large image data-bases while retaining the option of lossless transmission and archiving of the images.