Image coding based on zero-crossing and energy information

The purpose of this paper is twofold first to present two other variants of the algorithm shown by Basso, Geurtz, and Kunt (see SPIE-VCIP 93, volume SPIE 2094-I, p.976-986, Boston, USA, Nov. 8-11, 1993) that we call data restoring and error distribution algorithms. Second to discuss a new coding philosophy, based on modelization of the sign of 2D bandpass signals. The data restoring algorithm improves the reconstruction quality and convergence speed while the error distribution one can convergence faster with an acceptable reconstruction quality. In the proposed system a bandpass 2D signal is viewed as a baseband signal that is modulating a carrier. The great part of the information relative to the carrier component is known a priori. Furthermore it constitutes a major part of the sign information of the 2D signal. The method models the sign of carrier component and subtracts it from the sign of the original signal. This technique is applied in a subband based still image coding system. The image is decomposed in subbands that are modeled in terms of the carrier and modulation components. For each subband the sign of logical XOR between a carrier refined model and the original subband is coded, together with the carrier model parameters and refinement parameters. Interesting coding results have been obtained