Lecture Notes in Computer Science 6691

Compression of wavelet coefficient sign has been assumed to be inefficient for a long time. However, in the last years several proposals have been developed and, in fact several image encoders like JPEG 2000 include sign coding capabilities. In this paper, we present a new sign coding approximation using a genetic algorithm in order to efficiently predict the sign of wavelet coefficients. We have included that prediction in a fast non-embedded image encoder. Preliminary results show that, by including sign coding capabilities to a non-embedded encoder, the compression gain is up to 17.35%, being the Rate-Distortion (R/D) performance improvement up to 0.25 dB.