Image coding with discrete cosine transforms using efficient energy-based adaptive zonal filtering

Efficient 2-D discrete cosine transform (DCT) coding schemes based on the utilization of windows of various geometries whose sizes are adaptively controlled, are presented. In the frequency domain, zonal masks are applied to the transformed coefficients. Masks that generate the minimum number of coefficients and satisfying specific energy constraints are utilized. The coefficients are then quantized based on a scheme specific to the type of zonal mask employed. Since the sequential minimization of coefficient error and quantization error does not necessarily imply minimization of total error, a method of minimizing the two errors simultaneously is also developed. Results are compared to those from JPEG, obtaining similar compression. In this work, we omit parameter estimation, non-uniform or vector quantization, and variable bit rate coding, to focus just on the new algorithm. Including the latter option, higher compression can be achieved