Fast directional discrete cosine transform for image compression

Abstract

A fast directional discrete cosine transform (FD-DCT) is proposed for efficient representation of anisotropic edges in images. The transform is performed on the predefined direction lines similar to the intraprediction mode in H.264. Comparing to the directional discrete cosine transform (DDCT) now available, no interpolation is needed in FDDCT; thus, the amount of computation decreases by 80%. Simulation results indicate that the peak signal-to-noise ratios of images compressed using FDDCT are > 1 db higher than those using DDCT. © 2010 Society of Photo-Optical Instrumentation Engineers.