Title :
Image compression using adaptive lapped transforms
Author_Institution :
Dept. of Phys. & Comput., Wilfrid Laurier Univ., Waterloo, Ont., Canada
Abstract :
A new lapped block transform coding method is proposed which adapts on a block-by-block basis across an image. It is an extension of the mixture of principal components (MPC) method. Instead of a non-overlapping blocking projection, an overlapping projection is used to partition the image into blocks. When applied to image compression, the use of the adaptive lapped MPC method results in better rate-distortion performance relative to the standard JPEG method and the KLT. Further, the distortion is free from the block-effects apparent in non-overlapping methods and sinusoidal artifacts as those introduced by fixed, sinusoidal-based transformations such as the DCT
Keywords :
adaptive codes; data compression; image coding; rate distortion theory; transform coding; MPC method; adaptive lapped transforms; image compression; image partitioning; lapped block transform coding method; mixture of principal components method; overlapping projection; rate-distortion performance; Current measurement; Image coding; Karhunen-Loeve transforms; Neural networks; Nonlinear distortion; Performance evaluation; Physics computing; Principal component analysis; Statistics; Transform coding;
Conference_Titel :
Electrical and Computer Engineering, 1997. Engineering Innovation: Voyage of Discovery. IEEE 1997 Canadian Conference on
Conference_Location :
St. Johns, Nfld.
Print_ISBN :
0-7803-3716-6
DOI :
10.1109/CCECE.1997.608269