Title :
On implementing transforms from integers to integers
Author :
Xin Li ; Tao, Bo ; Orchard, Michael T.
Author_Institution :
Dept. of Electr. Eng., Princeton Univ., NJ, USA
Abstract :
Arbitrary linear transforms are often applied to integer valued data sequences, producing non-integer outputs. In such cases, it is often desirable to approximate the true transform with a non-linear transform producing integer outputs as close as possible to the true transform, while retaining the reversibility property of a transform. This paper proposes a new approach to implement transforms that map integers to integers in a reversible way. The approach aims to develop a reversible implementation that minimizes the average distortion between the integer outputs and the outputs of the true linear transform
Keywords :
channel bank filters; error analysis; filtering theory; matrix algebra; quantisation (signal); sequences; signal processing; transforms; 2D real valued matrices; approximation error; average distortion minimisation; filter banks; image procesing; integer outputs; integer to integer transform; integer valued data sequences; linear transforms; noninteger outputs; nonlinear transform; reversibility property; reversible implementation; signal processing; Density measurement; Discrete cosine transforms; Entropy; Image coding; Image processing; Lattices; Loss measurement; Nonlinear distortion; Quantization; Signal processing;
Conference_Titel :
Image Processing, 1998. ICIP 98. Proceedings. 1998 International Conference on
Conference_Location :
Chicago, IL
Print_ISBN :
0-8186-8821-1
DOI :
10.1109/ICIP.1998.999075