Title :
Statistical analysis of the DCT coefficients and their quantization error
Author :
Yovanof, Gregory S. ; Liu, Sam
Author_Institution :
Hewlett-Packard Co., Palo Alto, CA, USA
Abstract :
A detailed probabilistic analysis of the DCT coefficients and the error incurred during the quantization and reconstruction of natural images is presented. The distribution of both the DCT coefficients and their quantization error were modeled with the generalised Gaussian function (GGF) which includes the Gaussian and Laplacian PDF as special cases. The /spl chi//sup 2/ and Kolmogorov-Smirnov goodness-of-fit tests were used to determine the free parameter of the GGF that yields the best fit to experimental data. This analysis has yielded several new results regarding the distribution of the coefficients, their differences and their quantization error. The results of this analysis can improve the design of practical DCT based compression systems.
Keywords :
Gaussian distribution; data compression; discrete cosine transforms; error analysis; image coding; image reconstruction; quantisation (signal); statistical analysis; transform coding; DCT coefficients; Kolmogorov-Smirnov goodness-of-fit test; chi//sup 2/ goodness-of-fit test; generalised Gaussian function; image compression; image reconstruction; natural images; probabilistic analysis; quantization error; statistical analysis; Data analysis; Discrete cosine transforms; Gaussian distribution; Image analysis; Image coding; Image reconstruction; Laplace equations; Quantization; Statistical analysis; Testing;
Conference_Titel :
Signals, Systems and Computers, 1996. Conference Record of the Thirtieth Asilomar Conference on
Conference_Location :
Pacific Grove, CA, USA
Print_ISBN :
0-8186-7646-9
DOI :
10.1109/ACSSC.1996.601095
