Mojtaba OMAM AMINI, Farah AZAR TORKAMANI

Noise adjusted version of generalized principal component analysis

Principal component analysis (PCA) is a well-known tool in image processing, especially in dimension reduction schemes. Since using PCA is based on the vector representation of the image, the spatial locality of pixels in the image is not considered. However, generalized PCA (GPCA) could be applied to images in two dimensional spaces. Both schemes do not consider the noisy case of signals. Noise adjusted PCA (NAPCA) tries to find new coordinates for signal representation based on signal to noise ratio (SNR) maximization. In this paper we generalized noise adjusted GPCA to benefit the advantage of GPCA and SNR maximization case of NAPCA in two dimensional spaces. The experimental results on the huge databases show its reliability.

PDF

___

[1] Jolliffe IT. Principal Component Analysis. 2nd ed. New York, NY, USA: Springer, 2002.
[2] Vidal R, Ma Y, Sastry S. Generalized principal component analysis (GPCA). IEEE T Pattern Anal 2005; 27: 115.
[3] Jieping Y, Ravi J, Qi L. GPCA: An efficient dimension reduction scheme for image compression and retrieval. In: 2004 Knowledge Discovery and Data Mining Conference; 2225 August 2004; Seattle, WA, USA: SIGKDD04, pp. 354363.
[4] Green A, Berman M, Switzer P, Craig MD. A transformation for ordering multispectral data in terms of image quality with implications for noise removal. IEEE T Geosci Remote 1988; 26: 6574.
[5] Lee JB, Woodyatt AS, Berman M. Enhancement of high spectral resolution remote sensing data by a noise-adjusted principal components transform. IEEE T Geosci Remote 1990; 28: 295304.
[6] Roger RE. A fast way to compute the noise-adjusted principal components transform matrix. IEEE T Geosci Remote 1994; 32: 11941196.
[7] Chang CI, Du Q. Interference and noise-adjusted principal components analysis. IEEE T Geosci Remote 1999; 37: 23872396.
[8] Wang Z, Bovik AC, Sheikh HR, Simoncelli EP. Image quality assessment: from error visibility to structural similarity. IEEE T Image Process 2004; 13: 600612.