IR based color image preprocessing using PCA with SVD equalization

In this paper we presented a color image enhancement model to overcome the drawbacks associated with illumination-reflectance model of color image enhancement. In this work a new color image enhancement technique based on the Principal Component Analysis (PCA) and singular value decomposition is proposed and comparative analysis is made with IR based model using discrete wavelet transform (DWT) & SVD and Retinex model. The real color image is transformed from RGB to HSV space which is an orthonormal transform between achromatic and chromatic components. The chromatic component is decomposed in to illumination and reflectance using Homomorphic filtering and the reflectance image is accounted for the variation in brightness and is decomposed into four Principal components using (PCA) which involves decomposition of an image into feature based low frequency and high frequency sub bands. Estimates of singular value matrix are carried on low frequency which accounts for contrast of the image, and then modified reflectance is achieved from SVD equalized principal component. The experiment results reveal that the proposed method shows that the color images are enhanced with details preserved and `halos´ restrained. To indicate the impact of enhancement of true color images quantitative measurements like discrete entropy, relative entropy and quality metrics are computed.