Complex Valued Independent Component Analysis for Image Enhancement

Abstract

Independent Component Analysis (ICA), a category of blind source separation, can be effectively used for extraction of unknown independent source signals from signal mixtures in a wide range of signal processing applications. For example, ICA based methods can be applied in the areas of biomedicine, surveillance, face recognition and financial analysis. In this study, a new ICA based image equalization method is proposed. Our equalization method can be used to enhance the image quality by reducing shadowy and dark areas. The proposed method employs complex valued ICA for extracting equalized intensity component from the RGB components of an image. The extracted intensity component is then used to replace the value component of the hue-saturation-value (HSV) representation of the original image. Further, in order to identify the correct independent components that are used to generate the final equalized image, a new technique is proposed. The new image equalization method presented in this work could produce superior results and therefore, it is useful for implementing low cost vision based image enhancement applications.