Image Restoration and Enhancement Using Blind Estimation of Amplitude Distortion

Abstract—

Image restoration and correction are important steps for video analysis applications. This paper considers a new approach to assessing the characteristics of amplitude distortions in an image and constructing a method for correcting them. A “blind” method for determining the distortion function is developed and an algorithm for restoring and enhancing an image distorted by an unknown nonlinear amplitude transformation is proposed. Based on the signal model, it is shown that the source of information for the analysis of an amplitude distortion should not be the entire image, but only the boundary areas between its constituent objects. The concept of the function of local contrasts is introduced and the hypothesis about the form of this function in the absence of distortions is put forward. Based on this, a method of blind estimation of amplitude distortion and an algorithm for automatic image correction are developed. The principal feature of the proposed approach is that, as a result of the transformation, the brightness ratios between the image objects are preserved. Two possible variants of applying the algorithm to color images are proposed. Qualitative experiments demonstrated the effectiveness of the proposed method. The developed algorithm can be used in conditions when any knowledge about the distortion, to which the image is subjected, is absent, and only the resulting distorted image itself can serve as the source of information.