Design of Self-Adaptive Enhancement System for Infrared-Night-Vision Images

Abstract

This paper presents an efficient infrared-night-vision image enhancement system. This system mainly consists of wavelet-based image preprocessing, low-pass sub-band enhancement, and high-pass sub-bands enhancement. A preprocessing method is adopted to transform the image into low-pass main information sub-band and high-pass detail sub-bands. According to the characteristics of infrared-night-vision images, a self-adaptive dynamic range expansion strategy is performed on low-pass sub-bands to expand the grayscale of image's dark areas. Taking advantage of high-pass sub-bands' directivity, a texture-preserved Gaussian filter is proposed. Afterwards, a noise-suppressed edge enhancement algorithm is used to intensify texture while avoiding noise signal's amplification. The experimental results show that the proposed method outperforms histogram equalization algorithm in terms of perceptual quality. Meanwhile, the discrete information entropy of the images processed by our method is 48% - 58% higher than histogram equalization.