Adaptive Enhancement for non-uniform illumination Images via Pixel-wise Histogram Modification and Color Reconstruction

Non-uniform illumination images usually suffer from degenerated details and colors due to underexposure or overexposure problems. Traditionally, to improve local details, sophisticated mapping functions have been used to pull up dark intensities and pull down highly-bright regions. Inevitably, proper demarcation between underexposure and overexposure regions is fatal for the enhancing effects. In this work, in luminance channel, a pixel-wise demarcation was designed and cooperates with a nonlinear modification to global histogram. Then a color construction technique with color compensation was proposed to reconstruct color image using the modified luminance and the original chromatic information. Finally, in order to compensate for local contrast, a local comparison-dependent technique is employed to RGB channels. Experiments on several types of images show that our approach brings promising details for both underexposure and overexposure regions. Moreover, comparisons show that the proposed method performs well on globally overexposed images, and has the merit of protecting smooth uniform areas.