Color and white balancing in low-light image enhancement

Images acquired during night either appear too dark or either seem biased towards the color of lighting source. This paper proposes a low-light image enhancement technique that simultaneously preserves contrast and removes color saturation from an input image. Initially low and high frequencies are extracted from the low light image. Color balancing is performed on low frequencies using a sigmoid function. This is followed by chromatic balancing which is done using constrained linear least square minimization in YCbCr color space. The high frequencies are added back with respect to the ratio of enhanced and original low frequencies. The proposed method improves input contrast and object prominence while reducing artifacts as compared to state-of-art low light image enhancement methods.