Single Image Enhancement Using Inter-channel Correlation

This paper proposes a new approach for enhancing digital images based on red channel information, which has the most analogous characteristics to invisible infrared rays. Specifically, a red channel in RGB space is used to analyze the image contents and improve the visual quality of the input images but it can cause unexpected problems, such as the over-enhancement of reddish input images. To resolve this problem, inter-channel correlations between the color channels were derived, and the weighting parameters for visually pleasant image fusion were estimated. Applying the parameters resulted in significant brightness as well as improvement in the dark and bright regions. Furthermore, simple contrast and color corrections were used to maintain the original contrast level and color tone. The main advantages of the proposed algorithm are 1) it can improve a given image considerably with a simple inter-channel correlation, 2) it can obtain a similar effect of using an extra infrared image, and 3) it is faster than other algorithms compared without artifacts including halo effects. The experimental results showed that the proposed approach could produce better natural images than the existing enhancement algorithms. Therefore, the proposed scheme can be a useful tool for improving the image quality in consumer imaging devices, such as compact cameras.