Low-Artifact and Fast Backlit Image Enhancement Method Based on Suppression of Lightness Order Error

Abstract

Many image enhancement methods have been proposed to improve the visibility of backlit images. Although these methods can effectively improve the visibility of the subject and background compared to standard image enhancement methods, they may result in image quality degradation owing to non-negligible artifacts. In many cases, such artifacts are caused by a significant change in the Lightness Order Error (LOE) between the original and processed images. To address this problem, this paper proposes a low-artifact and fast backlit image enhancement method to effectively improve the visibility of images by suppressing the LOE. The proposed method uses adaptive luminance correction to generate lightnessenhanced images of the dark and bright areas of the backlit image. These images are then fused based on a weight map to calculate the lightness of the output image with a lower LOE. The final output, i.e., the enhanced color image, is obtained by multiplying the input color image by the ratio of the lightness component of the input image to the enhanced lightness component. The experimental results demonstrate the superiority of the proposed method in terms of low artifacts, natural enhancement, and high processing speed based on straightforward processing.INDEX TERMS Backlit image, image enhancement, lightness order error, low-artifact image.

Single Image Enhancement Using Gradient-Norm-Based Tone Curve for Images Captured in Wide-Dynamic-Range Scenes

“…In addition, contrast enhancement by the S-type tone curve can cause black-out of dark areas. For backlit image enhancement, Akai et al generated two images with improved image quality in the dark and bright regions of the image and blended them with a simple alpha blending [24]. To improve the image quality in dark areas, the brightness is adjusted by gamma correction and the contrast is enhanced by S-shaped function.…”

“…Various methods for improving the visibility of an image in which very bright and dark regions exist simultaneously have been proposed [9]- [25]. Some methods use image fusion method [18]- [24], while others use histogram stretching [9], [10], Retinex models [13]- [17] or deep learning models [25]. For improving the visibility of images, these methods brighten dark areas of the input image and enhance the contrast of the image.…”