Enhancement of low-light images using Sakaguchi-type function-based cost-effective filtering

Even though improving low-light (LOL) images is essential for a variety of applications in computer vision, recent methods often fail to maintain natural image quality and similarity with the quality of an ideal image. This paper presents the Sakaguchi-type function-based cost-effective filtering (SFCEF) method to enhance LOL images. We take linear combinations of the coefficient bounds derived from a Sakaguchi-type function class and Gegenbeur polynomial-based geometric function class to generate a convolution filter for the proposed method. The SFCEF method convolutes an LOL image with the filter to enhance the image. We implement the proposed method with some cutting-edge deep learning and traditional image enhancement methods in the enhancement of challenging LOL images containing noise and artifacts from some LOL datasets to compare their performances. The performance of the SFCEF method is validated regarding visual assessment and numerical measures, like AMBE, contrast improvement, NIQE, SSIM, PSNR, and corresponding bit-plane to bit-plane similarity. The results show the SFCEF method’s substantial advancements over the cutting-edge methods, providing natural color and similarity with the quality of an ideal image with enhanced contrast and denoised quality. The MATLAB® code of the proposed method is publicly available at: https://doi.org/10.13140/RG.2.2.16448.44807 .