Frequency-Adaptive Low-Light Image Enhancement Using Laplacian Pyramid and Blending Scheme

Image processing is essential in many fields, including security, medical image analysis, remote sensing, face detection, and pattern recognition. However, the enhancement of low-light images remains a significant challenge due to issues like detail loss, noise amplification, and limited dynamic range. Existing systems, including global contrast enhancement and histogram equalization, often fall short in preserving image details or introduce visual artifacts under poor lighting conditions. The Laplacian low-light image enhancement technique uses gamma correction to purposefully underexpose the image, which overcomes the difficulty of low-light conditions. Several exposed images are produced when frequency domain performance is improved. These images are then combined using a multi-scale Laplacian blending strategy to improve the output. By manipulating gamma values, a set of images with varying exposures is created, enabling the algorithm to capture a broader range of information in low-light conditions. Extensive experimental evaluations, encompassing qualitative and quantitative analysis, validate the method's effectiveness to preserve details while minimizing artifacts. The results show significant improvements in decreasing the mean-squared value (MSE) and improving the pixel-to-noise ratio (PSNR), especially in scenarios where traditional methods struggle. Optimizing the transformation in the frequency domain emphasizes the potential of multi-scale. Laplacian blending as a valuable tool in addressing challenges posed by low-light conditions, marking substantial advancements in image enhancement.