A Low-light Image Enhancement Algorithm Based on Optimized Multi-illumination Fusion

Low-light images blur information and make it difficult to extract details. We propose a Low-light image enhancement algorithm based on optimized multi-illumination fusion. It can improve the over-enhancement of the high dynamic range of the image and the under-enhancement of some dark areas in the existing image enhancement algorithms. The initial illumination is extracted based on the Retinex algorithm, and the smooth illumination components are obtained by the optimization method. An improved CRF model is introduced to obtain the illumination component that suppresses overexposure based on the morphological closing operation and guided filtering on the initial illumination. We design a brightening function as the third illumination component to enhance the global brightness to display image details. The fused lighting components are formed through the weight matrix. The resulting image is obtained by combining the fused illumination component with the reflection component. The enhanced image restores image details to the greatest extent and suppresses overexposure. Good performance was achieved on both subjective and objective assessments.