Image Enhancement Based on Reinforcement Learning and Zero-DCE
Against the backdrop of the continuous evolution of deep learning technology,image enhancement technology have garnered significant attention in the field of image processing.While traditional methods have made strides in impro-ving image quality,they often fall short in satisfying the image enhancement demands of complex scenarios such as high dynamic range,high noise,and low contrast.In this challenging environment,this paper adopts an innovative approach to image enhancement by cleverly decomposing the image into illumination and reflection components,achieving superior en-hancement results.By integrating Retinex,Zero-DCE,and reinforcement learning technology,the approach achieves not only enhanced image quality and visibility but also significant effects in handling complex scenes.The combination of clever reinforcement learning and composite loss functions contributes to a more pronounced image enhancement effect.The im-plementation based on the Retinex theory further strengthens the overall algorithm's performance.Additionally,by incor-porating the Zero-DCE method,the approach effectively distinguishes image enhancement needs among different scenes through deep curve estimation of illumination and reflection components.
万方数据
维普
