Dehazing Based on Deep Blind Denoising and Dark Channel Prior
Peng Yuan, Yijie Zheng · 2022
Imaging quality is often significantly degraded under hazy weather condition and noise interference. This paper makes an investigation of the dehazing and denoising effects for the single image. The goal is to get high quality image from degraded image by denoising and dehazing. In this paper, we propose an approach to dehazing with noise reduction based on deeply learning. First implement image denoising based on the training data set of the convolutional neural network framework. After obtaining the initial transmittance through the dark channel a priori, soft segmentation of the sky region is performed and the transmittance is optimized. Finally, the smoothing function of the guided filter is used to correct the boundary between the sky area and the non-sky area to restore a haze-free image. Experiments on benchmark images demonstrated that our method achieves superior performance over current state-of-the-art methods, yet keeps efficient and easy to use.