摘要：AbstractIn this paper, we propose an effective and efficient method to remove haze from a single input image. We first remove fine details from the minimum channel of RGB channels by a low-pass filter and then use it as a rough estimation of the transmission map. Then, we refine the rough estimation by using the saturation channel of HSI color space and local contrast. Based on the atmospheric scattering model, we then obtain a high-quality haze-removed image. Results on a variety of hazy images demonstrate the effectiveness and efficiency of our method.
