Dark channel prior principle and morphology based horizon detection method

Abstract

In order to detect the horizon line from the input images, especially the images with fog or other interference information, this paper proposes an effective method, which selects the dark channel image as the image preprocessing object. The dark channel prior method is effective to remove haze from the image, and after soft matting processing the dark channel image contains less complex interference information. Next, the dark channel image is processed in turn using image segmentation method based on Otsu algorithm, morphological processing method, and Sobel operator based edge detection method. Then the processed image basically retains only the edge information of the horizon. And In order to accurately obtain the horizon information, the straight line detection method based on Hough transform and least squares method are used as the final image processing step. Eventually, experimental results show that the proposed method can detect the horizon information more accurately.