Visibility Detection Based on Object Region Selection

Abstract

In response to the shortcomings of the dark channel prior and optimization methods in window selection and object depth variation, this paper proposes a visibility detection algorithm based on object region selection. The algorithm introduces superpixel segmentation to divide the image into superpixel blocks and extracts the superpixel blocks containing the target object for visibility calculation. In the visibility calculation, the dark channel and transmission rate extraction can be obtained only from the target object region, without the need to process the entire image region. This improves the accuracy of the target object parameter calculation and speeds up the computation.