Single Remote Sensing Image Dehazing Using Robust Light-Dark Prior

Haze, generated by floaters (semitransparent clouds, fog, snow, etc.) in the atmosphere, can significantly degrade utilization of remote sensing images (RSIs). However, existing techniques for single image dehazing rarely consider that haze is superimposed and shadow, they often aggravate degree shadow dark region. In this paper, a RSI method based on robust light-dark prior (RLDP) proposed, which utilizes proposed hybrid model to outlier pixels. RLDP method, first removed channel (RDCP). Then, with light (RLCP). Further, cube root mean enhancement (CRME)-based stable state search criterion solving difficult problem patch size setting. The experiment results benchmark Landsat 8 RSIs demonstrate could effectively remove haze.