Dark Channel Prior-Based Altitude Extraction Method for a Single Mountain Remote Sensing Image

The altitude information of single remote sensing image may aid in detecting natural disasters such as landslides or mud-rock flow. Accordingly, in this letter, an approach based on dark channel prior is proposed for the altitude extraction of single remote sensing image, and it also overcomes the effect of mountain shadow. We first detect the mountain shadows based on machine learning, and then adjust the brightness of each shadow with an adaptive adjustment parameter. Next, we estimate the altitude information based on dark channel prior, including atmospheric light calculation and soft matting. The experimental results with real mountain remote sensing images demonstrate that the proposed algorithm can obtain the accurate relative altitude information, which is effective for the extraction of the relative altitude of single mountain remote sensing image with shadows.