Closed water body extraction from low-solution and Low SNR meteorological satellite imagery based on MSER

Water body extraction from remote sensing imagery is significant for map registration and surveying, especially for low-resolution and low Signal Noise Ratio (SNR) meteorological satellite imagery. In single-channel imageries, short-wave infrared (SWIR) imagery presents a high contrast between water bodies and other ground objects. Otherwise, SWIR imagery is greatly disturbed by stripes and missing scan lines. We mainly use the method of moment matching in the process of restoration and MSER in the process of water body extraction. Experiments on meteorological satellite imageries show that the proposed methods achieve promising results. A high stability for extracting water bodies is obtained by the Hu moment invariants.