Cloud and shadow detecting method

The invention proposes a cloud and shadow detecting method. first, an image is sampled and a five-layer image pyramid is acquired. Then a saliency map is calculated on acquired each layer of the imagepyramid. Sampling is performed on the saliency map each layer and the saliency value is calculated. Finally, five saliency areas with the largest saliency values are selected for constructing cloud and shadow selection areas. Therefore, data redundancy is reduced effectively and other irrelevant object influence is weakened. The five saliency areas with the largest saliency values are assumed tocontain clouds and shadows and are sued for constructing possibility selection areas of cloud and shadow. A slide window is utilized for acquiring sample blocks that may be cloud or shadow in the constructed possibility selection areas and the sample blocks are input into a pre-trained classifier for classification. According to the invention, sampling is guided by saliency detection, so that positions of cloud and shadow