Abstract: The satellite’s storage device and downlink bandwidth are facing great challenges with the improvements of cameras’ resolution and swath width。However, 33%-50% of the remote sensing images are clouds, and these images are almost useless for losing the information on ground objects. This paper proposed a new multi-branches cloud discrimination algorithm to control the camera stop photo in cloud area. Firstly, the spectrum threshold method is used to distinguish between clouds and ground objects roughly. Then the texture analysis method is adopted after threshold method invalid. To reduce the false alarm rate, a new method based on wavelet SCM is used to extract texture properties, and a bi-judgement method based on ASM and entropy is proposed. The algorithm has been verified by 245 remote sensing images. The experimental results show that this algorithm can detect clouds and ground objects correctly, and the false alarm rate is lower than 5%.