The Independent Component Analysis (ICA) is showing a vital role in separating the image objects and dimensional reduction. The clouds are very important in the National Oceanic and Atmospheric Administration (NOAA) multispectral image. The removal of total cloud is very difficult, so these clouds are classified using this proposed technique. Using ICA and k-means clustering algorithm, the different types of clouds are classified and clouds are separated from the water and ground levels. These values are verified by temperature values of Ch4 and Ch5 bands, individual bands, Normalized Difference Vegetation Index (NDVI), and albedo reflectance values with Ch1 and Ch2. This algorithm shows optimum results compared with threshold and surface edge detection methods. The performance of this proposed method has been evaluated visually with good efficiency.