Cloud detection is preliminary task which shows the pathway to meteorological and cloud field research. This paper puts forward the effective cloud detection scheme for cumulus and cirrus cloud. First, Two Level Thresholding Algorithm is proposed for cirrus cloud detection. Second, In order to detect cumulus cloud, we conduct research and enhance fast marching method (FMM) by providing the provision towards appropriate selection of seed locations in digital image. Also, graph is constructed from image, edge weights are computed using gradient magnitude (Smoothness constraint) as well as grey level intensity difference between pixel and target cloud location. Edge weights are utilized to divide the image into segments. Experimental results illustrate that the proposed algorithm shows significant results for cloud detection and give better F-score than existing state-of-art cloud detection algorithms.