A satellite precipitation estimation algorithm based on wavelet features is investigated to find the optimal wavelet features in terms of wavelet family and sliding window size. In this work, the infrared satellite based images along with ground gauge (radar corrected) observations are used for the retrieval rainfall. The goal of this work is to find an optimal wavelet transform to represent better features for cloud classification and rainfall estimation. Our approach involves the following four steps: 1) segmentation of infrared cloud images into patches; 2) feature extraction using a wavelet-based method; 3) clustering and classification of cloud patches using neural network, and 4) dynamic application of brightness temperature (Tb) and rain rate relationships, derived using satellite observations. The results show that Haar and Symlet wavelets with sliding window size 5×5 have better estimate performance than other wavelet families and window sizes.