Abstract: Self-similarity is that property of being invariant with a different scale. The most of the researches used fractal dimension to calculate the self-similarity. In this study, we present a new algorithm, based on matching rang and domain fractal to find self-similarity properties of the data sets which can be used in data mining such as clustering and classification. This research focuses on two main points. Firstly, ranges and domains matching technique is used to extract self-similarity features from the images. Secondly, comparative between hash function and MapReduce to reducing time complexity. The experimental results show that the images from the same class are grouped to gather. The proposed methods of reducing time complexity results are presented and compared with traditional methods. The hash function reduced the complexity 0 (m×n) to 0 (mlogn) while MapReduce technique reduce the complexity 0(m×n) to 0(m×n/t) for the time where t is a number a of map task.