In this study, it is aimed to use sub-block technique for the purpose of speeding up the segmentation of images within clustering algorithms. Due to the fact that all of the image data is given to clustering algorithms, generally the clustering process takes a lot of time, and it causes delays in real-time segmentation applications. In this study, in order to minimize the delays, the dividing of images into the sub-blocks, and using the average values of sub-blocks for the clustering process are proposed. As a result, the size of clustering data is relatively decreased. In the experimental studies, besides the images of travertine plates, well-known images such as "Lena" and "Baboon" were also used. The proposed method was compared with K-means, Fuzzy C-means, K-medoids and Spectral clustering methods, and its speed increased 2-4 times. Furthermore, it was observed that the image quality did not change too much in case of small size of blocks.