Content Based Image Retrieval (CBIR) for image segmentation is a concern this year, especially in the development of computer vision. The object discussed in this study is about interest, which uses a dataset from ImageCLEF2017 by taking 8 flower samples. Image of flowers in the dataset is still a lot of noise such as the initial background behind objects such as leaves, tree trunks or others. So we need a method to eliminate the noise, this method for cleaning noise is done by color clusters using the K-means method. By color clustering using K-Means and using color clusters k=2,3,4,and5. After that, a morphological process is carried out in order to obtain a clean area so that it can be compared with the original image and the Blob values formed. Blob analysis is calculated after the process of cleaning the noise is done in order to get the best value in the process of recognition of images with objects of interest. The results of the segmentation process that have been done are the highest MSE and RMSE values are at k-means results with k=4, while for PNSR are at k=2, and for the lowest MSE and RMSE values are at k=5, while the lowest PNSR is at k=4