Abstract: In order to investigate the relationship and difference in brain functional connectivity when subjects process different valence of emotional images, a new method is employed to extract the relatively weak connected regions more accurately. First, a K-means algorithm based on density is proposed to analyze the brain functional connectivity and extract the brain structure model which has the functional connectivity. Then, aggregation index is introduced to evaluate the positioning accuracy of activated brain regions. The above results are also compared with the results using independent component analysis(ICA) algorithm; Finally, the advantage of K-means algorithm based on density in the field of brain functional connectivity analysis is demonstrated in terms of the intensity of voxel activation and the position precision of activated brain regions. The experimental results show that relatively obvious activity areas mainly distributed in the frontal lobe, cingulum and hypothalamus in the process of emotional stimulation processing, which provides a more reliable method for subsequent clinical observation and diagnosis.