In this work, a histogram-based colour image fuzzy clustering algorithm is proposed for addressing the problem of low efficiency due to computational complexity and poor clustering performance. Firstly, the presented scheme constructs the red, green and blue (short for RGB) component histograms of a given colour image, each of which is pre-processed to preserve their smoothness. Secondly, the proposed algorithm multi-thresholds each component histogram, using some dominant valleys identified from a fast peak-valley location scheme in each global histogram. Thirdly, a new histogram is reconstructed by applying a histogram merging scheme to the RGB three-component histograms, and multi-thresholding this new histogram again using some dominant valleys obtained from the fast peak-valley location scheme. Thus, the proposed approach can easily identify the initialisation condition of cluster centroids and centroid number. Finally, we construct a new dataset composed of some pre-segmented small regions using the WaterShed algorithm, and the FCM (Fuzzy C-Means) algorithm is executed on this dataset, instead of on pixels, in combination with the initial cluster centroids. Experimental results have demonstrated that the proposed algorithm is more efficient than the DSRPCL (Distance Sensitive Rival Penalised Competitive Learning) algorithm and the HTFCM (Histogram Thresholding Fuzzy C-Means) algorithm with respect to run times and PRI (Probability Rand Index) values.
Li CM, Huang R, Ding ZH et al (2011) A level set method for image segmentation in the presence of intensity inhomogeneities with application to MRI. IEEE Trans Image Process 20(7):2007–2016MathSciNetCrossRefGoogle Scholar