Color Image Clustering using Hybrid Approach based on CIELab Color Space

In contrast with the problem of phantom colors which fails the slow spatial variation along the edges results in the poor clustering of pixels from the color image science.[1] In this paper color clustering results are improved using an hybrid approach where we are combining the benefits of bilateral filter with an improved ant based clustering algorithm. Bilateral filter is a non iterative, local and simple method for smoothening and edge preserving of an image. It combines the colors based on both their vicinity in the domain and range. in our approach ants are created dynamically on the spatial grid. Ants pick its respected data item using CMC as pheromone. As CMC similarity measure is best suitable for CIELab color space to quantifies the perceptual visual differences. The initial knowledge of the number of clusters is not required during the clustering process. As MSE is the global quality measure we applied here using Euclidian distance to evaluate the performance of proposed technique that decay with dissimilarity in clusters.