An Algorithm for Image Clustering and Compression

This paper presents a new approach to image compression based on fuzzy clustering. This new approach includes pre-filtering, and fuzzy logic image enhancing to reduce undesirable noise effects on segmentation result; separation of image into 4x4 blocks and two dimensional discrete cosine transform; obtaining of peak values of cosine membership functions by combining of performing the zig-zag method with discrete cosine transform coefficients; obtaining of membership values and cluster centroids; and finally, creation of segmented image and compression. After applying the new method on sample images at different number of clusters, better compression ratio, performing time and good validity measure was observed. Possibility to reach incorrect results and local minima is also prevented for clustering by this new method.