Compression of Color Images Using Clustering Techniques

This paper deals mainly with the image compression algorithms and presents a new color space normalization (CSN) technique for enhancing the discriminating power of color space along with the principal component analysis (PCA) which enables compression of colour images. Context-based modeling is an important step in image compression. We used optimized clustering algorithms to compress images effectively by making use of a large number of image contexts by separating a finite unlabeled data set into a finite and discrete set of natural, hidden data structures, rather than provide an accurate characterization of unobserved samples generated from the same probability distribution. Since images contain large number of varying density regions, we used an optimized density based algorithm from a pool. PCA is used to express the large 1-D vector of pixels constructed from 2-D color image into the compact principal components of the feature space. Each image may be represented as a weighted sum (feature vector) of the eigen values, which are stored in a 1D array. PCA allows us to compute a linear transformation that maps data from a high dimensional space to a lower dimensional space. It covers standard deviation, covariance, eigen vectors and eigen values. These values are provided at the compression side. So the decompressed image has not much loss of information. Experiments using different databases show that the proposed method by combining color space discrimination and PCA can improve compression of color images to a great extend.