Implementation of Image Quality and Design Time for Block-based Lossy VQ Image Compression using K-Means and K-Medoids Algorithm in Spatial Domain

Dr. Ali Tariq Bhatti, Dr. Jung H. Kim, Dr. Robert Li

Data Mining (DM) technologies are one of the vast and hot topics in today’s era. In the digital and computing world, information is generated and collected at a rate that rapidly exceeds the boundary range. It is increasing significantly in bio-medical, bio-informatics, engineering and health-care research. Large amount of biological and clinical data have been generated and collected at an unprecedented speed and scale. It seems to be a big disadvantage during storage and transmission. It raises the problem of reducing the memory size of a digital image. This research has focused on the lossy Vector Quantization (VQ) image compression to reduce the data volume in the spatial domain. We are using two unsupervised clustering algorithm for lossy VQ technique such as K-Means and K-Medoids. These two algorithms are proposed to analyze the performance in data mining and to other applications. To evaluate the clustering quality in image compression, the distance between two data points are taken for analysis. For two unsupervised clustering techniques, different block size are used as 4x4 (16) and 8x8 (64) with the codebook size of 25 and 50 in this research paper.  In spatial domain, K-Medoids was compared with K-Means in terms of execution time and quality (PSNR). The research results show that the K-Means algorithm yields the best results compared with K-Medoids algorithm.