An Adaptive Method for Lossy Compression of Big Images

• Activity Number: 163
• Type: Topic Contributed
• Date/Time: Monday, August 4, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Computing
• Author(s): Geoffrey Thompson*+ and Ranjan Maitra
• Companies: Iowa State University and Iowa State University
• Keywords: clustering ; compression ; sequential sampling ; k-means algorithm ; model selection ; image compression

Abstract

Compression algorithms are an important part of storing large images such as those obtained at high resolutions using cameras. We propose here an adaptive lossy algorithm which uses an iterative clustering algorithm to reduce storage requirements for images while still preserving image quality. Choosing an optimal number of clusters for a large image is done with a multi-stage sequential algorithm. First, an initial sample is clustered. The observations which can be reasonably described by this set of centers are filtered out, and then the sampling and fitting procedure is repeated until all of the observations have been classified. This yields an optimal number of clusters in one pass. Experimental results indicate good general performance for this clustering algorithm and, specifically, good image quality and compression.