A study of distortion measures for vector quantization of images

Vector Quantization is a term used to describe the determination and implementation of quantization steps for image vectors, essentially similar to its more familiar special case, scalar quantization, in the image and speech compression area of digital signal processing. The term "compression" means lossy data reduction by removing redundancy and unessential parts of the data in order to reduce the hardware and bandwidth requirements for storing and/or transmitting the data. Vector quantization is gaining more attention lately because of its property of high compression ratio.; The objective of this research is to determine the advantages and disadvantages of distortion measures in VQ, primarily focusing on the mean squared distortion. Furthermore, the selection of distortion measures is iterated for a particular imaging application. Two distortion measures are studied for two different applications. Example applications studied are the compression of ordinary life images with edge enhancement and the compression of Particle Image Velocimetry (PIV) images using a distortion measure in a fluid flow vectors domain.; PIV is a fluid mechanics application used to determine various fluid flow quantities using laser speckle and particle displacement measurement techniques. In our experiments we used vector quantization and genetic algorithms together to compress the PIV images and compared the results with ones that are obtained using LBG algorithms and JPEG which is another popular image compression algorithm that employs transform coding.