A0950
Title: Model-based color quantization with Gaussian copula mixtures
Authors: Ranjan Maitra - Iowa State University (United States) [presenting]
Samipan Majumder - Iowa State University (United States)
Abstract: Color quantization is used in computer graphics to reduce the number of colors in an image without appreciably losing its visual quality. The importance of this process comes from a need to display images on devices that are not completely capable of displaying all the colors in a digital image. Each pixel, or picture element, in an image is represented in terms of its primary components, namely Red, Green, and Blue. Therefore, each pixel has a certain amount of Red, a certain amount of Green, and a certain amount of Blue. This way of representing color is known as the RGB format. Model-based color quantization is performed in a way that more accurately models the RGB channels of a digital image. The model uses a Gaussian copula with Kumaraswamy marginals, while also practically incorporating spatial context through penalization. The methodology is demonstrated on several digital images.
