Color quantization for image processing using self information

Abstract

A digital picture generally contains tens of thousands of colors. Therefore, most image processing applications first need to apply a color reduction scheme before performing further sophisticated analysis operations such as segmentation. While a lot of color reduction techniques exist in the literature, they are mainly designed for image compression and are unfortunately not suited for many image processing operations (e.g. segmentation) as they tend to alter image color structure and distribution. In this paper, we propose a new color reduction scheme (SICR), using probabilities and information theory elements to balance between the information provided by the selected colors and the necessity to accurately represent the selected colors. We also advocate for the use of perceptually accurate metrics for evaluation. Experimental results on a diversified dataset of images selected from the internet show that our technique performs well compared to other color reduction schemes.

Heat Map Based Feature Ranker: In depth comparison with popular methods

“…However, since the idea is to build a generalization of the original data, we apply a technique called Color Quantization that allows the mapping of a true color to a lower depth color scale, in this case, we reduce each true color to a 4-Bit 16-Colors scheme. The idea is to discard small data variations or uniqueness, and produce a new set of data which is more general and consistent, the process of quantization is performed by looping over the 16 reference colors and find the minimum Euclidean distance with the true color as this approach is very popular for the task and it has been found to produce satisfactory results [30].…”

Analysis of Colorimetric Stationary Criterion: Application to Colour Image Quantization

“…Indeed, when quantizing a colour image, the most representative colours are selected from that colour image for the purpose of displaying that image by those colours without generating visual degradations [1]. This has been the subject of several pieces of research [1], [2], [3], [4], [5], [10], [11]. The quantization techniques proposed in the literature fall into two categories:…”

“…With adaptive division, the colour space is divided up in function of certain criteria relating to visual perception, to colorimetric distribution in the colour space, and to the frequency of occurrence of colours in the scene. This process seems more realistic in terms of quantization, but it is more difficult to implement and some aspects, such as colour representativeness, are not related solely to the frequency of occurrence in the scene studied [4]. It seems too that a good quantization process must not only faithfully reproduce the original colour image with a limited number of colours but also be fast.…”