IMAGE CLASSIFICATION ACCORDING TO THE DOMINANT COLOUR

摘要

The aim of this work is to develop a user-friendly software allowing him to classify images according to their dominant colour expressed through linguistic expressions. With this aim in view, images are processed and stored in a database. The processing consists in assigning a profile to each image. To do this, we consider the pixels of the images in the colorimetric space HLS and then a restricted number of colours classes are built. These classes depend on the hue (H). For each colour class a certain number of subclasses depending on the lightness (L) and the Saturation (S) are defined. Finally the profile is drawn using the pixels membership of the classes and subclasses. Thus starting from a linguistic expression of a colour, the user can extract images from the database.