Colour Image Retrieval by Self-Organizing Maps

摘要

The actual necesity for systems that searches for images by visual features is making the growing of a new reasearch field exclusivity centered in this problem. This paper introduces a new method to organice the image database of a Content-Based Information Retrieval System. Our first implementation of an image retrieval systems uses a L~*a~*b~* colour histogram as image feature. The feature vectors are organiced by a neural net, a Self-Organizing Map. This net classifies the image database in subgroups of images where the system can apply any type of ranking metric to show the results to the final user.