Adding fuzzy color information for image classification
Juan I. Forcen, Miguel Pagola, Humberto Bustince, Jose Manuel Soto-Hidalgo, Jesús Chamorro-Martínez · 2017
Color is a powerful feature for image analysis but it is usually not used in image classification schemes. We propose a method to combine fuzzy color information with the result obtained from a One-Versus-All classifier (OVA) trained with Bag-of-features. This method consists in weighting the outputs of the OVA classifier based on the distances between the new image to be classified and the classes. Experimental results show that our approach improves OVA classifier performance.