Trademark logo recognition: preliminary results on a comparative between Haar-like features and local binary patterns

Abstract

In this paper, a comparative study between two computer vision methods for recognizing trademark logos in images, is presented. Trademark logo recognition has different real-world applications, for example the identification of the source of a store. Logos in images present different conditions that affect their automatic recognition, such as shape, scale, location, illumination and perspective, among others. We present a training and detection phase to recognize logos using Haar-like features and Local Binary Patterns, two well-known methods in the computer vision community that obtain high accuracy for object recognition. Our preliminary results, using an image database of logos, show that the method of Haar-like features obtains a 20% better performance than Local Binary Patterns. Also, the reported results indicate a fast processing time, making it suitable for real-time applications such as augmented reality marketing applications