Improving Logo Spotting and Matching for Document Categorization by a Post-Filter Based on Homography

Abstract

Digital document categorization based on logo spotting and recognition has raised a great interest in the research community because logos in documents are sources of information for categorizing documents with low costs. In this paper, we present an approach to improve the result of our method for logo spotting and recognition based on keypoint matching and presented in our previous paper [7]. First, the keypoints from both the query document images and a given set of logos (logo gallery) are extracted and described by SIFT, and are matched in the SIFT feature space. Secondly, logo segmentation is performed using spatial density-based clustering. The contribution of this paper is to add a third step where homography is used to filter the matched keypoints as a postprocessing. And finally, in the decision stage, logo classification is performed by using an accumulating histogram. Our approach is tested using a well-known benchmark database of real world documents containing logos, and achieves good performances compared to state-of-the-art approaches.

Document Retrieval Based on Logo Spotting Using Key-Point Matching

“…In recent years, there has been a number of approaches proposed for logo detection [11,16,17,18], logo recognition [19,20] and logo spotting [6,13,7,8]. In the field of logo retrieval, M. Rusinol et al [14] introduce a method for organizing and indexing logos based on a description using a variant of the shape context descriptor.…”

“…Learning-based methods in [11,13,18], methods based on shape context descriptor [14,12], methods based on context-dependent keypoint matching [10], and key-point matching method [3,6] were popular. Among these, a key-point matching method using the nearest neighbor matching rule with ambiguity rejection based on the two nearest neighbors, has achieved good results in [3,6,7,8]. In these methods, a threshold I in (0,1) is used for matching key-points; if the ratio of distances to its two nearest neighbors is greater than I, then it means that the matching is not reliable, as there is a possible ambiguity between the two nearest neighbors.…”

Key-point matching with post-filter using SIFT and BRIEF in logo spotting

“…6). According to the two-step method proposed in [8] based on homography using RANSAC, a transformation between the pairs of matched key-points in each candidate region is computed to integrate spatial relationships between the key-points in the query image and those in the candidate region. Then, the transformation is computed again with all the pairs of matched key-points within a bounding box which is estimated based on the size of logo image (Fig.…”

“…A number of approaches for logo detection [12,17,18,19], logo recognition [20,21] and logo spotting [6,14,7,8] are proposed. Many different techniques in pattern recognition and spotting have been employed and developed throughout research work.…”

“…Learning-based methods in [12,14,19], methods based on shape context descriptor [15,13], methods based on context-dependent key-point matching [11], and keypoint matching method [3,6] were popular. Among these, a key-point matching method using the nearest neighbor matching rule with ambiguity rejection based on the two nearest neighbors, has achieved good results in [3,6,7,8].…”