Logo matching in Arabic documents using region based features and SURF descriptor

The detection and matching of logos have widely attention for the document images retrieval projects. In this paper an automatic effective logo matching has been developed and applied to different Arabic documents. The proposed technique based on matching between logos by using two types of significant features. The first is region based feature which represent the global features of an object. This feature is used as first matching step that distinguish logos according to their exterior shapes and will greatly reduce the search space of dataset for the local feature. The second type is a local feature detector and descriptor called SURF (Speeded Up Robust Features). SURF is scale invariant representation of object used as fast and final matching between logos in segmented documents images. The proposed method has been tested on variety type of Arabic documents images with different shapes, sizes, colors, and resolutions. Experimental results show the performance of the proposed logo matching technique with 97% for recall, 99% for precision and 98% for F-score.