Bag of Visual Words for Word Spotting in Handwritten Documents Based on Curvature Features

In this paper, we present a segmentation-based word spotting method for handwritten documents using Bag of Visual Words (BoVW) framework based on curvature features. The BoVW based word spotting methods extract SIFT or SURF features at each keypoint using fixed sized window. The drawbacks of these techniques are that they are memory intensive; the window size cannot be adapted to the length of the query and requires alignment between the keypoint sets. In order to overcome the drawbacks of SIFT or SURF local features based existing methods, we proposed to extract curvature feature at each keypoint of word image in BoVW framework. The curvature feature is scalar value describes the geometrical shape of the strokes and requires less memory space to store. The proposed method is evaluated using mean Average Precision metric through experimentation conducted on popular datasets such as GW, IAM and Bentham datasets. The yielded performances confirmed that our method outperforms existing word spotting techniques