Segmentation-free word spotting in historical Bangla handwritten document using Wave Kernel Signature

Abstract

In this paper, we present a segmentation-free word spotting method based on Wave Kernel Signature (WKS) under the foundation of quantum mechanics. The query word and the document page are smoothened first, then SIFT detector is used to obtain the keypoints in both the query image and the document page. A window is placed centered at each keypoint to obtain the WKS descriptors. The WKS descriptors represent the average probability of measuring a quantum mechanical particle at a specific location based on quantum energy. We use an efficient search technique which calculates minimum energy difference between query word and document image to spot where the query word appears in the document image. The proposed method is tested on three historical Bangla handwritten datasets, one Bangla handwritten dataset, one old Bangla-printed dataset and one historical English handwritten dataset. To substantiate the goodness of the proposed method, its performance is measured using standard metrics.