Bag of Visual Words Based on Co-HOG Features for Word Spotting in Handwritten Documents

Introduction

Document Image Analysis (DIA) is one of the active research areas, which attracts research community due to its complexity and the increasing requirement for accessing the content of digitized contents. Optical Character Recognition (OCR) has been investigated for a few decades with huge achievement, which helps to automate the human process. The automatic recognition by traditional OCR system is suitable for modern high quality printed documents with simple layouts and known fonts. The poor quality printed text and handwritten text is not feasible by traditional OCR system. Processing of these documents through the OCR system requires high computation cost because of complexity involved in understanding the page layout of documents, irregular writing styles, faded ink, stained paper and other undesirable factors. In order to overcome these problems, the DIA community has developed a technique called word spotting. Word spotting technique is a moderately new alternative for character recognition and retrieval in both printed and handwritten documents.

Word spotting can b e defined as the pattern recognition task aimed at locating and retrieving a particular word from a document image collection without explicitly transcribing the whole corpus. The word spotting approaches do not require the recognition of every letter of the query word or the target word and thus are capable of similar word retrieval in the presence of small distortions. Generally, a typical word spotting system consists of three main modules: preprocessing, features extraction and feature matching. Among them, features extraction is one of the most important factors for achieving high retrieval performance, because features with strong discriminative information can be well classified even using simplest classifier. The literature survey reveals that Histogram of Oriented Gradients (HOG) descriptor is widely used in several recognition applications because of its discriminating ability compared to other existing feature descriptors. The HOG feature descriptor is developed by Dalal et al. (2005) for human detection using Support Vector Machine (SVM) classifier. The HOG has been successfully applied in many research fields such as word spotting task (Rodrıguez 2008; Terasawa et al., 2009), body parts detection (Corvee et al., 2010), face recognition (Deniz et al., 2011; Shu et al., 2011), character recognition (Newell et al., 2011), text/non-text classification problem (Minetto et al., 2013) and vehicle detection in traffic video (Arrospide et al., 2013).

Rodriguez, J. A. et al. (2008) have proposed local gradient histogram features for word spotting in unconstrained handwritten documents. A sliding window moves from left to right over a word image. At each position, the window is subdivided into cells, and in each cell, a histogram of orientations is accumulated. Slit style HOG features for handwritten document image word spotting is proposed by Terasawa et al. (2009). Newell et al. (2011) have extended the HOG descriptor to include features at multiple scales for character recognition. Saidani et al. (2015) have proposed a novel approach for Arabic and Latin script identification based on Histogram of Oriented Gradients feature descriptors. HOG is first applied at word level based on writing orientation analysis. Then, they are extended to word image partitions to capture fine and discriminating details. The unsupervised segmentation-free HOG based word spotting method was proposed by Almazan, et al. (2014). Documents are represented by a grid of HOG descriptors, and a sliding-window approach is used to locate the document regions that are most similar to the query.