Asaf Tzadok

Word-Based Adaptive OCR for Historical Books

The aim of this work is to propose a new approach to the recognition of historical texts by providing an adaptive mechanism that automatically tunes itself to a specific book. The system is based on clustering together all the similar words in a book/text and simultaneously handling entire class. The paper describes the architecture of such a system and new algorithms that have been developed for robust word image comparison (including registration, optical flow based distortion compensation, and adaptive binarization). Results for a large dataset are presented as well. Over 23% recognition improvement is demonstrated.

Transforming Japanese archives into accessible digital books

Digitized physical books offer access to tremendous amounts of knowledge, even for people with print-related disabilities. Various projects and standard activities are underway to make all of our past and present books accessible. However digitizing books requires extensive human efforts such as correcting the results of OCR (optical character recognition) and adding structural information such as headings. Some Asian languages need extra efforts for the OCR errors because of their many and varied character sets. Japanese has used more than 10,000 characters compared with a few hundred in English. This heavy workload is inhibiting the creation of accessible digital books. To facilitate digitization, we are developing a new system for processing physical books. We reduce and disperse the human efforts and accelerate conversions by combining automatic inference and human capabilities. Our system preserves the original page images for the entire digitization process to support gradual refinement and distributes the work as micro-tasks. We conducted trials with the Japanese National Diet Library (NDL) to evaluate the required effort for digitizing books with a variety of layouts and years of publication. The results showed old Japanese books had specific problems when correcting the OCR errors and adding structures. Drawing on our results, we discuss further workload reductions and future directions for international digitization systems.

Using image analytics to monitor retail store shelves

Using image analytics to monitor the contents and status of retail store shelves is an emerging trend with increasing business importance. Detecting and identifying multiple objects on store shelves involves a number of technical challenges. The particular nature of product package design, the arrangement of products on shelves, and the requirement to operate in unconstrained environments are just a few of the issues that must be addressed. We explain how we addressed these challenges in a system for monitoring planogram compliance, developed as part of a project with Tesco, a large multinational retailer. The new system offers store personnel an instant view of shelf status and a list of action items for restocking shelves. The core of the system is based on its ability to achieve high rates of product recognition, despite the very small visual differences between some products. This paper covers how state-of-the-art methods for object detection behave when applied to this problem. We also describe the innovative aspects of our implementation for size-scale-invariant product recognition and fine-grained classification.

Hybrid Approach to Adaptive OCR for Historical Books

Optical character recognition (OCR) technology is widely used to convert scanned documents to text. However, historical books still remain a challenge for state-of-the-art OCR engines. This work proposes a new approach to the OCR of large bodies of text by creating an adaptive mechanism that adjusts itself to each text being processed. This approach provides significant improvements to the OCR results achieved. Our approach uses a modified hierarchical optical flow with a second-order regularization term to compare each new character with the set of super-symbols (character templates) by using its distance maps. The classification process is based on a hybrid approach combining measures of geometrical differences (spatial domain) and distortion gradients (feature domain).

Fine-Grained Recognition of Thousands of Object Categories with Single-Example Training

We approach the problem of fast detection and recognition of a large number (thousands) of object categories while training on a very limited amount of examples, usually one per category. Examples of this task include: (i) detection of retail products, where we have only one studio image of each product available for training, (ii) detection of brand logos, and (iii) detection of 3D objects and their respective poses within a static 2D image, where only a sparse subset of (partial) object views is available for training, with a single example for each view. Building a detector based on so few examples presents a significant challenge for the current top-performing (deep) learning based techniques, which require large amounts of data to train. Our approach for this task is based on a non-parametric probabilistic model for initial detection, CNN-based refinement and temporal integration where applicable. We successfully demonstrate its usefulness in a variety of experiments on both existing and our own benchmarks achieving state-of-the-art performance.

Page Curling Correction for Scanned Books Using Local Distortion Information

The correction of page curling in scanned document images has attracted a lot of attention in recent years. Fixing page curling is essential because of the resulting damage in the visual perception of the scanned text and the ensuing reduction in OCR performance on the distorted image. It has been generally concluded that correcting the distortion due to page curling will serve as a solid basis for increased OCR accuracy. We present a novel approach for the efficient correction of page curling in the images of scanned book pages. The approach is based on the fact that approximately 70% of the words in any book are recurring terms. Thus, for many distorted words, a distinct and clear reference word can be found. Our work computes a global, polynomial transformation-based correction for the page distortion. This correction is based on the estimation of various local distortions in the given page, which are characterized by located words. Experiments on the scanned page images of an 18th century book printed in Old Gothic font have demonstrated the effectiveness of the proposed technique.

alpha-Shape Based Classification with Applications to Optical Character Recognition

We present a new classification engine based on the concept of

User Collaboration in Mass Digitisation of Textual Materials

\alpha

A New Physically Motivated Warping Model for Form Drop-Out

-shapes. Our technique is easy to implement and use, time-effective and generates good recognition results. We show how to efficiently use the concept of

Automatic location of address information on parcels sent by mass mailers

\alpha