Search

One of the most recent challenging issues of pattern recognition and artificial intelligence is Arabic text recognition. This research topic is still a pervasive and unaddressed research field, because of several factors. Complications arise due to the cursive nature of the Arabic writing, character similarities, unlimited vocabulary, use of multi-size and mixed-fonts, etc. To handle these challenges, an automatic Arabic text recognition requires building a robust system by computing discriminative features and applying a rigorous classifier together to achieve an improved performance. In this work, we introduce a new deep learning based system that recognizes Arabic text contained in images. We propose a novel hybrid network, combining a Bag-of-Feature (BoF) framework for feature extraction based on a deep Sparse Auto-Encoder (SAE), and Hidden Markov Models (HMMs), for sequence recognition. Our proposed system, termed BoF-deep SAE-HMM, is tested on four datasets, namely the printed Arabic line images Printed KHATT (P-KHATT), the benchmark printed word images Arabic Printed Text Image (APTI), the benchmark handwritten Arabic word images IFN/ENIT, and the benchmark handwritten digits images Modified National Institute of Standards and Technology (MNIST).

Unconstrained handwriting text recognition is a stimulating field in the branch of pattern recognition. This field is still an open search due to the wide variability of human writing. Recent trends show a potential improvement of recognition by adoption a novel representation of extracted features. In the present paper, we propose a novel feature extraction model by learning a Bag of Features Framework for handwritten text recognition based on Deep Sparse Auto-Encoder. The Hidden Markov Models are then used for sequences modeling. For features learned quality evaluation, our proposed system was tested on two handwritten text datasets IFN/ENIT word images benchmark and MNIST handwritten digits. Our method achieves promising recognition on both datasets.

The relevant entity extraction from scanned document image is a very challenging task due to highly heterogeneous templates, and several structure layouts. These problems lead to inaccuracy for document image recognized by OCR. In this paper, we propose an effective solution for these problems, in which the relevant entities are extracted from Arabic and Latin scanned invoices. The input of the system is an invoice image which is submitted to an OCR without layout analysis. After, invoices are labeled in the text recognized by the OCR. By combining the logical and physical structures, a local graph model is built for extraction entity. Finally, we implement a correction module which requires the mislabeling correction by eliminating the superfluous parts detected by labeling step. We evaluate the obtained results with 1050 real invoices as reported in experimental section.

Identifying scripts in natural images is an important step in document analysis. Recently, Convolutional Neural Network (CNN) has achieved great success in image classification tasks, due to its strong capacity and invariance to translation and distortions. A problem with training a new CNN is that it requires a large amount of labelled images and extensive computation resources. Transfer learning from pre-trained models proves to ease the application of CNN and even boost the performance in some circumstances. In this paper, we use transfer learning and fine-tuning in document analysis. Indeed, we deal with the scene script identification quantitatively by comparing the performances of transfer learning and learning from scratch. We evaluate two CNN architectures trained on natural images: AlexNet and VGG-16. Experimental results on several benchmark datasets namely, SIW-13, MLe2e and CVSI2015, demonstrate that our approach outperforms previous approaches and full training.

In recent years, growing attention has been paid to recognizing text in natural scenes images. Scene Character recognition (SCR) is an important step in automatizing the process of reading text in natural scenes.

Nowadays, the number of mobile applications based on image registration and recognition is increasing. Most interesting applications include mobile translator which can read text characters in the real world and translates it into the native language instantaneously. In this context, we aim to recognize characters in natural scenes by computing significant points so called key points or features/interest points in the image. So, it will be important to compare and evaluate features descriptors in terms of matching accuracy and processing time in a particular context of natural scene images. In this paper, we were interested on comparing the efficiency of the binary features as alternatives to the traditional SIFT and SURF in matching Arabic characters descended from natural scenes. We demonstrate that the binary descriptor ORB yields not only to similar results in terms of matching characters performance that the famous SIFT but also to faster computation suitable for mobile applications.