Showing papers by "Neeta Nain published in 2015"

A Hybrid Feature Extraction Algorithm for Devanagari Script

Abstract: The efficiency of any character recognition technique is directly dependent on the accuracy of the generated feature set that could uniquely represent a character and hence correctly recognize it. This article proposes a hybrid approach combining the structural features of the character and a mathematical model of curve fitting to simulate the best features of a character. As a preprocessing step, skeletonization of the character is performed using an iterative thinning algorithm based on Raster scan of the character image. Then, a combination of structural features of the character like number of endpoints, loops, and intersection points is calculated. Further, the thinned character image is statistically zoned into partitions, and a quadratic curve-fitting model is applied on each partition forming a feature vector of the coefficients of the optimally fitted curve. This vector is combined with the spatial distribution of the foreground pixels for each zone and hence script-independent feature representation. The approach has been evaluated experimentally on Devanagari scripts. The algorithm achieves an average recognition accuracy of 93.4p.

Fast Face Detection Based on Skin Segmentation and Facial Features

Abstract: Human face detection plays an important role in various biometric applications such as crowd surveillance, photography, human computer interaction, tracking, automatic target recognition and many security related areas. Varying illumination conditions, color variance, brightness, pose variations are major challenging problems for face detection. This paper proposes a localized approach for face detection based on skin color segmentation and facial features. Skin color segmentation approach decreases the computational complexity and increases the accuracy since the skin region is previously determined. For skin color segmentation, we have used Y CbCr color image. The advantage of using Y CbCr is to remove the illumination component that is represented by Y. This method is tested on two databases: Bao database: contains 157 images and Muct database: contains 751 images. The algorithm achieves an average accuracy of 96:73%. Comparison with Viola Jones and Face Detection using Skin Color Model methods has also been done.

An Adaptive Codebook Model for Change Detection with Dynamic Background

Abstract: Codebook model is a widely used method forsegmenting foreground pixels. However, it is often generateerroneous positive result in case of dynamic background. This paper proposed an adaptive codebook model for change detection to disparate static background from dynamic background. To eliminate shadow/illumination effects cone-shaped color distance map is utilized in lieu of cylindrical. Moreover multi-layer codebook model is proposed containing codebooks for each pixel that is used to separate dynamic background from static background region. Proposed method reduces erroneous positive foreground pixels detected conventionally when pixel belongs to background shows dynamic behavior. During experimentation proposed method is tested over numerous videos with complexillumination and background situations. The experimental result shows improvement over basic codebook model and other state of-art background subtraction model.

A Framework for Compilation of Multi-lingual Handwritten Database: Four Levels XML Ground-Truth

Abstract: In this paper, we are presenting a semi-automatic framework for annotating multi-lingual handwritten texts document images. There is a significant need for a structure that can annotate the coordinate segmentation information of the text present in a handwritten document image to provide a platform for OCR algorithm evaluation. In this paper, we describe an XML based four level annotations of handwritten text image that contain the ground-truth information of script text image in Unicode format. In order to collect the huge amount of data for linguistic researchers, structure provide a way to store and annotate at different four levels: Image, Lines, Words and Characters which aids for benchmarking of various OCRs. Structure would be best source for compilation of an annotated handwritten corpora in systematic and scientific way by storing a labelling(markup) information of image script texts in a Unicode and an XML file format that encapsulates the bounding box pixel information of each level in a collaborative manner. The structure provides useful results based on the annotation for various quantitative and statistical corpus approaches to linguistic analysis.

Foreground Object Extraction Using Thresholding with Automatic Shadow Removal

Abstract: In Computer Vision, Visual Surveillance or in caseof Intelligent Transportation System, one of the most challengingtasks is the extraction of moving object or foreground extraction. Background Subtraction is the most intuitive method using singlefixed threshold for foreground extraction. Some techniques makeuse of correlated nature of RGB intensities, but can't resolvethe ambiguities of using a single fixed threshold. Moreover, segmentation has side effects like moving cast shadows and SelfShadow, both of which reduce accuracy. In this paper, we presentan automatic shadow removal method to extract objects. Theanalytical comparison with other foreground extraction, shadowdetection and removal algorithms and their results are alsopresented for better understanding. The proposed method extractsilhouette without any pre or post processing step of shadow removal.

Video partitioning by segmenting moving object trajectories

Abstract: Video partitioning may be involve in a number of applications and present solutions for monitoring and tracking particular person trajectory and also helps in to generate semantic analysis of single entity or of entire video. Many recent advances in object detection and tracking concern about motion structure and data association used to be assigned a label to trajectories and analyze them independently. In this work we propose an approach for video portioning and a structure is given to store motion structure of target set to monitor in video. Spatio-temporal tubes separate individual objects that help to generate semantic analysis report for each object individually. The semantic analysis system for video based on this framework provides not only efficient synopsis generation but also spatial collision where the temporal consistency can be resolved for representation of semantic knowledge of each object. For keeping low computational complexity trajectories are generated online and classification, knowledge representation and arrangement over spatial domain are suggested to perform in offline manner.

A unified approach for development of Urdu Corpus for OCR and demographic purpose

Abstract: This paper presents a methodology for the development of an Urdu handwritten text image Corpus and application of Corpus linguistics in the field of OCR and information retrieval from handwritten document. Compared to other language scripts, Urdu script is little bit complicated for data entry. To enter a single character it requires a combination of multiple keys entry. Here, a mixed approach is proposed and demonstrated for building Urdu Corpus for OCR and Demographic data collection. Demographic part of database could be used to train a system to fetch the data automatically, which will be helpful to simplify existing manual data-processing task involved in the field of data collection such as input forms like Passport, Ration Card, Voting Card, AADHAR, Driving licence, Indian Railway Reservation, Census data etc. This would increase the participation of Urdu language community in understanding and taking benefit of the Government schemes. To make availability and applicability of database in a vast area of corpus linguistics, we propose a methodology for data collection, mark-up, digital transcription, and XML metadata information for benchmarking.

An improved multi-layer codebook model to eliminate ghost region

Abstract: Codebook model is a widely used method for segmenting foreground pixels. However, it often generates the erroneous detection results with a dynamic background. This paper proposed a multi-layer codebook model for segmenting foreground pixels and a separate layer is utilize to disparate static background from dynamic background. It improve processing speed as system keep history of uncovered background region. Proposed method reduces erroneous positive pixels detected conventionally as ghost region when pixel belongs to background suddenly start moving. To eliminate shadow/illumination effects cone shaped color distance map is utilized in lieu of cylindrical. During experimentation proposed method is tested over numerous videos with complex illumination and background situations. The experimental result shows improvement over rudimentary codebook model and other state-of-art background subtraction model by reducing erroneous positive pixels detected such as ghost region and enhance foreground detection.

CALAM: Linguistic Structure to Annotate Handwritten Text Image Corpus

Abstract: In this paper, we report our effort in building a multi linguistic structure Cursive and Language Adaptive Methodology (CALAM) to create, annotate and validate linguistic dataset. CALAM provides a way for fetching and retrieval of information in a scientific and systematic manner through design and development of an annotated corpus of handwritten text image. It is a useful tool to annotate multi-lingual handwritten image dataset (Hindi, English, and Urdu etc.). The annotation is not limited with the grammatical tagging, but structural markup is also done. Annotation of handwritten text image is done in a hierarchical manner starting from handwritten form to segmented lines, words, and components. The component level markup is useful for finding strokes and list of ligatures in Urdu language. Along with a hierarchical access structure, CALAM provides the functionalities of Indexing, Insertion, Searching and Deletion of words and phrases in handwritten form. Apart from dataset fetching and retrieval it also automatically generates XML tagged file for each annotated handwritten text image for all dataset.