Fajri Kurniawan

Self Organizing Features Map with Improved Segmentation to Identify Touching of Adjacent Characters in Handwritten Words

This paper presents an intelligent technique for segmentation of off-line cursive handwritten words particularly on touching characters problem. In this study, Self Organizing Feature Maps (SOM) is implemented to identify the touching portion of the cursive words. The image of the connected characters is preprocessed and the core-zone is detected to overcome ascender and descender of the touched character. Prior to clustering, the pixels of the image are mapped into coordinate system as features vector. These features vector are clustered into three classes: left, right and middle region, and the vertical segmentation is performed using SOM to determine the winner node of middle region.The experiments are conducted using syntactic CCC database.The results show that the proposed algorithm yields promising segmentation output and feasible with other existing techniques.

An automatic approach for line detection and removal without smash-up characters

Abstract In composite document image, handwritten and printed text is often found to be overlapped with printed lines. The problem becomes critical for obscure and broken lines at multiple positions. Consequently, line removal is unavoidable pre-processing stage in the development of robust object recognisers. Moreover, the restoration of the smash-up characters after removal of lines still persists to be a problem of interest. This paper presents a new approach to detect and remove unwanted printed line inherited in the text image at any position without character distortion to avoid restoration stage. The proposed technique is based on connected component analysis. Experiments are conducted using single line images that scanned and extracted manually from several documents and forms. It is demonstrated that our approach is equally suitable to deal with line removal in printed and handwritten text written in any language circumvent restoration stage. Promising results are reported in comparison with the other researchers in the state of the arts.

Two-Layer Fragile Watermarking Method Secured with Chaotic Map for Authentication of Digital Holy Quran

This paper presents a novel watermarking method to facilitate the authentication and detection of the image forgery on the Quran images. Two layers of embedding scheme on wavelet and spatial domain are introduced to enhance the sensitivity of fragile watermarking and defend the attacks. Discrete wavelet transforms are applied to decompose the host image into wavelet prior to embedding the watermark in the wavelet domain. The watermarked wavelet coefficient is inverted back to spatial domain then the least significant bits is utilized to hide another watermark. A chaotic map is utilized to blur the watermark to make it secure against the local attack. The proposed method allows high watermark payloads, while preserving good image quality. Experiment results confirm that the proposed methods are fragile and have superior tampering detection even though the tampered area is very small.

From contours to characters segmentation of cursive handwritten words with neural assistance

This paper presents a novel algorithm to resolve an open problem to correctly locating letter boundaries in off-line unconstrained cursive handwritten word images. The proposed algorithm is based on vertical contour analysis. Following preprocessing, during the course of pre-segmentation vertical contours are analyzed from right to left. Furthermore to improve accuracy of segmentation, trained ANN is employed to validate segment points. For fair analysis, experiments were performed on IAM benchmark database. Results obtained thus show that the proposed approach is capable to accurately locating the letter boundaries for unconstraint cursive handwritten words

A review on 2D ear recognition

Biometric systems have been implemented on numerous public facilities that able to enhance the security system. Nowadays, fingerprint and face are the most popular biometric. In addition, emerging technology has introduced potential biometric such as hand geometry, palm print, lips, teeth and vein. However, most of this biometric requires a special device to capture it. This will added more cost to implement the system. Ear is one of emerging biometric and proven having a great potential for identifying a person. Recently, a study shows human ear can easily captured just using an ordinary camera and it can be extracted from existing face profile image. In general, human ear is captured using a camera, after that the ear recognition system will attempt to identify it. Numerous methods have been proposed by the community to establish ear recognition system. Hence, this paper summarized, reviewed and critically discussed various recent advances in 2D ear recognition, in order to find out the research gap.

Performance Comparison between Contour-Based and Enhanced Heuristic-Based for Character Segmentation

This paper presents a performance comparison between the proposed segmentation method with the existing methods [17]. Generally, the heuristic segmentation of proposed approach namely contour-based segmentation (CBS) is working on contour analysis and using two global features, which are estimation of stroke width and average character width. Meanwhile, enhanced heuristic segmentation (EHS) is based on boundary and connected component analysis along with several estimation of global feature includes average character width, average stroke width, and word height. Besides that, CBS assigns prospective segmentation according to ligature types. In contrast, EHS has dissimilar scheme that assign the prospective segmentation at minima on the lower contour using modified vertical histogram analysis. The last difference is; CBS is using reduction scheme to reduce over-segmentation, whereas EHS removes over-segmentation that sliced at a character holes. Regarding to those differences, a performance comparison is conducted. The objective is to highlight the success of CBS as compared to EHS.

An automated approach to remove line from text bypassing restoration stage

In composite document image, handwritten and printed text is often found to be overlapped with printed lines. The problem becomes critical for obscure and broken lines at multiple positions. Consequently, line removal is unavoidable pre-processing stage in the development of robust object recognizers. Moreover, the restoration of the handwritten area after removal of lines still persists to be a problem of interest. This paper presents a new approach to detect and remove unwanted printed line inherited in the text image at any position without characters distortion to avoid restoration stage. The proposed technique is based on connected component analysis. Experiment is conducted using single line images that scanned and extracted manually from several documents and forms. It is demonstrated that our approach is equally suitable to deal with line removal in printed and handwritten text written in any language circumvent restoration stage.

Performance Analysis of Segmentation Approach for Cursive Handwriting on Benchmark Database

The purpose of this paper is to analyze improved performance of our segmentation algorithm on IAM benchmark database in comparison to others available in the literature from accuracy and complexity points of view. Segmentation is achieved by analyzing ligatures which are strong points for segmentation of cursive handwritten words. Following preprocessing, a new heuristic technique is employed to over-segment each word at potential segmentation points. Subsequently, a simple criterion is performed to come out with fine segmentation points based on character shape analysis. Finally, the fine segmentation points are fed to train neural network for validating segment points to enhance accuracy. Based on detailed analysis and comparison, it was observed that proposed approach increased the segmentation accuracy with minimum computational complexity.

Geometrical and eigenvector features for ear recognition

Unconstrained ear biometric means an ear image that has variance in view and pose. This situation is challenging in ear recognition because one ear has various presentation. In this study, two features are considered to handle unconstrained ear image. The features called geometrical feature and eigenvector features. In eigenvector feature, the ear is extracted from six regions then the eigenvector is computed from each of those regions. Each region has capability to represent particular part of the ear image. Another feature is called geometrical feature that reflecting the shape of ear image. The widely used classifier is utilized and it trained with both features. Proposed method outcome is measured to evaluate the recognition rates among single features and fused features. The experiment is carried out on benchmark database collected by University of Science and Technology Beijing (USTB). It shows the proposed method can achieved promising result.