We develop a common theoretical framework for combining classifiers which use distinct pattern representations and show that many existing schemes can be considered as special cases of compound classification where all the pattern representations are used jointly to make a decision. An experimental comparison of various classifier combination schemes demonstrates that the combination rule developed under the most restrictive assumptions—the sum rule—outperforms other classifier combinations schemes. A sensitivity analysis of the various schemes to estimation errors is carried out to show that this finding can be justified theoretically.

Abstract not found

Extract the Punjabi Word from image has been a very intensive area of research during last decades due to it is wide range of solution to real world problems. A lot of work has been done in languages like Chinese, Arabic, Devnagari, Urdu and English. A neural network based Gurmukhi recognition system has been developed. Range free skew detection and correction algorithms for de-skewing Gurmukhi machine printed text skewed at any angle have been developed. If different classifiers cooperate with each other group decisions may reduce errors drastically and achieve a higher performance. The whole process consists of two stages. The first, feature extraction stage analyzes the set of isolated characters and selects a set of features that can be used to uniquely identify characters. The performance depends heavily on what features are being used. Main advantage of this system is its accuracy to extract the Punjabi word. Input to the system is the scanned images from newspaper, magazines and old books and Extract the Punjabi Word from Machine printed Document Images.

In this paper, a feature extraction and hybrid classification scheme, using binary decision tree and nearest neighbour, for machine recognition of Gurmukhi characters is described. The classification process is carried out in three stages. In the first stage, the characters are grouped into three sets depending on their zonal position ( upper zone, middle zone and lower zone). In the second stage, the characters in middle zone set are further distributed into smaller sub-sets by a binary decision tree using a set of robust and font independent features. In the third stage, the nearest neighbour classifier is used and the special features distinguishing the characters in each subset are used. One significant point of this scheme, in contrast to the conventional single-stage classifiers where each character image is tested against all prototypes, is that a character image is tested against only certain subsets of classes at each stage. This enhances computational efficiency. 1.

In the name of Allah, the Most Gracious and the Most Merciful All praise and glory goes to Almighty Allah (Subhanahu Wa Ta’ala) who gave me the courage and patience to carry out this work. Peace and blessings of Allah be upon His last Prophet Muhammad (Sallulaho-Alaihe-Wassalam) and all his Sahaba (Razi-Allaho-Anhum) who devoted their lives towards the prosperity and spread of Islam. First and foremost gratitude is due to the esteemed university, the King Fahd University of Petroleum and Minerals for my admittance, and to its learned faculty members for imparting quality learning and knowledge with their valuable support and able guidance that has led my way through this point of undertaking my research work. My deep appreciation and heartfelt gratitude goes to my thesis advisor Dr. Mo-hamed Deriche for his constant endeavour, guidance and the numerous moments of attention he devoted throughout the course of this research work. His valuable ii suggestions made this work interesting and knowledgeable for me. Working with him in a friendly and motivating environment was really a joyful and learning ex-perience. I extend my deepest gratitude to my thesis committee members Dr. Asrar Sheikh and Dr. Mohandes.M for their constructive and positive criticism, extraordinary at-tention and thought-provoking contribution in my research. It was surely an honor and an exceptional learning to work with them. Acknowledgement is due to my senior fellows Saad Azher and Mohammad Moin Uddin for helping me on issues relating to LATEX and MATLAB. I also appreciate the help provided by my fellow Mudassir Masood in programming on MATLAB. Sincere friendship is the spice of life. I owe thanks to my house mates, colleagues and my friends for their help, motivation and pivotal support. A few of them are Moin

Extract the Punjabi Word from image has been a very intensive area of research during last decades due to it is wide range of solution to real world problems. A lot of work has been done in languages like Chinese, Arabic, Devnagari, Urdu and English. A neural network based Gurmukhi recognition system has been developed. Range free skew detection and correction algorithms for de-skewing Gurmukhi machine printed text skewed at any angle have been developed. If different classifiers cooperate with each other group decisions may reduce errors drastically and achieve a higher performance. The whole process consists of two stages. The first, feature extraction stage analyzes the set of isolated characters and selects a set of features that can be used to uniquely identify characters. The performance depends heavily on what features are being used. Main advantage of this system is its accuracy to extract the Punjabi word. Input to the system is the scanned images from newspaper, magazines and old books and Extract the Punjabi Word from Machine printed Document Images.