Study On Off-Line Character Handwriting Identification Based On Texture Feature Extraction

The off-line character handwriting identification technique is a technique that aims to decide the identity of writers according to the character handwriting. With the continuous expansion of applied fields, the off-line character handwriting identification technique becomes an active area of computer vision and pattern-recognition.Because this technique involves all of the typical problems in image processing and pattern-recognition, for example, image pre-processing, feature extraction, classifier design and so on, so it is a formidable task in the image processing and pattern-recognition. The main purpose of this paper is to study the composition and the main algorithm of key parts of off-line character handwriting identification system,and offering technical support for system's realization.The system includes three parts: pre-processing, feature extraction, and classifier design.1. Pre-processing is the first step of off-line character handwriting image identification, in this paper, a series of complete pre-processing algorithm is adopted for off-line character image, includes removing the background and noise of image, and the processing of image binarization and normalization.2. The stability and efficiency of feature have crucial influence on the identification system. From the aspect of texture feature, this paper adopts a feature extraction algorithm base on multi-channel Gabor filter, this algorithm is text independent, it denotes the mean and standard deviation of each channel, they represent important messages of each channel's texture feature. 3. In the aspect of classifier design, this paper adopts a character handwriting classification method based on SVM theory. Through the training of SVM, and the tested feature data as tested samples, then the character handwriting type can be judged by means of the trained SVM according to the method adopted in this paper.In my experiment, through inputting character handwriting by various people and feature extraction, finally, it acquires good results by SVM training, the experimental data shows the algorithm proposed here is simple and effective.