Preprocessing Method for NaXi Pictographs Character Recognition Using Wavelet Transform

Abstract

Preprocessing, a major component of Character Recognition System, has direct effect on the recognition system by its performance. Using wavelet transform, this paper mainly focuses on noise filtering and Edge Detection in a Preprocessing method for NaXi Pictographs Character Recognition. Image denoising is an important issue in image preprocessing. Based on the analysis of thresholding function and thresholding, which are very important in image denoising, we put forward the procedures of wavelet denoising for NaXi Pictographs. The experimental results turned out that wavelet denoising is a better method for NaXi Pictographs Recognition Preprocessing than linear denoising, median denoising or adaptive denoising. The images are broken down into the approximate parts and detail parts by wavelet. The approximate part is an approximate representation for the original image after filtering the high frequency part. Removed the high frequency component, the approximate part can detect the edge which the original image cannot detect. The wavelet based edge detection also has better effect than the traditional operators like Prewitt, Sobel, Roberts, Laplacian of Gaussian, Canny and others.