Optimization of pattern recognition of hijaiyah letters using normalized cross correlation tchniques

Abstract

The development of analysis in digital image increasingly developed with various methods, one of which is in recognition of letter patterns. Each letter written using handwriting must have different writing patterns, such as the thickness and shape of the letter pattern. This research will be doing on the pattern recognition of hijaiyah letters of handwriting by applying the Normalized Cross Correlation (NCC) technique. NCC is a technique used to match two images. Before the NCC process, it should be done with the preprocessing using convolution and without convolution using the binary image. The convolution technique used was the Sobel and Prewitt edge detection with the aimed to get the edge of an object and compared the number of matching letters between using edge detection and without edge detection. The tests were done by using the different sized image of 32x32 pixels, 64x64 pixels and then match it against a similar sample data, a different sample data, a different objects font sample data and a different sample data of original image size. The results show that the matching of the letter pattern depends on the size of the image that is more matches to the image of 32x32 pixels. The binary image had better matching numbers than the convolution techniques. While in convolution techniques, Prewitt edge detection had the higher accuracy and matching results compared to the image using Sobel edge detection.