A new distinguishing algorithm of connected character image based on Fourier transform
Xiaoyan Zhu, Yifan Shi, Song Wang · 1999
Segmentation is the most difficult problem in a handwritten character recognition system and often contributes major errors to its performance. To reach a balance of speed and accuracy, a filter distinguishing a connected image from an isolated image is required for multi-stage segmentation. The Fourier spectrum is promising in this problem. Since it is influenced by the stroke width, we propose a Fourier spectrum standardization method. Based on the standardized Fourier spectrum, a set of features and a fine-tuned criterion are presented to classify connected/isolated images. A theoretical analysis proves their rationality. Experimental results demonstrate that this criterion is better than other methods.