Text segmentation in color images based on tensor voting and component labeling method
Xiuhong Chen · Jisuanji gongcheng yu sheji · 2009
In natural scene,text elements are corrupted by many types of noise.There effects make the automatic segmentation very difficult.We identify image using hue and intensity respective and normalization in the same range.Analysis each layers using tensor voting,remove image's noise effectively and restore the lost information.Excellent results are achieved by many times compute because confirm the threshold depend on experience.We use improved component labeling method confirm threshold and improve efficiency.Excellent results are achieved in experiments.