Research On Text Location And Segmentation Technique In Image

With the coming of the era of big data,massive information retrieval becomes an urgent need.Thus text extraction in images has more and more important application value.Text information in images is an important part of understanding the whole image.A great number of text extraction methods have been widely applied in specific situations such as video security monitoring,real-time license plate recognition,auto-driving and content-based image indexing.Although a lot of research has been done,text localization remains to be a challenging problem.This is because the images often have complex background with low contrast,unconstrained fonts,sizes,color and alignment of characters.These factors make the text extraction become a challenging problem.Text extraction can be roughly devided into three steps: text localization,text segmentation and text recognition.Text recognition is very mayture now with a lot of commercial software.So the paper focus on the two steps,we studied the text localization and segmentation methods for images with complex background.The main contents are as follows:1.The Stroke Width Transform(SWT)text location method is sense to complex background,aiming at this problem,we present a new method based on the improved SWT.Firstly,The Maximally Stable Extremal Region(MSER)method is used to detect text candidate regions.Secondly,the SWT algorithm is used in the candidate regions,which can improve the edge detection rate compared with tradition SWT method.Finally,the Frequency-tuned(FT)visual saliency is introduced to remove false-text candidate regions.The experiment results show that,the method can achieve good robustness for complex background with multi-orientation perspective,various characters and font sizes.2.In order to overcome the shortcomings of ghost and disconnection problems of traditional Niblack method,an improved binarization method for image is proposed in this paper.The method has considered the global information of the images and the local information of the images.First,Tradition Otsu and Niblack thresholds are used for initial binarization.Second,we introduced the difference between maximum and minimum values in the local window as a third threshold to generate two images.Finally,with a logic AND operation of the two images,great results were obtained.The experiment results prove that the proposed method is reliable and effective.