Font Size: a A A

Research On Algorithm Improvement Of Text Image Tilt Detection Based On DSP

Posted on:2019-11-28Degree:MasterType:Thesis
Country:ChinaCandidate:Y Q ChengFull Text:PDF
GTID:2428330545988672Subject:Control Engineering
Abstract/Summary:PDF Full Text Request
In the digital information age,the image has become an important information carrier and information resource.The extraction of information from text picture has become an important way to utilize image resources.Besides,converting text image into text message has become an important technology of digital office.Mobile phones,cameras and other multimedia devices have become the main tool of image collection.Image skew is a common cases,and character recognition is very sensitive to image tilt effect.At present,there are still some improvements in the image tilt angle detection technology.The main product of image recognition is still in the form of software based on PC,and the research of mobile hardware equipment is relatively small.In view of the above problems,the main contents of this paper are as follows:In view of the above problems,on the basis of the predecessors,this paper studies a method of pure text region screening in detail.Through the method of multi region width interception,the text area is located from the image,the pure text area is intercepted,the oblique direction of the image,the line inclination angle,the width of the line and interline blank are estimated,besides,the starting position of the upper and lower edges of the text line are marked.Then a step edge search method is proposed,which combines the information obtained in the front to search the edges of the lines.Finally,the edge points are transformed by Hough transform,and the angle is voted cumulative to get the angle of the final word image.In this paper,we can detect the oblique angle of text images including illustrations and tables by using the method of filtering the multi area width text area and the straight line detection method.It can detect the tilt angle of the literal image in the range of the tilt angle.And it has higher detection range,detection speed and detection accuracy compared with several traditional methods,and these are proved from theory and simulation experiment based on MATLAB.In the software design part based on DSP,the dual core architecture of ARM+DSP is selected as the core of the system.The related contents of the CCS experimental platform,the DSP/BIOS real-time operating system and the related content of the design and optimization of the code are introduced.
Keywords/Search Tags:Word recognition, tilt angle detection, Hough transform, text area extraction, DSP
PDF Full Text Request
Related items