Algorithms Research On Thinning, Feature Extracting And Similar Chinese Characters Recognition For Off-line Handwritten Chinese Character Recognition

Off-line Handwritten Chinese Character Recognition(HCCR) has become a white-hot research point in pattern recognition which is largely different from Printed Chinese Character Recognition(PCCR) and on-line HCCR. Compared with PCCR, the style of handwritten chinese character is diverse and optional which is hard to find rules in it. Furthermore recognition of similar character in off-line HCCR is much more difficult than that in PCCR because of the ubiquity of handwritten distortion. On the other hand, compared with on-line HCCR, off-line HCCR dosen't have any real-time information. Therefore traditional thinning, feature extraction and similar chinese characters recognition are not fit for off-line HCCR, new algorithms should be proposed according to the speciality of off-line HCCR.Firstly, thinning character images may bring not only the elimination of image information redundancy but also the predigestion of character recognize pick-up procession and the improvement of its veracity and speediness by a long way. How to make thinning is the precondition of character recognize. Upon that an efficient thinning algorithm based on borderline (ETAB) is proposed which treats with individual pixels point in each step and picks up unilateral edge points sequence as the framework of a character image. Also, simulation results prove it can hold the basic information of character images better and can wipe off the redundancy pixel points of character images which solve the problem of intersection separates and the problem of shortened stroke segments. To sum up, the algorithm provides an efficient method to compartmentalize connective strokes, which achieves good thinning results.And then, a segment feature extract algorithm based on ETAB (SFE_ETAB) is proposed. The basic idear of this algorithm is to find the inflexion, end point and broken point in the thinned image by utilizing the specialty of which complicated segements naturally disassemble basic segements after ETAB thinned a character primarily. After that, it joins the segments to get the beeline corresponding to each segement. Simulations show that the new algorithm can extract segement features of a character image efficiently, and can attain good effect on segment feature extract.On the basis of the research above, the improper recognition reason of similar chinese characters is analyzed and a new similar chinese characters recognition algorithm is proposed. The requests of differentiating similar chinese characters under each condition are attained by analyzing. And feature points are extracted efficiently. What's more, the algorithm induced the concept of end points pair and broken points...