The Recognition Of Western Xia Character Based On Moment Invariant

The paper aims at research of moment technique in Tangut Character recognition.In general,moments describe numeric quantities at some distance from a reference point or axis.Moments are commonly used in statistics to characterize the distribution of random variables.The use of moments for image analysis is straightforward if we consider a binary or gray level imagine segment as two-dimensions density distribution function.In this way,moments may be used to OCR(Optical Character Recognition)and extract properties that have analogies in statistics.The research work of this paper can be classified in following respects:(1)Based on the 4 corner code theory created by Professor Li and combined with special structure of Tangut Character,the essay proposed the coding strategy according the different part,in that the original coding method is be enlarged.This method is useful to the recognition of components of 4 corners of Tangut character.(2)The paper proposed the method of image segmentation algorithms about Ancient Tangut character.Baes on the non-recursion connected domain and merging algorithm,we make the improvement to the existing method.(3)Components image recognition is completely by the extraction of invariant image feature using Hu invariant moments.According the Hu's invariant moments having no scale invariance,normalization is being used to adjust the moments.In addition,the deduction of the formulation using the continuous function,the essay renews it in the discrete condition.(4)Without the properties of the orthogonality,the Hu invariant moments has the low anti-interference.Thus,the essay takes advantage of the Zernike moments,which among the orthogonal invariant.,meantime,anti-interference ability and noise resistance performance is higher.However,Zemike moments doesn't exit the rotation invariant,the paper has put forward the regularization method to correct it.The paper used the correlation coefficient and Minimum Distance to analysis the two-different invariant,the result denotes that two invariants are recognizing the character.But based on the statistical analysis,the Hu's invariant moments has the disadvantage of poor anti-interference,in that reduce the recognition rate.In general,under the same condition,the recognition rate under the Zernike is better than the Hu's,in addition,take a better recognition effect.