Calculation Error Analysis And Algorithm Optimization Of Image Moments Based On Iteration Method

Image moment is an algorithm which is used to extract image features.It has been widely used in the fi elds of target recognition,pattern re cognit ion and i mage processing s ince 1960 s.Compared wi th other image features,image moments have unparalleled adva ntages in the appli cations of the above fields.It c an be said that no ot her image features can be found a t this t ime compared to efficiency and stabilit y.However,image mome nt a lgorithm have the pr oblem of computational ac curacy as with other feature extraction a lgorithms.In the pr ocess of extra cti ng and processing of the image with the mome nt function,the error will be magnified with the order.The error s will be gradually e nlarge d in the operation process,whic h causes the image moments algorithm uns tabl e and eventually damages the accuracy of the image moments,.At present,only a s mall number of re searchers studies this issue,And t here are no researchers s ystematically,qua lit ati ve study of the problem.After re viewing domestic and foreign literature I found that there is no judgment of the convergence of the i mage of the criteria.In this c ase,the author inte nds to study the error gener ati on and transfer mechanism of the image moments al gorithm systematically,then tryi ng to fi nd some criteria for the convergence of t he moments of the evaluation image,and on t his basi s,make some Opti mizati on and improvement in order to suppress the error in the ca lculation process.The main contents of this paper are as follows:(1)we int roduce the types of errors in the matrix function of the iterative method and the origin of the errors,the influence of t hese e rror s on t he algorithm precision a nd the convergence of the algorithm,and the conventional image error eval uati on criteria.(2)The author conver ts the error transfer expression i nto a s econd order discre te error system to study the convergence of the algorithm by judgi ng t he st abil ity of the error syste m.In orde r to j udge the stability of the error s ystem,the author uses the Lyapunov method,the nor m metric method and the SVD decomposition method.According to these three methods,several crit eria for judgi ng t he convergence of the conve ntional image moment s algorithm are proposed.Finally,the fea sibility of the above criterion is verified by image recons truct ion expe r iment.(3)We Put forward two kinds of optimi zation suppression methods of transmissi on error,then int roducing the parameter opt imizat ion method of the second met hod emphatically.The me thod can modi fy the unstable paramete r image moment iterative algorithm wil l not converge into the convergence of the algori thm,which can effectively inhibit the transmission error.Finally,this method is used to reconstruct theima ge and analyze the e rror.The results show the feasibility of the met hod.