Research On Enhancement And Recognize Of Mongolian Household Pattern Based On SWT-LWT And Weighted Transformation

The indentification of Mongolian furniture pattern is mainly by human eyes from the subjective perspective,which causes inaccurate results of the classification.However,the application of computer can help get an intelligent recognition.And the sampling of Mongolian furniture pattern would be affected by environmental factors,sampling equipment,historical conditions,etc.As a result,the blurry pattern and the missing pattern information would affect the effect of the pattern recognition.On the basis of traditional image enhancement,the author proposes an enhancement algorithm of combining LWT-SWT(Lift Wavelet Transform-Stationary Wavelet Transform)with weighted transformation.That is to remove the pixels with fewer pixel frequency in the histogram of the original image,and then calculate by weighting function to improve the quality of Mongolian furniture pattern.The main contents of this thesis:1.It will introduce two traditional enhancement methods of histogram equalization bicubic interpolation and deal with Mongolian furniture patterns to compare the advantages and disadvantages of the two algorithms.2.Based on the existing traditional image enhancement algorithms,the author proposes an enhancement algorithm of combining LWT-SWT(Lift Wavelet Transform-Stationary Wavelet Transform)with Weighted transformation.Firstly,the furniture pattern is decomposed by the lifting wavelet transform(LWT)and stationary wavelet transform(SWT).The high frequency subband decomposed by LWT is interpolated into the high frequency subband decomposed by SWT.The generated high frequency subband and the untreated low frequency subband will be combined through the ILWT(Inverse Lift Wavelet Transform)to obtain the pattern with enhanced resolution.Secondly,the histogram of the pattern with enhanced resolution will be modified.By setting the threshold,the frequency histogram of pixels less than the threshold will be removed.Then the average histogram and weight function will be calculated.Finally,the enhanced furniture pattern will be obtained by converting pixels.3.The author will compare and subjectively analyze the Mongolian furniture pattern enhancement algorithm based on LWT-SWT and weighted transformation with histogram equalization bicubic interpolation algorithm,and the experimental results.At the same time,the thesis will employ PSNR(Peak Signal-to-Noise Ratio),MSE(Mean Square Transform)objective evaluation,andSSIM(Structure Similarity Index Measurement)to make an objective evaluation.4.The author will pre-process the Mongolian furniture pattern based on the enhancement algorithm of LWT-SWT and weighted transformation,histogram equalization,bicubic interpolation algorithm.Animal,plant and geometric furniture patterns will be used as identification samples.The(Entropy,E),(Contrast,C),(Homogeneity,Homo),(Deficit From,DF)will be employed for identification.Also,this thesis will use the(Support Vector Machine,SVM)to identify and analyze the results.