LETTER FUSION BASED ON ARMLET AND BALANCED MULTIWAVELET

Letter fusion is one of the most important stages in letter recognition. Tang Yuan Yan first used wavelet transform (WT) in this area. However, multiwavelet possesses desirable properties in compassion with scale-valued wavelet. In this paper, a new algorithm for letter fusion is presented based on multiwavelet transform (MWT). We apply various wavelets, in particularly, the analysis-ready multiwavlets (Armlet) which was constructed by Lian jian-ao recently to our experiment. Experimental results show that balanced multiwavlet transform (BMWT) is more effective than MWT on merging letters, BMWT of higher balanced order indicates larger entropy and Armlet has better performance that approximates to the balanced wavelet.