【摘要】 详细研究了图像轮廓的 CODON编码方法 ,将传统的基于封闭光滑轮廓的 CODON编码推广至任意非封闭非光滑曲线 ,并提出自然尺度和多尺度 CODON特征的概念 ,使图像的 CODON编码能进一步抑制噪声的影响 ,准确表征图像的轮廓信息更多还原

【Abstract】 The representation method of planar curves is studied, COD ON code which extends the set of CODONs so that both the open and the closed cur ves containing straight lines and cusps can be completely represented. The main drawback of CODONs is their sensitivity to noise due to the differential operato r. Thus,the multi-scale CODON is presented to solve the problem of noise created by the differential operator. In order to represent curves in e fficient scales, a derivation method is studied in which curves can be represen ted without redundancy and noise resistant. The number of zero-cross of curves is selected to determine the efficient scales. The CODON code is used in the ima ge matching technique so that the edge and contour features of images can give f ull play of their stability in imaging noise. Experimental results show the val idity of the improved CODONs.更多还原