Abstract: Single pixel edges or figure are needed for border recognition in the field of 3D garment human body modeling.Integration of multi-scale plays an important role in step-form edge detection used by wavelet transform.In this paper，a method was applied to determine the double thresholds adaptively based on the histogram of the local maxima.Then，by restraint of local no-maxima to further edge thinning，and linking edge points to attack edge discontinuity using the double thresholds，the suitable human figure image can be detected.The experimental results showed this algorithm is feasible and effective.

摘要： 在个性化的服装人体三维图像测量技术中，人体边界识别需要得到准确的单像素级细化轮廓。在综合采用多尺度小波变换进行图像灰度的阶跃型突变点检测基础上，对小波模极大值的直方图均化处理以自适应地选取双阈值，通过抑制非极大值点对边缘线细化，并利用双阈值跟踪连接边缘点，从而得到单像素级的人体轮廓图像。对人台图像的实验证明了算法的可行性和有效性。