为从纺织品图案中剔除纹理的同时保留主结构信息,提出了一种惩罚算子,在加权最小二乘法处理纺织品图案的过程中将该算子用于对图案主结构和纹理的模糊处理.阐述了惩罚算子对主结构和纹理进行不同程度惩罚的原理.建立了纺织品图案主结构样本库和纹理样本库并对惩罚算子的有效性进行了验证.为解决最小二乘法求解过程中大型稀疏矩阵求解耗时长的问题,采用分层预处理模型来减少稀疏矩阵求解的时间消耗以加快主结构的提取.采集大量实际纺织品样本,建立了纺织品图案样本库,利用纺织品图案样本库验证了分层预处理可将主结构提取时间缩短50%左右.%In order to remove the texture information from the textile pattern while maintaining the main structure information, a penalty operator was proposed to smooth the main structure and texture of the textile pattern differently in the process of weighting the least square method. The main structure sample database and the texture sample library of the textile pattern were established to verify the validity of the penalty operator. In order to solve the problem of large time cost of large sparse matrices in the solving process of the least squares method, a hierarchical preprocessing model was adopted to reduce the time consumption of sparse matrix solution to speed up the extraction of the main structure. Using the textile pattern sample library, it is verified that the hierarchical preprocessing can reduce the time for main structure extraction by about 50%.
展开▼
