基于统计滤波和稀疏度优化的印花配准算法

在数码印花技术上，传统全局印花图案配准方法在精度和效率上无法满足需求，局部印花图案配准方法存在配准误差较多导致匹配误差大的问题，还存在图像变形控制点过多导致算法效率低等问题.该文提出了一种新的局部印花图案配准方法.该方法基于统计滤波优化配准算法，减少FLANN匹配点中的误差点；并在分析图像变形中冗余控制点特性的基础上，提出一种基于稀疏度的冗余优化算法降低控制点数量.实验结果表明：该方法可有效滤除配准后的误差点，优化控制点集，总体提高配准精度和算法效率.

Printing registration algorithm based on statistical outlier removal and sparsity optimization

In digital printing technology， the traditional global printing pattern registration method cannot meet the requirements in terms of accuracy and efficiency. The local printing pattern registration method has the problem of large registration errors resulting in large matching errors， and low algorithm efficiency coused by too many image distortion control points. This paper proposes a new registration method for partial printing patterns is proposed. This method is based on statistical outlier removal registration optimization algorithm to reduce the error points in the FLANN matching points. Based on analyzing the characteristics of redundant control points in image deformation， a redundancy optimization algorithm based on sparsity is proposed to reduce the number of control points. Experimental results show that this method can effectively filter out the error points after registration， optimize the set of control points， and improve the overall registration accuracy and algorithm efficiency.