Research on Characteristics of Template Matching Algorithm Under Image Distortion
Traditional template matching methods suffer from image local occlusion and deformation. To improve the matching performance of template matching algorithm when the image is distorted,a local template matching algorithm based on equidistant effective points is proposed. Firstly,candidate local eigenvectors are obtained by using local image grid processing method. Thereafter,a method to determine equidistant effective points is proposed based on local similarity of corresponding coordinate locations. Finally,the comprehensive similarity is obtained by normalizing the number and similarity of equidistant effective points. Local similarity is used to filter out distorted parts, which makes the eigenvalues of the parts that participate in the calculation of comprehensive similarity more accurate and further enhances the authenticity and reliability of the comprehensive similarity. The matching rate of the algorithm proposed is 100.0％ in both the original and the disturbed image set,and it is 94.1％ and 83.3％ in the occlusion and deformation image set respectively. The experimental results show that by virtue of the proposed algorithm,strong anti-interference is obtained in the condition of image distortion,which is robust to noise and illumination changes.
万方数据
维普
