CAJ | 학술논문

针对SIFT算法是一种提取局部特征的算法，不能完全体现整幅图像信息，并容易产生失配的问题，提出了一种改进方法。该方法利用swr算法的匹配结果计算两幅图像之间的对极几何关系，利用该约束关系对SIFT算法的结果进行筛选以提高SIFT算法匹配点的正确率。实验结果表明，匹配正确率随着误差精度的提高而提高，反映出算法具有良好的噪声鲁棒性。
침대SIFT산법시일충제취국부특정적산법，불능완전체현정폭도상신식，병용역산생실배적문제，제출료일충개진방법。해방법이용swr산법적필배결과계산량폭도상지간적대겁궤하관계，이용해약속관계대SIFT산법적결과진행사선이제고SIFT산법필배점적정학솔。실험결과표명，필배정학솔수착오차정도적제고이제고，반영출산법구유량호적조성로봉성。
A new enhanced algorithm based on epipolar gemometry is presented to reduce mismat- ches in SIFT, which extracts local feature of image and can＇ t express image information fully. This meth- od describes the epipolar geometric relationship between two input images, generated by the SIFT algo- rithm using the matched points. And then the correct matches are estimated using the relationship. The experimental results show that the correct rate is improved due to the raising of the error precision and the above method has good noise robustness.