Robust and Fast Feature Points Matching

Aiming at the problem that most stereo matching algorithms cost too much time,an algorithm in which feature points are detected in frequency domain and feature vectors are extracted in spatial domain was proposed.Firstly,effective coding theory was studied.Secondly,the salient features were located in the image and their scales were computed.Finally,patterns whose scales are matched with the feature points' scales were constructed to extract features,and then,features were matched by the nearest neighbor rule.The experiment results show that the propesed method has high computational effciency,less time consumption,strong robustness to scale and affine tranformation and make a balance between speed and performance.