An Improved SIFT Operator Based on the Theory of Zero\|crossing on Feature Extraction

An improved SIFT operator based on the theory of zero\|crossing is proposed for automatic remote sensing image matching. Firstly, in the improved SIFT algorithm based on zero\|crossing theory, the array of scale space pyramid zero\|crossing point detection is constructed. Then the search window around the center pixel for the zero\|crossing point in the adjacent three layers from the array of scale space detection pyramids is established. If the result of zero\|crossing detection is greater than the detection threshold, the center pixel will be marked as a feature point. Feature extraction based on the theory of zero\|crossing in Gaussian scale space introduced the image geometry feature in scale space detection, resulting in higher repeatability and more stable characteristics. The experiments carried with close\|range digital stereo image which has large rotation angle and low altitude aerial stereo\|image showed significant increase in the aspects of feature repeatability rate, correct number of corresponding points and matching correct rate with the improved SIFT operator applied to automatic remote sensing matching image.