Improved SIFT algorithm based on adaptive contrast threshold

Abstract

How to adjust the feature points number adaptively according to the images in different scenes is one of the key issues in improving detection efficiency. In this paper, an improved SIFT algorithm based on adaptive contrast threshold was proposed. Firstly, back propagation neural network and analytic hierarchy process were used to analyze the mathematical models of feature points number, image information and SIFT contrast threshold in different scenes from the perspective of image complexity, so as to realize the dynamic adjustability of contrast threshold. Then, a new SIFT algorithm framework was constructed by using the adaptive control module based on the mathematical model, and ultimately the number of feature points was coordinated. Compared with the two existing algorithms, the experimental data verified that the proposed algorithm had higher efficiency and accuracy, and that it realized the efficient control of feature point number in multi-scene.