Research of sub-pixel location algorithm based on image correlation

Digital image correlation as a tool for surface deformation measurement has been widely used in the field of experimental mechanics. The method is known to resolve deformation gradient fields with sub-pixel accuracy. In this paper, we address the application of digital image correlation to the image location with sub-pixel accuracy to estimate displacement of multiple frames of video sequences. The estimation effect depends on various factors such as image noise and the correlation algorithm chosen. Algorithms of the sub-pixel location on image are analyzed: Gray-value Interpolation based Image Correlation and Correlation Coefficient Distribution based Fitting. However, Gray-value Interpolation needs a large amount of computational consumption although has high accuracy and it is apt to be influenced by noisy. Correlation Coefficient Distribution has low accuracy but high effective performance. According to the characteristics of these algorithms, a mix algorithm is introduced to improve both accuracy and computational consumption. The imaging process and algorithm execution are simulated using MATLAB. Further more, we could evaluate the displacements of moving objects between two frames of real video sequences and obtain the reconstructed images through displacement data. The validity of the mixed image location algorithm is obviously verified by comparison between original frames and reconstructed image.