Research of remote sensing image matching with sub-pixel accuracy

To increase sub-pixel matching speed with a certain sub-pixel accuracy,the conventional algorithms to obtain sub-pixel accuracy in template matching was discussed,a new algorithm was proposed to combine the resampling and surface fitting methods.At first,it should use the sequential similarity detection algorithm to find the best match point,then only the model was need to be resampled n-times,next the normalization correlations among each sub-template and the image in the best match point was calculated,and the position of the sub-template which was corresponded to the maximum normalization correlation in the image was calculated.At last the method of curve fitting in the 3×3 area of that position was applied to obtain the absolute coordinates of the positioning with sub-pixel accuracy.The experiment shows that the accuracy obtained by the new method base on 10 times resampling is less than 0.04 pixel.