Automatic optimal distance threshold prediction for microscope image matching

Feature extraction and matching are a fundamental step in the image stitching. The quality of image stitching depends on a quality of matching output. In the process of matching the distance threshold is set to select the output. If the distance threshold is set too high or too low, the stitching will be inaccurate. This paper presents algorithm for automatically estimate the optimal distance threshold. First, we analyze relationship between error of matching and accuracy of the result image. Then, we develop a model to evaluate the optimal distance threshold automatically by numerical method. Ours matching approach can be used to improve matching result of RANSAC.