A fully automatic endoscopic image correction method using Harris corner detection

We present a fully automatic endoscopic image correction method using Harris corner detection. Harris corner detection on a chessboard pattern image is performed and the interest points are grouped automatically according to lines they belong to. A metric function taking the distortion coefficients as parameters which measures the straightness of the lines consisting of the interest points is introduced. The distortion coefficients are determined by minimizing the metric function using the Levenberg-Marquardt method. No relative pose of the camera with respect to the chessboard pattern is assumed during the image acquisition and no user interaction is needed when the proposed algorithm is running. Experimental results show that our approach yields a good linear camera model with sub-pixel accuracy and is a simple flexible method especially for clinical application.