A Multiscale Sub-pixel Detector for Corners in Camera Calibration Targets

Hessian matrix is an important tool to analysis local characters of images, and detectors based on it are used widely in camera calibration. This paper derives a new representation of Hessian matrix by wavelet transform modulus and furtherer proposed a multiscale sub-pixel corner detector based on Hessian matrix and wavelet. It is expected to detect the corners at different scales and overcome the drawback of the single-scale detectors, which may usually either miss significant corners or detect false corners due to noise. For other kinds of corners besides X-corners mentioned in this paper, relative multiscale detectors based on Hessian matrix can be easily designed in the same way as this article. Computer simulation experiments show that at low noise level our algorithm is slightly more accurate than traditional single scale algorithm, and that at high noise level our algorithm is still robust enough to detect most corners when the traditional detector fails.