Subpixel Target Location Method Based On Phase Correlation

Photo-topography is a new technique in the metrology based on optical measurement. Including electronics, computer technology, laser technique and image processing, it has been widely used in geometrical, telemetry and profile measurements. With the increasing need for the higher precision measurements, the traditional measurement based on the pixel level is not enough. To increase the precision of the image measurement system, the general methods include improving the performance of the image sensors and imaging systems. However, the optical systems are known to be limited in resolution because of diffraction caused by their limited numerical aperture. And the size of the CCD sensor pixels cannot be infinitely reduced to obtain an improvement in the resolution, for the basic reason that they must collect a minimum energy to provide a convenient SNR. Therefore subpixel technique as a software method to improve the system performance becomes a hotspot of the current research in phototopography.The target location is a basic problem in phototopography. For the location of a known target, the problem focuses on image matching. For subpixel image matching, the popular method is using the cross correlation between the object image and reference image to locate the object on the pixel level, and then the subpixel algorithm is taken to refine the location. In this thesis, we proposed a phase correlation method which is a popular method in image subpixel registration. A comparison of the performances of traditional cross correlation and the phase only correlation of images is given. Numerical results show that the phase correlation method can be applied to subpixel target location of digital images. Furthermore, it has demonstrated that the phase correlation method has its own features. The phase correlation method has a higher robustness to various types of noise and it can also handle blurred images.We derived the analytic expressions for the phase correlation of down sampled images and directly extended it target subpixel location problem. Numerical experiments have been carried out using different types of images. The numerical results show that this method can get a good approximation in the...