Identification of image feature edge gradients during visual scanning to improve edge profile accuracy

Abstract

A new method to identify spatial gradients of edges using grey level variations is presented. This method is especially useful when camera resolution is low. A high accuracy edge profile can be obtained by blending the extracted edge gradients and edge positions. This paper experimentally proves that the grey level gradient is a function of the spatial gradient of a profile that transits across a pixel. A calibration equation is established relating the grey level gradients and edge gradients. This equation can be used to extract the gradients of an unknown profile. As positions along the profile are readily available, by blending in the profile gradients, an object profile with increased accuracy can be obtained. Experiments were conducted on a specially built rig to verify the validity of the theoretical formulations. An interpolation algorithm is presented to demonstrate how improved accuracy can be obtained using gradient blending.