C.3 - A simple cue-based camera calibration method for digital production of moving images

Abstract

One of the keys to new-generation digital image production applicable even to domestic uses is to construct simple methods for estimating the camera's motion, position, and orientation from a moving image sequence observed with a single domestic video camera. This chapter discusses a method for camera calibration along with an estimation of the focal length of the camera accurately by using four definite coplanar points as a cue. The practical computational algorithms for the cue-based method of camera calibration are composed of simple linear algebraic operations and arithmetic operations, and hence they work so well as to provide accurate estimates of the camera's motion, position, and orientation stably. Experimental simulations demonstrate that the cue-based camera calibration method works well for the digital moving image mixing task. The key to the accurate cue-based camera calibration is to accurately detect the feature points used as a cue in an input image. Sub-pixel accuracy is possibly required for the detection task. To do this, one should enhance the spatial resolution of the image region containing the feature points in advance.