Automatic Control of a Robot Camera for Broadcasting Based on Cameramen's Techniques and Subjective Evaluation and Analysis of Reproduced Images.

Abstract

With the goal of achieving an intelligent robot camera system that can take dynamic images automatically through humanlike, natural camera work, we analyzed how images were shot, subjectively evaluated reproduced images, and examined effects of camerawork, using camera control technique as a parameter. It was found that (1) A high evaluation is obtained when human-based data are used for the position adjusting velocity curve of the target; (2) Evaluation scores are relatively high for images taken with feedback-feedforward camera control method for target movement in one direction; (3) Keeping the target within the image area using the control method that imitates human camera handling becomes increasingly difficult when the target changes both direction and velocity and becomes bigger and faster, and (4) The mechanical feedback method can cope with rapid changes in the target's direction and velocity, constantly keeping the target within the image area, though the viewer finds the image rather mechanical as opposed to humanlike.