| Summary: | 碩士 === 龍華科技大學 === 電子系碩士班 === 94 === In three-dimensional head motion, the two-dimensional movement can be expressed by the rotational degrees in horizontal and vertical directions computed from simple geometry projection method. Furthermore, if the depth information of head can be obtained, the physical movement of head can be simulated and applied in the applications of human-machine interfaces. In this thesis, we propose a measurement technique that can compute the rotational degrees and depth of head from the images captured by only one normal camera.
In this thesis, the proposed method combines the techniques of depth from defocus and geometrical projection. Firstly, the depth information of head is estimated by using depth from defocus technique. After knowing the depth information of head, the current captured image is adjusted to form a reference image. Finally, a simple geometrical projection technique is exploited to compute the rotational degrees of head. The main idea of depth from defocus method is to derive the depth information from the difference of blurring amount in images by using the direct relationships among the depth, camera parameters and the amount of blurring in images. In order to obtain accurate measurement of difference of blurring amount, a particular object with clear and thick edge is used and placed at the same head’s depth. When the head’s depth changed, the blurring amount of that object altered too. By calculating the difference of blurring amount, the physical depth of head can be estimated. In the computation of head’s rotational degrees, we use inherent features of user’s face and applying the general principles of optical image formation to derive the 2-D geometric projected result that can be used to estimate the physical rotational degrees.
|
|---|
