| Summary: | 碩士 === 國立宜蘭大學 === 資訊工程學系碩士班 === 102 === In recent years, 3D vision equipment and content are easier to obtain and getting cheaper price. 3D tracking technology is more progress because of technological breakthroughs. Traditionally, 3D object-tracking is the use of image processing methods. However, these methods are more time consuming. Based on time considerations, a new depth sensor uses the principle of infrared reflection to make fast object-tracking. Unfortunately, it cannot be detected correctly when the sensor behind the glass or try to detect a reflective objects. In order to overcome those problems, this paper presents a fast phase correlation based disparity estimation technique. A hierarchical scheme is proposed to improve the speed and accuracy for the image depth estimation. Noting the problem that local phase correlation based disparity estimation may fail in some image areas either featureless or with significant similar texture between an image pair. A multi-size image pair based disparity estimation called hierarchical disparity estimation, is design to improve the unreliable process around these areas. Moreover an pixel-square integral image is employed to speed up the computing time of variance estimate, which is used for deciding the size of image pairs, from O(n2) to O(1).From the result of our experiments, the proposed algorithm achieved a faster and more accurate disparity estimation and produce a efficiently image depth map for the interactive system.
|
|---|
