Layer by Layer Depth Assignment System For 2D to 3D Video Transform

• Main Author: 蕭論豪
• Other Authors: 章定遠
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/37333595472481718411

碩士 === 國立嘉義大學 === 資訊工程學系研究所 === 101 === This thesis is aim to apply the depth during the 2D to 3D image. We effectively use various depth information such as clarity of the image, lightness and intensity of texture in our images to layer the depth. Although each feature is calculated in different ways, they are inspired by a common concept, that is, feature information can help us to find areas of potential interest in the image. According to the application of each stage, this paper has four topics, including searching image features, SVM classification, Cutting block layer-by-layer and depth assignment. We assign different depth features by different types of images. To define the depth of the features, we use SVM voting system to separate the depth feature from near to far into four classes, the strongest depth feature is selected to be the first layer, and so on. After that, we use image segmentation to identify the main objects of each layer, finally, we give the proper depth to objects corresponding to the layer.