| Summary: | 碩士 === 國立臺灣科技大學 === 電機工程系 === 101 === 3D monitors have attracted attention for the past few years. A viewer can enjoy a 3D movie by gearing with 3D glasses. Yet, with the advance of the 3D displayers, it is very likely that the naked-eye 3D becoming a trend. As it can be anticipated, the 3D market can be even bigger. Currently, many stereo information reconstruction technologies have been developed. One of them is the 2D to 3D image conversion, while it is very challenging to create a universal approach to adapt to various circumstances with the limited information given by a single image. Thus, this thesis proposes a semi-automatic 2D to 3D image transformation system.
In this thesis, the region growing scheme is employed to segment foreground and background. Subsequently, a user can manually determine the orientation of the depth. With the aid of various cues, the depth map can be estimated. The final two-view result can be further rendered with the DIBR, and output to a stereo monitor.
As documented in the experimental results, the proposed method can provide superior performance compared to the former related schemes in terms of the computation complexity and perceptual quality.
|
|---|
