A regression approach to visual assessment for the perception of 2D to 3D conversion in stereoscopic displays

• Main Authors: Pin-Hsiang Hsu, 徐彬翔
• Other Authors: Shu-Chiang Lin
• Format: Others
• Language: en_US
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/19400527376533501504

碩士 === 國立臺灣科技大學 === 工業管理系 === 102 === ABSTRACT The technology of stereoscopic image goes on development in recent year. The related application of stereoscopic image has breakthroughs. However, it is not enough to increase the purchasing motivation of 3D products by counting on stereoscopic display, 3D-play equipment, and scarce 3D movies because of high cost of making 3D content. Due to high cost of stereo production, there are lots of studies indicate by converting 2D to 3D for cost down. Some scholars bring up Naked Eye 3D algorism to improve defect from the process of 2D to 3D conversion. Converting 2D image to 3D is a good technique to reduce both the production cost and to satisfy users’ need of 3D content. There are 4 control factors in this experiment, which are Image (15 pictures) Repetition (3 levels), DOG (Difference of Gaussian) Filter (4 levels) and Sigma (6 levels). We changed the levels of those independent factors for 3D image conversion, and then we ask participants to evaluate the perception of Depth, Distortion, Edge, and Blur which are dependent factors for the converted 3D images. Moreover, the results of ANOVA can be used for building up the Regression analysis model. The analysis showed that, the factors to the negative first power is the best transformed linear model of Regression. And the Regression coefficients of independent factors are minus. Above all, the higher value of independent factors is the better result of perception of stereograph will be. In addition, it is obvious that different images will have huge impacts on perception of stereograph. It is worth to make a thorough inquiry for the common of images which are better for perception of stereograph.