研究生: |
黃建修 Chien-hsiu Huang |
---|---|
論文名稱: |
基於色彩分析之風景照片立體化技術 2D to 3D Conversion for Landscape Photos Based on Color Analysis |
指導教授: |
孫沛立
Pei-Li Sun 陳致曉 Chih-Hsiao Chen |
口試委員: |
羅梅君
none 陳鴻興 Hung-Shing Chen 溫照華 Chao-Hua Wen |
學位類別: |
碩士 Master |
系所名稱: |
電資學院 - 電子工程系 Department of Electronic and Computer Engineering |
論文出版年: | 2011 |
畢業學年度: | 99 |
語文別: | 中文 |
論文頁數: | 80 |
中文關鍵詞: | 影像深度圖 、立體視覺 、2D轉3D 、色彩心理 、影像檢索 、線性迴歸 、影像特徵 |
外文關鍵詞: | image depth map, stereoscopic vision, 2D to 3D conversion, color psychology, image retrieval, linear regression, image features |
相關次數: | 點閱:339 下載:5 |
分享至: |
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報 |
本論文首先提出一種「基於影像分割」的2D風景影像深度評估技術。該技術先以K-means色彩分群方式粗略地分割影像,並根據垂直方向的位置,給定每個區塊大致的深度值。再將YCbCr影像的三通道數值加權混合,作為击顯影像細節深度差異的增益值。然而,「基於影像分割」的方式容易因分割錯誤,導致影像深度錯亂。 為了改善上述方法的缺點,本論文另外提出「基於影像檢索」以及「基於影像特徵值迴歸」的兩種影像深度評估技術:前者是透過計算測詴影像與個別訓練影像之間的影像特徵差異度D,推算出不同位置的影像深度;後者則是由訓練影像資料事先建立了不同位置上,「影像特徵值」與「深度值」之間的迴歸方程式,將測詴影像的特徵值導入該位置所屬之迴歸方程式,推算出該位置的影像深度值。 心理視覺實驗結果顯示,基於「基於影像特徵值迴歸」結合YCbCr加權色彩資訊作影像細節的深度微調,其效果是最佳的。未來若能以更精準的方式,收集大量的影像深度圖作為資料庫,應該能使其有更好的表現。
The study first proposed a color-clustering-based method to generate depth map for 2D landscape photos. It segments blocks and assigns different image disparities to each blocks and modifies details of depth map by weighting YCbCr color information. However, the accuracy of resulted depth map is very sensitive to the accuracy of image segmentation. On the other hand, to overcome the disadvantage of the color-clustering-based method, we propose another two methods based on image retrieval and linear regression to generate depth map. The first method assigns image disparities by calculating the differences of image feature vectors between the test image and the training images on proximal fields, and the second method assigns image disparities by the linear regression between the local image feature vectors and the corresponding disparities of a large number of training images. The results of psycho-visual experiment show that the regression method combines lower color information is better mode. It should have better performance if the training depth maps are accurate and plentiful.
G. Bradski and A. Kaehler (2008). Learning OpenCV, OReilly, p.417.
L. Czuni, D. Csordas and G. Csaszar (2004). Distance Map Retrieval, ICIAR , pp. 811-817
P. Engledrum (2000). Psychometric Scaling, Imcotek Press, p.50.
D. Minoli (2010). 3DTV: Content Capture, Encoding and Transmission, Willey, p.50.
M. C. Kao and T. C. Shen (2009). A Novel Real-time 2D to 3D Conversion Technique Using Depth Based Rendering, IDW ’09, pp. 2033-2034.
H. Murata, Y. Mori, S. Yamashita, A. Maenaka, S. Okada, K. Oyamada and S. Kishimoto (1999). A Real Time 2D to 3D Image Conversion Technique Using Computed Image Depth, SID Symposium, Vol. 29, pp. 919-922.
Q. Q. Wei (2005). Converting 2D to 3D: A Survey, Technical Report, pp. 11-18.
W. J. Tam, C. Vazquez and F. Speranza (2009). Three-dimensional TV: A novel method for generating surrogate depth maps using colour information, Proc.SPIE, Vol. 7237, 1A.
C. T. Lin, S. C. Hsu, C. Y. Lin and J. F. Lee (2010). A Fully-Automated 2D to 3D Image Conversion Technique Applied on Stereo Display, CGVIP2010, pp. 1230-1231.
J. Y. Huang and P. L. Sun (2009). Converting 2D Photos to Stereoscopic 3D Images based on Image Segmentation, CVGIP 2009, pp.1315-1320.
C. C. Cheng, C. T. Li and L. G. Chen(2010). An Ultra-Low-Cost 2-D/3-D Video-Conversion System, SID 10 Digest, pp766-769.
J. B. MacQueen (1967), Some Methods for classification and Analysis of Multivariate Observations, Proceedings of 5-th Berkeley Symposium on Mathematical Statistics and Probability , University of California Press, 1:281-297.
John Canny(1986), A computational approach to edge detection, Pattern Analysis and Machine Intelligence, IEEE Transactions on, PAMI-8(6):679–698.
孫慶文、趙怡婷、簡苑如、廖元渝、翁慈憶(民 98),虛擬環境中色彩意象與空間深度知覺之關聯性研究,第十屆國際色彩學研討會,第6頁。
賴文能、陳韋志(民99),淺談2D至3D視訊轉換技術,影像與識別 2010, Vol.16 No.2。