WEKO3
アイテム
Estimation of Atmospheric Turbidity from a Sky Image and Its Applications
https://doi.org/10.15083/00004056
https://doi.org/10.15083/0000405633fa789d-0077-4385-9862-fd4f51c7f1bf
| 名前 / ファイル | ライセンス | アクション |
|---|---|---|
48097417.pdf (3.4 MB)
|
|
| Item type | 学位論文 / Thesis or Dissertation(1) | |||||
|---|---|---|---|---|---|---|
| 公開日 | 2012-10-29 | |||||
| タイトル | ||||||
| タイトル | Estimation of Atmospheric Turbidity from a Sky Image and Its Applications | |||||
| 言語 | ||||||
| 言語 | eng | |||||
| 資源タイプ | ||||||
| 資源 | http://purl.org/coar/resource_type/c_46ec | |||||
| タイプ | thesis | |||||
| ID登録 | ||||||
| ID登録 | 10.15083/00004056 | |||||
| ID登録タイプ | JaLC | |||||
| その他のタイトル | ||||||
| その他のタイトル | 天空画像からの大気混濁係数推定とその応用 | |||||
| 著者 |
Zhao, Hongxun
× Zhao, Hongxun |
|||||
| 著者別名 | ||||||
| 識別子Scheme | WEKO | |||||
| 識別子 | 9374 | |||||
| 姓名 | 趙, 宏勛 | |||||
| 著者所属 | ||||||
| 著者所属 | 東京大学大学院情報理工学系研究科電子情報学専攻 | |||||
| 著者所属 | ||||||
| 著者所属 | Department of Information and Communication Engineering, Graduate School of Information Science and Technology, The University of Tokyo | |||||
| Abstract | ||||||
| 内容記述タイプ | Abstract | |||||
| 内容記述 | Atmospheric turbidity is defined as the ratio of optical thickness of the haze atmosphere as opposed to molecules, it is widely used in the computer graphics and vision community for modeling the outdoor illumination. This thesis proposes a novel method for estimating the atmospheric turbidity and provides three applications. Specifically, the atmospheric turbidity is estimated by matching the brightness distribution between a sky model and an actual image. By using the estimated turbidity, we recover the spectral sensitivity and white-balance parameters of digital cameras from sky images, render the virtual object with turbidity-based aerial perspective calculation, estimate the reflectance of outdoor diffuse object from a single perspective image. Camera spectral sensitivity plays an important role for many physics-based computer vision methods, such as demosaicing, color correction and illumination estimation. However, less attention has been paid to estimating the spectral sensitivity. This is unfortunate, because this parameter significantly affect the image colors. We have proposed a novel method for estimating the spectral sensitivity from images. A basic idea is to use a sky image from which the spectra of the sky can be inferred. Given the input image, the atmospheric turbidity is estimated by fitting the brightness distribution to a sky model. Assuming the sun direction with respect to the camera view direction can be estimated, the sky spectra is calculated from the estimated turbidity. Having obtained the pairs of image RGB values and corresponding spectra, the spectral sensitivity is calculated by using basis-functions. The basis functions are extracted from the database which consists of collected spectral sensitivities of different digital cameras. The utilization of the basis functions makes the estimation accurate and robust. Virtual object rendering has been widely used in the academy, the architecture, the entertainment and the heritage preservation. Aerial perspective plays an important rule for making the rendered image appear realistic, especially when the object is distant from the camera. The aerial perspective is caused by the scattering of particles in the atmosphere, it models the slight color change of distant object. The virtual object rendering with the aerial perspective effect is described. The spectral radiance of the sun and skylight is first calculated from the estimated turbidity, and the aerial perspective effect is calculated as the summation of in-scattered and out-scattered light along the camera view direction. Experimental results shows the accuracy of this technique. Three-dimension models have been widely used in many fields. The reflectance is necessary for making a realistic 3D model. However, how to estimate the accurate reflectance still remains a challenge, especially for outdoor objects, because it is difficult to precisely model the outdoor illumination. Many trials have been made to capture the outdoor illumination, but these methods need expensive equipments and massive calibrations. We propose a novel method for estimating the reflectance of outdoor diffuse object from a single perspective image. The atmospheric turbidity is first estimated from the input image, and then the whole illumination condition is recovered from the estimated turbidity. The reflectance is calculated from the surface radiance and irradiance value. The proposed method also solves the problem of inter-reflection, which exists between concave surfaces. We assume the object surface consists of hundreds of small facets, and the inter-reflection is calculated as the incoming light energy from all other facets. Experimental results show the accuracy achieved by the proposed method. The main contributions of this thesis are that a novel atmospheric turbidity estimation method and its three applications. It can be summarized by the four following points: First, the atmospheric turbidity is estimated by matching the brightness distributions between a sky model and an actual image. Second, the estimated atmospheric turbidity is used for recovering the camera spectral sensitivity and white balance parameters. Third, the estimated atmospheric turbidity is used to calculate the aerial perspective effect which makes the rendered image appear realistic. Fourth, the estimated atmospheric turbidity is used to model the outdoor illumination environment for estimating the reflectance of outdoor diffuse object. | |||||
| 書誌情報 | 発行日 2012-09-27 | |||||
| 日本十進分類法 | ||||||
| 主題Scheme | NDC | |||||
| 主題 | 451 | |||||
| 学位名 | ||||||
| 学位名 | 博士(情報理工学) | |||||
| 学位 | ||||||
| 値 | doctoral | |||||
| 学位分野 | ||||||
| Information Science and Technology (情報理工学) | ||||||
| 学位授与機関 | ||||||
| 学位授与機関名 | University of Tokyo (東京大学) | |||||
| 研究科・専攻 | ||||||
| Department of Information and Communication Engineering, Graduate School of Information Science and Technology (情報理工学系研究科電子情報学専攻) | ||||||
| 学位授与年月日 | ||||||
| 学位授与年月日 | 2012-09-27 | |||||
