Image Synthesis from Real and Generated Images by Using A Unified Camera Model

使用统一相机模型从真实图像和生成图像进行图像合成

• 批准号: 08680410
• 负责人: ASADA Naoki
• 金额: $ 1.6万
• 依托单位: Hiroshima City University
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Scientific Research (C)
• 财政年份: 1996
• 资助国家: 日本
• 起止时间: 1996 至 1998
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/grant/KAKENHI-PROJECT-08680410/
• 关键词: Camera Model; Camera Calibration; Image Synthesis; Zoom Lens; Computer Vision; Computer Graphics; ズ-ムレンズ; 工学的歪み; 口径触現象; 可変円筒モデル

We have investigated a method to synthesize realistic images from real and generated images based on camera calibration.1. Calibrated Computer Graphics : We have proposed a new scheme for image synthesis called "Cali-brated Computer Graphics" and presents a method to produce defocused images based on a camera model and lens parameters obtained by calibrating a real camera. Using the reversed projection image blurring model, we have developed a multiple Z-buffer algorithm to generate images of arbitrary focus settings. Experimental verification demonstrated that the quality of reality of the generated images was as high as that of the real images.2. Improvement of Image Generation Algorithm : The computation of the multiple Z-buffer algorithm is rather heavy because the standard Z-buffer algorithm is applied on every pixel of an image to be generated. We have developed a multiple pinhole algorithm in which images are generated by accumulating a set of shifted pinhole images. and discuss … More ed the improvement of both algorithms by evaluating the trade-off between the reduction of computation time and the degradation of image quality.3. A Unified Camera Model of Zoom, Focus and Iris Parameters : We have proposed a unified camera model of zoom, focus and iris parameters for "Calibrated Computer Graphics". We first discuss the properties of the camera parameters and present a thin-lens-based camera model that describes the relationship between zoom, focus and iris parameters. Using the multiple Z-buffer algorithm with the calibration results, we generated images and compared them with real images of the same scene.4. Photometric Calibration of Zoom Lens Systems : We have developed a calibration method of zoom lens systems to eliminate the photometric distortion due to vignetting. A novel camera model including the variable cylinder is proposed to account. for the marginal intensity reduction observed in images. We presented the calibration method to estimate the intensity reduction from the lens parameters of zoom and iris. Less

我们研究了一种基于摄像机标定的从真实图像和生成图像合成真实图像的方法。校准计算机图形：我们提出了一种新的图像合成方案，称为“校准计算机图形”，并提出了一种基于摄像机模型和通过校准真实摄像机获得的镜头参数产生离焦图像的方法。利用反向投影图像模糊模型，我们开发了一种多z缓冲算法来生成任意焦点设置的图像。实验验证表明，生成的图像与真实图像的真实质量相当。图像生成算法的改进：由于生成图像的每个像素都使用标准Z-buffer算法，因此多重Z-buffer算法的计算量相当大。我们开发了一种多针孔算法，该算法通过积累一组移位的针孔图像来生成图像。通过评估计算时间的减少和图像质量的下降之间的权衡，进一步改进了这两种算法。变焦、对焦和光圈参数的统一相机模型：提出了“校准计算机图形学”中变焦、对焦和光圈参数的统一相机模型。我们首先讨论了相机参数的性质，并提出了一个基于薄透镜的相机模型，该模型描述了变焦、对焦和光圈参数之间的关系。利用多重z缓冲算法与标定结果生成图像，并与同一场景的真实图像进行对比。变焦镜头系统的光度校正：我们开发了一种变焦镜头系统的校正方法，以消除因渐晕引起的光度畸变。为此，提出了一种包含可变圆柱的摄像机模型。用于在图像中观察到的边缘强度降低。提出了从变焦和光圈的透镜参数估计光强衰减的标定方法。少

项目成果

Asada, Naoki: "Edge and Depth from Focus" International Journal of Computer Vision. Vol.26, No.2. 153-163 (1998)

浅田直树：“焦点的边缘和深度”国际计算机视觉杂志。

Naoki Asada: "Edge and Depth from Focus" International Journal of Computer Vision. 26・2. 153-163 (1998)

浅田直树：“焦点的边缘和深度”国际计算机视觉杂志 26・2（1998）。

Baba, Masashi: "Calibrated Computer Graphics : A new approach to image synthesis based on camera calibration" Transactions of Information Processing, Society of Japan. Vol.39, No.7. 2180-2188 (1998)

Baba, Masashi：“校准计算机图形：基于相机校准的图像合成新方法”，日本学会信息处理学报。

Amano, Akira: "Photometric Calibration of Zoom Lens Systems : Analysis and correction of vignetting distortion by the variable cylinder model" Transactions of Institute of Electronics, Information and Communication Engineers D-II. Vol.J80-D-II, No.6. 1458

Amano, Akira：“变焦镜头系统的光度校准：可变柱面模型对渐晕失真的分析和校正”电子、信息和通信工程师协会汇刊 D-II。

Asada, Naoki: "Seeing Behind the Scene : Analysis of photometric properties of occluding edges by the reversed projection blurring model" IEEE Transactions on Pattern Analysis and Machine Intelligence. Vol.20, No.2. 155-167 (1998)

Asada、Naoki：“了解幕后：通过反向投影模糊模型分析遮挡边缘的光度特性”IEEE 模式分析和机器智能汇刊。