Lytro Depth for Standard Plenoptic Cameras - Part 02

The data consists of seven datasets acquired under different zoom and focus settings for a 1st generation Lytro camera. The zoom and focus step settings of each dataset were determined by placing a target object at a pre-determined depth of the encasing of the camera and autofocusing on this object. This allows to define a plane in focus by the main lens that is close the target object. Thus, the focus depth is assumed to be the depth of the target object. The Part 02 contains the datasets captured with a focus depth of 1.5 m.

These datasets encompass images for calibration and depth range assessment. Each dataset is provided with a set of calibration plenoptic images since the camera parameters are different for each dataset. The calibration images are different from the depth images to ensure the results do not suffer from any type of overfitting effect.

The calibration plenoptic images were captured using a 19×19 calibration grid of 3.18 mm cells placed at different poses and at different depths close to the target object depth bearing in mind that a minimum of 10 poses are required.

On the other hand, the depth plenoptic images were captured using two different grid sizes: 19 × 19 grid of 470 6.10 × 6.08 mm cells and 5 × 7 grid of 26.50 × 26.38 mm cells. The grids for the depth plenoptic images were placed parallel to the encasing of the camera and at a regular spacing of 0.05 m from the camera for depth values ranging from 0.05 to 2.00 m. The two grid sizes are used for the depth plenoptic images since the depth range evaluated is wide and it is necessary to have a reasonable number of detections to assess the depth accuracy. The smaller grid size was placed up to a maximum depth of 1.0, 1.5 and 2.0 m according to the focus depth considered 0.05, 0.50 and 1.50 m. The bigger grid size was placed considering all depth range evaluated.

该数据集由七个在不同放大和聚焦设置下采集的Lytro第一代相机数据集组成。每个数据集的放大和聚焦步长设置是通过在相机外壳预定深度放置目标物体并进行自动对焦于此物体而确定的。这允许通过主镜头定义一个靠近目标物体的聚焦平面。因此，聚焦深度被视为目标物体的深度。第二部分包含聚焦深度为1.5米的采集数据集。这些数据集涵盖了用于校准和深度范围评估的图像。由于每个数据集的相机参数不同，因此每个数据集都提供了一套校准全息图像，以确保结果不受任何类型过拟合效应的影响。校准全息图像是通过在目标物体深度附近的不同姿态和深度放置一个由3.18毫米单元格组成的19×19校准网格进行采集的，考虑到至少需要10个姿态。另一方面，深度全息图像则是使用两种不同的网格尺寸进行采集：470 6.10×6.08毫米单元格的19×19网格和26.50×26.38毫米单元格的5×7网格。深度全息图像的网格放置在相机外壳平行且与相机保持0.05米的标准间距的位置，以覆盖从0.05到2.00米的深度值。由于评估的深度范围较广，需要合理的检测数量以评估深度精度，因此采用了两种网格尺寸。较小的网格尺寸根据0.05米、0.50米和1.50米的聚焦深度放置，直至最大深度1.0米、1.5米和2.0米。较大的网格尺寸则考虑到所有评估的深度范围。