Translational sequences of panoramic high dynamic range images in natural environments

This database contains 37 sequences of panoramic high dynamic range images recorded in natural environments with a wide range of depth structures. Each sequence consists of up to 100 images that were subsequently recorded on a straight path of one meter, with 1 or 2 cm distance between the images depending on the sequence. Therefore, the image sequences reflect a translational movement. The images are panoramic in full 360° in azimuth and between -58° below and 47° above the horizon in elevation. Any rotational yaw movement can, thus, be calculated via software from individual images.
We used a spectral filter to limit the camera’s spectral sensitivity to a wavelength range of 480-560 nm (green). This filtering mimics the spectral sensitivity of photoreceptors R1-R6 that provide the input of the fly motion vision system. As a consequence, the mapping of colors to gray values in these images is similar to the green color channel in RGB images.
The raw images have a resolution of approximately 1 megapixel (928x928) and 12-bit. The images have a high dynamic range covering the entire brightness range encountered in natural environments (excluding the solar disc) After linearization the resulting image values had a dynamic range of 1:23,900 covering 3,955 intensity steps. Note, however, that the pixel brightness values cannot be recalculated to a SI unit like candela, though the values are proportional to luminance in the green spectral range. For more technical details about the recording of the image sequences see Schwegmann et al. (2014b).
In addition to the raw camera images, unwrapped and linearized panorama images are provided with a resolution of 927 x 250.
Additionally, the distance map of the environment for each frame of the image sequences is provided within this database given as distance in cm for every pixel. See Schwegmann et al. (2014a) for detailed information on how the distance maps were obtained.
Additional Matlab .m scripts that can be used for processing the data in this archive is provided in: Schwegmann, A., Lindemann, J. P., Egelhaaf, M. (2014) Matlab .m scripts for processing panoramic HDR images. doi:10.4119/unibi/2693180
Please refer to the readme.pdf contained in the data archives for detailed usage information.
Note: The .rar files are archives that can be opened with programs as Winrar or Winzip, for example.
Ref.:
Schwegmann, A; Lindemann, JP; & Egelhaaf, M (2014a): Depth information in natural environments derived from optic flow by insect motion detection system: a model analysis. Front. Comput. Neurosci. 8: 83. doi: 10.3389/fncom.2014.00083
Schwegmann, A., Lindemann, J. P., & Egelhaaf, M. (2014b). Temporal statistics of natural image sequences generated by movements with insect flight characteristics. PLoS ONE

本数据库收录了37组在自然环境中以广泛深度结构记录的全景高动态范围图像序列。每组序列包含最多100张图像，这些图像随后在一条一米的直线上以1至2厘米的间隔进行拍摄。因此，图像序列反映了平移运动。图像在方位角上为全360°全景，在高度角上介于地平线下58°至地平线上47°之间。因此，通过软件可以从单个图像中计算出任何旋转偏航运动。本研究采用了光谱滤波器来限制相机的光谱灵敏度，使其在480-560纳米（绿色）波长范围内。这种滤波模拟了光感受器R1-R6的光谱灵敏度，这些光感受器为昆虫运动视觉系统提供输入。因此，这些图像中颜色到灰度值的映射与RGB图像中的绿色通道相似。原始图像的分辨率为约100万像素（928x928），位深为12位。这些图像具有高动态范围，涵盖了自然环境中遇到的整个亮度范围（不包括太阳盘）。经过线性化后，得到的图像值动态范围为1:23,900，覆盖3,955个强度级别。请注意，尽管像素亮度值不能重新计算为SI单位如坎德拉，但这些值与绿色光谱范围内的亮度成正比。有关图像序列记录的更多技术细节，请参阅Schwegmann等（2014b）。此外，数据库还提供了展开和线性化的全景图像，分辨率为927 x 250。对于图像序列的每一帧，还提供了环境距离图，以厘米为单位表示每个像素的距离。关于距离图获取的详细信息，请参阅Schwegmann等（2014a）。此外，提供了一系列Matlab .m脚本，可用于处理存档中的数据。这些脚本由Schwegmann、Lindemann和Egelhaaf（2014）编写。请参阅数据存档中包含的readme.pdf，以获取详细的用法信息。注意：.rar文件是存档文件，可以使用Winrar或Winzip等程序打开。参考文献：Schwegmann, A; Lindemann, JP; & Egelhaaf, M (2014a): 从昆虫运动检测系统通过光流获取自然环境中深度信息：模型分析。Front. Comput. Neurosci. 8: 83. doi: 10.3389/fncom.2014.00083；Schwegmann, A., Lindemann, J. P., & Egelhaaf, M. (2014b). 具有昆虫飞行特征的移动生成自然图像序列的时间统计。PLoS ONE