Jump to content

jerpol

Members
  • Posts

    2
  • Joined

  • Last visited

Everything posted by jerpol

  1. Thanks, Greybeard. Your comment prompted me to delve deeper into what else is happening in the pipeline. Turns out that the RAF files have the correct image dimensions, although the orientation flag does change. We normally process through DXO Photolab and don't apply any changes to image rotation/orientation. However, it seems that Photolab swaps the width and height written in the EXIF information for any images that have a rotation flag that indicated portrait orientation! Naughty, naughty. That means I can put in place a workaround where we search for any images with an orientation flag that's different from 'Horizontal (Normal)' and use exiftool to overwrite the flag to the desired value of '1'. It's a bit tedious, but can be done. However, ideally I'd like to have the camera disable orientation detection and simply write all images as orientation '1'. Do you know if this is possible? Thanks for your help.
  2. I'm currently testing the X100V as a possible sensor for aerial image capture for photogrammetry purposes. For such applications, knowing the image orientation is critical and so normal practice involves disabling any on-camera auto image rotation. Cameras usually handle auto image rotation in two parts: Detection of camera orientation during capture, which determines what is written to the EXIF information; and Detection of camera orientation during viewing, which determines what is displayed (based on the file EXIF information). The first part determines what value is written to the image file's EXIF orientation flag (documented in various references). However, what these references don't mention is that the orientation also ends up affecting the image width and height, which are also written into the EXIF information. While the orientation flag can be modified using various tools (or settings on cameras or software packages) so that the images can be viewed in whatever orientation is desired, the image width and height do not change. So, why does this matter for photogrammetry? Well, a key aspect of the process involves determining intrinsic camera parameters that are expected to be identical for all images in the set. One of these parameters is the image size, so we need the width and height to be the same for all images! When a camera decides to randomly go and write different values for the width and height depending on how the camera is oriented (something that's hard to control in aerial capture), that becomes a big problem. I use a Nikon D850 for a lot of my work and that camera provides the ability to properly disable auto image rotation, i.e. the camera writes exactly the same EXIF information for orientation, image height and image width no matter how I hold the camera. The Sony mirrorless cameras I've tested, on the other hand, don't seem to provide that same capability. Now, in testing the X100V, I'm concerned that I may be facing the same problem. From what I understand of the X100V, there are only two settings relating to camera rotation that can be controlled: Display Rotation, which sets whether the viewfinder and LCD monitor indicators are allowed to rotate to match camera orientation; and Autorotate PB, which sets whether images are displayed in the "correct" orientation during playback. The first setting is irrelevant to this discussion. The second setting only affects the second part of the image rotation problem (during viewing). I've looked carefully through the X100V menus and documentation and searched various forums, but cannot find any way to completely disable image rotation detection and the subsequent impact on the EXIF information. Very much looking forward to hearing from anyone that's delved into this topic and has more familiarity with the X100V than I do!
×
×
  • Create New...