Posted by John Freeman on November 25, 2014 at 4:52 PM
This thread has been moved from the Landscape Forum for a post I made titled "Bumper Crop". Mark Seaver suggested I move it here. Most of my photography work is exclusively stitched images. On occasion I find some banding that is especially noticeable if the scene has nearly perfectly blue skies as in the example photo attached. I am trying to identify the source. The series of 30 photos that this image was stitched from were shot at f20, 150th of a second, ISO 200 and 200mm with my Canon 20-200 f2.8 IS lens. I was using my Canon 5Diii which is about 1-1/2 years old.
As noted above I notice vertical bands of brightness changes. I have had this show up in some of my stitched images. I use Autopano Giga Pro for stitching. In it's blending process it seems to average the luminosity in vertical bands if you have lighter or darker objects in particular area resulting the sky brightness varying. In a nearly cloudless sky this becomes quite noticeable. If anyone out there is doing stitching I would be curious how you are doing your brightness match and what software you use. Also, I have been trying to determine if this issue is exacerbated when I do image size reductions using Photoshop's image resize. It doesn't seem to be as prominent when I am viewing the original full size image. Here is a link to a high resolution viewer of another shot I did the same morning. http://www.abbascreationsphotography.com/Pumpkin_Field_Tour/Pumpkin%20Field.html (Rolling your mouse will allow you to zoom into 200%)
I went back into my raw files to verify that I was in full manual shooting mode when I shot these. I was. My process flow is to use Camera Raw to batch process the RAW images to tif files. When shooting the series I will typically use "Expose to the Right" picking the brightest portion of the scene and set the exposure to about +2 on the exposure meter. I do this to get the best possible noise in the dark areas of the image. Then the first step in Camera Raw is to move the exposure slider so the histogram has no over exposed areas indicated. In the original image I ended up setting the Exposure to -1.85. I then select all the images and Synchronize the setting to all images and export as 16 bit TIF files to a new sub-directory. I then stitch the image using Autopano Giga Pro.
As I reviewed each individual image of the top row with mostly sky, I noted that the histogram for the images that exhibit bright bands have most of the image information shifted further to the right. I am wondering if my camera is needing cleaned and thereby not giving me correct exposure. Also, I believe that I have read somewhere that with higher f-stop values greater that f-11 (f-20 in this case) there is more exposure variation because the lens doesn't do a good job of closing the aperture to the same size every time. I was using higher f-stop of course to get the best depth of field possible.
Has anyone seen this?
Posted by Mark Seaver on November 25, 2014 at 11:16 PM
John, my biggest pano stitch to date is 11 vertical images stitched into one pano. I also do a lot of stacking where it may take as many as 30 or 40 images to get an extended subject sharp. Especially in the stacks where there should not be any brightness changes (unless the sun goes under a cloud), I routinely see 1/2 to 2/3rds of a stop variability in the histogram even though the shutter speed has stayed the same (even in manual mode). I know that there is a certain amount of variability allowed in the actual shutter speed relative to the recorded speed and I think that this is most likely the culprit. In both pano stitches and in stacks, part of my RAW processing (in Lightroom, but it's the same as Adobe Camera Raw in PS) routine is to adjust the exposures so that all of the images that will get combined have similar brightnesses and the same white balance settings. Often, there's a peak in one of the colors that can be used for these adjustments.
In shooting panos, exposure adjustments are more challenging because there may be regions that should be darker or lighter. Even then, your eyes do a very good job of matching exposures if you quickly shift back and forth between images.
I'm shooting a 2.5 year old 5DIII
Posted by Gary Minish on November 26, 2014 at 8:05 PM
John, I use AutopanoGiga quite a lot and normally don't have that problem but I have seen it on occasion. One quick check, first of all: look at the detection tab in the settings dialog box and make sure the "auto color correction" box is checked in the "automate" section (that's the only thing I turn on in that section). Beyond that, the most common cause of banding seems to be a change of white balance between frames. Most people think to set the camera to full manual and lock the exposure but forget about the white balance. if the white balance is in auto it can significantly change the colors from one frame to the next as the light is different at different angles. Gamma/color correction is the most difficult obstacle for the software when it comes to blending the images. Even when I forget to set the white balance to manual the software usually does an amazing job of compensating and blending though. On the few occasions where it's had troubles I've been able to fix it by manually setting the color adjustment method for each frame. To do that, after you've opened the stitched pano in edit mode, click on the rainbow circle icon that has "auto" in the middle. You will then see markers in each frame and one of the frames will have a thumbtack in it. That frame is the anchor and everything else is adjusted based on it. You can right click on the markers and change the adjustment options for each frame. You probably want to look it up in HELP to understand what you're doing but you can also just try different choices and undo if it doesn't work.
Side note: If the aperture and shutter speed are locked in manual the histograms for each image will still be different because each frame is seeing a different scene. Therefore I don't think your camera is having problems. You're doing the right thing by setting the exposure using the brightest scene if you don't want to lose the highlights.
Hope that helps you
Posted by Roger Clark on November 27, 2014 at 8:48 AM
Hello John,
I do a lot of mosaics (e.g. there is a 54 frame mosaic on my home page now). There are two things that can cause this problem. In your raw converter, are you enabling the lens profile and correcting for light fall-off? If not, that will definitely cause the problem. It will not be noticeable in areas with detail (the field) but will be in the sky. In ACR turn on lens profile corrections and keep the light fall-off at 100%.
The second way this problem can show is if you are saturating one channel, and in this case, the blue channel. In your metering, are you showing the red, green and blue histograms on the camera and making sure the blue is not saturating in the blue channel? Your vertical banding on the sky is bright bands, it indicates either you overcompensated the light fall-off or your blue channel is saturating. When a channel is saturating, and you apply fall-off correction it can impart a slope to the saturated channel. It does not appear to me that exposure is your problem, at least in the presented image.
If you applied lens profile and light fall-off correction (popularly called vignetting but that is incorrect use of the term), it may be that the lens profile data is not as good as it should be.
Finally a comment on using f/20. In these kinds of mosaics, there really is no need to use such a slow f/ratio. I let the focus point change with each frame. One only needs enough depth of field for all the objects within one frame to be in good focus. For this reason, I keep the camera in landscape orientation so the top to bottom of the frame is a smaller focal range. It looks like your scene could have been done at about f/11 and thus produced a sharper image with higher contrast in the fine details (less impact from diffraction).
More on this subject: http://www.clarkvision.com/articles/large_mosaics/
The 54 frame night scene currently on my home page was done at f/1.4 by shifting the focus on the near objects at the bottom of the scene. As this is a night scene, the focus was done manually with each frame. In a daytime mosaic, I select the center focus point and let the AF system focus, then turn off AF when reaching sky.
The 54 frame mosaic is also here (it will be off the home page soon):
http://www.clarkvision.com/galleries/gallery.nightscapes/web/santorini.night.rnc06.07.2014.IMG_5176-424_54frames.e-bin12x12s.html
Discussion
3 repliesAs noted above I notice vertical bands of brightness changes. I have had this show up in some of my stitched images. I use Autopano Giga Pro for stitching. In it's blending process it seems to average the luminosity in vertical bands if you have lighter or darker objects in particular area resulting the sky brightness varying. In a nearly cloudless sky this becomes quite noticeable. If anyone out there is doing stitching I would be curious how you are doing your brightness match and what software you use. Also, I have been trying to determine if this issue is exacerbated when I do image size reductions using Photoshop's image resize. It doesn't seem to be as prominent when I am viewing the original full size image. Here is a link to a high resolution viewer of another shot I did the same morning. http://www.abbascreationsphotography.com/Pumpkin_Field_Tour/Pumpkin%20Field.html (Rolling your mouse will allow you to zoom into 200%)
I went back into my raw files to verify that I was in full manual shooting mode when I shot these. I was. My process flow is to use Camera Raw to batch process the RAW images to tif files. When shooting the series I will typically use "Expose to the Right" picking the brightest portion of the scene and set the exposure to about +2 on the exposure meter. I do this to get the best possible noise in the dark areas of the image. Then the first step in Camera Raw is to move the exposure slider so the histogram has no over exposed areas indicated. In the original image I ended up setting the Exposure to -1.85. I then select all the images and Synchronize the setting to all images and export as 16 bit TIF files to a new sub-directory. I then stitch the image using Autopano Giga Pro.
As I reviewed each individual image of the top row with mostly sky, I noted that the histogram for the images that exhibit bright bands have most of the image information shifted further to the right. I am wondering if my camera is needing cleaned and thereby not giving me correct exposure. Also, I believe that I have read somewhere that with higher f-stop values greater that f-11 (f-20 in this case) there is more exposure variation because the lens doesn't do a good job of closing the aperture to the same size every time. I was using higher f-stop of course to get the best depth of field possible.
Has anyone seen this?
In shooting panos, exposure adjustments are more challenging because there may be regions that should be darker or lighter. Even then, your eyes do a very good job of matching exposures if you quickly shift back and forth between images.
I'm shooting a 2.5 year old 5DIII
Side note: If the aperture and shutter speed are locked in manual the histograms for each image will still be different because each frame is seeing a different scene. Therefore I don't think your camera is having problems. You're doing the right thing by setting the exposure using the brightest scene if you don't want to lose the highlights.
Hope that helps you
I do a lot of mosaics (e.g. there is a 54 frame mosaic on my home page now). There are two things that can cause this problem. In your raw converter, are you enabling the lens profile and correcting for light fall-off? If not, that will definitely cause the problem. It will not be noticeable in areas with detail (the field) but will be in the sky. In ACR turn on lens profile corrections and keep the light fall-off at 100%.
The second way this problem can show is if you are saturating one channel, and in this case, the blue channel. In your metering, are you showing the red, green and blue histograms on the camera and making sure the blue is not saturating in the blue channel? Your vertical banding on the sky is bright bands, it indicates either you overcompensated the light fall-off or your blue channel is saturating. When a channel is saturating, and you apply fall-off correction it can impart a slope to the saturated channel. It does not appear to me that exposure is your problem, at least in the presented image.
If you applied lens profile and light fall-off correction (popularly called vignetting but that is incorrect use of the term), it may be that the lens profile data is not as good as it should be.
Finally a comment on using f/20. In these kinds of mosaics, there really is no need to use such a slow f/ratio. I let the focus point change with each frame. One only needs enough depth of field for all the objects within one frame to be in good focus. For this reason, I keep the camera in landscape orientation so the top to bottom of the frame is a smaller focal range. It looks like your scene could have been done at about f/11 and thus produced a sharper image with higher contrast in the fine details (less impact from diffraction).
More on this subject: http://www.clarkvision.com/articles/large_mosaics/
The 54 frame night scene currently on my home page was done at f/1.4 by shifting the focus on the near objects at the bottom of the scene. As this is a night scene, the focus was done manually with each frame. In a daytime mosaic, I select the center focus point and let the AF system focus, then turn off AF when reaching sky.
The 54 frame mosaic is also here (it will be off the home page soon):
http://www.clarkvision.com/galleries/gallery.nightscapes/web/santorini.night.rnc06.07.2014.IMG_5176-424_54frames.e-bin12x12s.html
Roger