Preparing Pictures Before Using RC (Raw)

Comments

12 comments

  • Avatar
    Götz Echtenacher

    Hi Mike,

    nice topic!

    And to hopefully get a vivid discussion started I will be controvercial:

    After feeding RC with the exact same handful of images, same alignment settings and a few GCPs and High reconstruction I literally cannot pick out any meaningful difference between: camera JPG (I'm not kidding), RAW (exported as JPG 92%) with only exopsure corrected (I shot slightly underexposed) plus color noise corrected, the same plus blacks pulled up, another one with additionally adjusting sharpness, local contrast, micro contrast and luminar noise reduction and finally all that but exported as TIF. The difference is literally on a pixel level, meaning if you do pixel peeping and switch back and forth, you can see pixels shifting a bit. That's it. And again, only by a pixel! The geometry is also essentially identical, which is no surprise since it's literally the basis for the texture. It really surprised me. The differences are academic in my view and not relevant in my daily use, which is based on accuracy. The only difference worth mentioning is of course the color spectrum of the texture.

    Fuji X-T20 with Fujinon 10-24 and RawTherepee 5.4

    0
    Comment actions Permalink
  • Avatar
    Mike Simone

    I would have thought the same, but I have noticed a HUGE difference when using the settings above.  Its night and day for alignment, reconstruction and texturing. Right now I am religiously using Highlights, Shadows, Whites, Blacks and Clarity. 

    Keep in mind i'm doing mostly interior models but I would say to anyone they should try to play around with these settings.

    Here are some of the results im getting (I have cleaned up the model here using Modo, but everything is razer sharp because of the pictures.

    https://skfb.ly/6AIXK

    https://skfb.ly/6AItR

    0
    Comment actions Permalink
  • Avatar
    Götz Echtenacher

    Ah, that's what I wanted to add: My sample was all well lit to begin with and also a sculpture.

    But I totally agree with you that it is all about trying it out and judge for oneself!

    I also agree with you that for interiors, especially cranking up the shadows can make a huge difference.

    Just so that we're clear: are we talking about the texture or the geometry?

    Because on the one hand, your results look amazing (*envy*) but that is largely due to the fact that you rebuilt 93% of the geometry, right? Cleanup is putting it mildly!  :-)  That's no criticism, on the contrary. How much time did you spend doing that? And did you re-texture in RC?

    Maybe the difference is more pronounced with "not-ideal" image sets? Bearing in mind that interiors have a natural limitation as to how ideal they can get - I would argue 50% is already pushing it. There are just way too many restricting factors like occlusion, material, very bad angles (as in almost parallel to surfaces) etc...

    0
    Comment actions Permalink
  • Avatar
    Mike Simone

    It helps for both the geometry and that ends up helping a lot with the texturing. If you have poor geometry, that normally means you dont have enough pictures or your alignment is off. Even ever so slightly, a poor alignment makes everything very difficult later when using your own low poly meshes.

    That family room I completed in about 4 hours of my own time but the kitchen took a couple days (Longer as I was learning some better modelling techniques)

    For interiors you will always need to rebuild at least 50% of what RC reconstructs but you use the RC mesh as a proper outline for your work. Now back to alignments, if your alignment is off by even a little, no matter how much you model on your own, its veeeeery hard to get sharp images without some sections being blurry or areas being uneven.

    So for best practices and results, always nail down your images first. :) (That being use RAW and optimize your results by using recommended settings such as I have provided)

    But I know there are much better experts then me on these forums :) hopefully they can chime in here as well.

    Cheers!

    0
    Comment actions Permalink
  • Avatar
    Götz Echtenacher

    4 hours is not bad! Thanks for hunmoring me.

    Hmm, I'm just not entirely convinced that it will help overly much with the alignment in general. In my experience it's more about getting better results in the dark corners.Your samples seem to be on the dark side (may the force be with you) and maybe that explains why it made such a difference? Would you mind sharing a before and after sample?

    0
    Comment actions Permalink
  • Avatar
    sairus.tro

    After experimenting with RC I am sorta reaching the conclusion that raw files (I am shooting Canon 600D) directly being fed to the program end up with the best results for me.

    It is rather refreshing because I do not have to worry about processing of the jpeg images (I have done the same things earlier as you guys, correct WB, bring up the blacks, pull down the whites, crank up clarity) and exporting into jpeg, but this way I ran into alignment issues... going directly raw seems to solve it pretty much.

    I watch out to have as close exposures as possible throughout the shots (therefore I am shooting in aperture priority).

    Just gotta make sure your in camera WB is right and you're set.

    Anyone with a similar conclusion?

    0
    Comment actions Permalink
  • Avatar
    Mike Simone

    I always use TIF and keep the same level of detail as the raw image. I find JPG never gets me good results.

    0
    Comment actions Permalink
  • Avatar
    lukasas74

    Yeah, me also using similar settings but I avoid doing clarity and other "more contrast" tools. Sometimes that makes my textures bit off.

    0
    Comment actions Permalink
  • Avatar
    Götz Echtenacher

    Amazing how different the experiences can be!

    I wonder if it's not just the person in front of the computer and behind the camera that's causing those differences but maybe also the camera itself?

    Would anyone be interested in a little experiment? Each of us would shoot a little 10 image set and then we all try to get the best (whatever that may be) results with each set?

    0
    Comment actions Permalink
  • Avatar
    sairus.tro

    Of course Götz!
    The camera and.. the settings!
    I'm down with the experiment.. expect images later today!

    0
    Comment actions Permalink
  • Avatar
    andrew breckill

    I use a Linear colour space for my workflow, which is imported as a texture file into Maya or Houdini for rendering. My setting use the 2010 ACR profile after RAW image capture using gretag macbeth colour chart. My setting are everything to zero in the basic settings, use the colour chart to set the whites mids and bolack points in the tone curve shelf, lens corrections turn distortion correction off, vignetting upto 200, remove CA. camera calibration uses the 2010 preset and the preset made for my camera with the colour checker software. This workflow is to give a linear colour space to work correctly with Mantra and Arnolds shaders.

    0
    Comment actions Permalink
  • Avatar
    PANORAMA-ULB

    Having made a lot of test, I always correct my shoot (approximatively the same ways as in the head of this post) and recover a lot of holes, mesh pinch, and some more CP by picture + better color (it's normal because I work with a color chart too). I edit raw (Nikon D810 uncompressed) and use jpg without compression. I also correct the distortion (LPC made by our lab with different calibrations) and gain some % on error (little but better than using RC correction).

    0
    Comment actions Permalink

Please sign in to leave a comment.