Hi All I posted this in my WIP thread but thought this would be a better place to ask!
I wanted to get some different HDR’s for lighting that I wouldn’t have copyright issues with so decided to try making some myself, I made this over the weekend I think I have it correct I have full res version as well this is abut 3rd of the size at 8k the origional is around 23k! It comes out a bit darker than others I have used have to pump it up to around 50-100 to use it but it looks balanced as I do that I’ve been running with 2048 map resolution and multiimportance settings in the world settings.
I think it’s dark just because of the dynamic range I captured this was about 80 pictures each one made out of 7 RAW photos combined shot at 1 stop intervals so -3 to 3 then stitched it all together you can def crank it up without blowing out like a jpg/png etc… I think it’s just because it was a challenging scene to photograph and I couldn’t get it in one shot my camera only does around 7 stops I’d guess I didn’t meter it but probably 13-14 stops in reality.
I’ve done another one in a wooded scene I’m working on and that comes in around the 1-2 emission range which is more like others I have used but it was overcast day and pretty flat lighting anyway.
If anyone has a sec and want’s to give it a test run and let me know if it works ok for them, I’ll post it on blend swap under a creative commons licence once I know it’s good.
This is what it looks like, quite funky lighting which is why I wanted to get it, don’t use this image it’s just tonemaped JPG will look rubbish use the dropbox link above around 128meg. Can’t help the slight purpling it’s the uv filtering on the digital camera at some angles with a wide angle lens that’s pretty challenging lighting conditions tbh, had my trusty leica 15mm prime out
Thx I used Photomatrix pro to merge the exposures and create a tone mapped proxy and PTGui pro to stitch the proxy then the exr, basically followed the tut here http://adaptivesamples.com/2013/11/17/how-to-create-your-own-hdr-environment-images/ both sets of software are newer now so it’s slightly different but easy enough to figure out!
Heres the other one I did, the sky looks blown out in the HDR but it’s not as you can see from the jpg I produced from it dropping the highlights with colour curves. This is a really flat light but that’s what it was overcast and hardly any shadows and that’s what gets rendered, think this one would obviously be good fow lowlight woodland scenes you could pump the greens a bit etc. Didn’t have to run this as high just 4-5 I think because overall it’s a far flatter image in general which makes sense, getting a feel for it now! Might do some more at the weekend.
Link to HDR
Example jpg don’t use this one use the HDR file from the link
Thanks for sharing! You are aware of it, but the first one is indeed too dark. The second one is pretty good though. Much better than most of the (free) ones out there. If you want I can share some tests.
Thanks for feedback! I can’t decide on the first one, I know it’s dark but when I increase the strength that’s linear and when I get to about 50 it looks about right and all the detail is there and the highlights look ok, it’s a really challenging scene to photograph and both the images are tone mapped and don’t reflect what it really looks like i.e. it really is very dark to very bright, let me have a look at it again tomorrow now I know a bit more after doing the second one now and see if I screwed something up
I’m as always happy to be wrong But I’ve gone back over the creation process and looked at the intermediary files and I’m pretty sure the first darker hdr one I posted is correct for what it is. That’s the stitched together intermediary exr files created when you merge the bracketed raw files. All the colour details in there if you drop on an rgb curve for instance you can pull back all the shadows and highlights etc no problems and it looks fine in scopes and it’s what all the tone mapped images were based off. I could apply a curve and re-save it as an hdr again so t looked more like people might expect but I’d rather just leave it raw and if anyone wants a different look they can just tweak it in the environment. Thinking about what I want to try and get at the weekend now
That barn’s a tough scene for a first attempt. You got good detail in the dark areas but you didn’t have a wide enough dynamic range to capture the highlights. If you zoom in on the skylights or the light fixtures and adjust the exposure you’ll see they’re blown out. If you primarily want to use these for IBL it’s those light sources that you really want to capture. In an image that fully captures the dynamic range you should be able to see the interior of the light fixture and the filaments of an incandescent bulb.
If you’re shooting raw you don’t need to shoot brackets at 1 EV intervals, jumping up to 2 EV intervals should be fine and will vastly increase you’re dynamic range for the same number of shots. If you’re shooting jpgs sticking to 1 EV intervals is probably a good idea.
You’re images are dark, but they’re not technically wrong. There is no hard rule about how HDR pixel values must map to measured luminance values in the the scene, however there is a convention. That convention is that a pixel value of 0.18 corresponds to an 18% gray card and 1.0 corresponds to a diffuse 100% reflecting surface. Any pixel value above 1.0 is either a light emitting object or a specular reflection. The brightest pixel in your barn scene has a luminance of 0.26.
Fortunately, your image still contains all of the necessary data, it’s just a matter of normalization. Suppose that when you shot the panorama you left a gray card or ball sitting on the ground near your tripod. After creating the HDR panorama you measure the pixel values on that card and they are approximately 0.004. You know that they should be 0.18 and that the 0 end of the range (no light) is fixed, therefore it is a simple matter to figure out what you need to multiply the measured value by to get the target value (0.18/.004). For this example, that works out to 45 so you multiply all of your pixel values by 45. Note that after applying this multiplication your data is still a scene-referred linear measurement of light within the scene and therefore suitable for IBL. This is different than if you were to simply apply an RGB curve to the image to make it look more pleasing. The curve transform would not necessarily be linear and therefore the resulting image may no longer be scene-referred, making it less useful for IBL, as you’ve rightly intuited.
For extra credit add an object with an 18% gray material to your CG scene and see how closely you can get it to match your on-set reference.
Awesome feedback thanks jedfrechette totally makes sense how you have explained it, I’m going to reshoot it at the weekend, I don’t have a setting for 2 stops but my camera exposes -3 to +3 and I can get off 7 raw at 1 stop intervals which I guess is probably a lot of overlap but that gives me the best overall range I can get with it looking at the available settings.
I exposed for 0 like I would if I was averaging a single shot in there then went 3 over 3 under I’ll expose at-3 making sure I get the highlights next time then calculate 0 from that as a baseline and shoot from there I think that should get it in there. I’ve got grey cards etc so can get some reference shots, the maths makes sense and corresponds to what I saw visually i.e. around 50 it looked like I would expect in the environment map. I’ll reshoot the wood one as well if I get time and better expose the sky well I’ll see what nature deals me!
Thanks so much for that that’s the detail I was missing really helpful, good fun shooting these
edit, was just playing around to get the filament I’m going to have to go to manual bracketing I think to get that one shot, is that normal or instead of doing 7 1 stop perhaps I set up 2 custom manual exposures and do it like that one to get more range, will have a think that’s trickier than I was expecting but not insurmountable! Just thinking about the quickest way to get the shots.
When shooting HDRs for IBL it isn’t unusual at all for your ‘normally’ exposed image to be at the very end of the bracket sequence. It can be quite surprising how much brighter the lamps are than the rest of the scene. For IBL you don’t really care about what’s in the shadows so that’s fine but for photographic purposes you might want to do a double burst of brackets to capture the full range.
You’re well on your way and half the fun is figuring out a system that works reliably for you. One resource I would recomend is hdrlabs.com, where you’ll find lots of good technical tips.
Ok that makes sense did some quick tests and I think for that scene I’ll have to do 2 bracketed burst shots I can get my shutter speed up to 1/8000 and min 1sec in exposure burst mode so I can get 14 full stops over 2 exposure bursts giving me 14 images, I could get away with 7 to cover it with raw but there is not that spacing setting in burst mode I can always just drop every other image after I’ve got 64 gig fast cards so can manage that volume so may as well grab them. Thx!
Will try at weekend, my next task is to understand exactly how the software calculates the luminance for the raw hdr after it’s fused the images together, I can’t get my head around how it knows how bright the image is relative to the other images if say one setr of images burst has 7 stops of information and another set pointing at the floor has say 3 stops how does it know where that 3 stops of data should sit from a luminance point of view. It seems to just work but I don’t like magic happens here as I’ll get caught out down the line This is more of a rhetorical question I’ll do some research and reading it must be buried somewhere how it calculates it, perhaps it uses the camera exif as well or perhaps the panorama code is levelling it I’ll figure it out!
Thanks for the help much appreciated I’ll hopefully get a revised one up over the weekend and see what that looks like!
