Page 1 of 1

Render sizes

PostPosted: Mon Dec 17, 2012 2:47 pm
by kimaldis
I've a set of 57 images in 3 EV sets. I'm stitching each EV set into a panorama so I can merge the resulting 3 into a single HDR image. The requirement of the merge tool is that each panorama be identical in size. Despite the fact each set is identical, other than exposure (tripod shot) the final render size of each is different. How can I ensure that each image is rendered to the same size. I've tried cropping but that's specified in fractions, not pixels.

thanks.

PostPosted: Mon Dec 17, 2012 3:44 pm
by marzipano
I seem to remember having this problem once and I used Irfanview (free download) to specify exact width and height pixel dimensions.

If I recall you can choose whether or not to override the aspect ratio if the shapes of each of your panoramas is slightly different

good luck
Martin

PostPosted: Mon Dec 17, 2012 4:14 pm
by klausesser
kimaldis wrote:I've a set of 57 images in 3 EV sets. I'm stitching each EV set into a panorama so I can merge the resulting 3 into a single HDR image. The requirement of the merge tool is that each panorama be identical in size. Despite the fact each set is identical, other than exposure (tripod shot) the final render size of each is different. How can I ensure that each image is rendered to the same size. I've tried cropping but that's specified in fractions, not pixels.

thanks.

Stitching each set individually wouldn´t work. No two seperate stitches from the same source-images are identical.

You need to import all the bracketed images and use them as stacks for stitching them. This way you can fuse them in APG or export them as .hdr/.exr or you can export them as seperate layers for processing them to hdr in a seperate HDR-application.
But exporting them as .hdr/.exr basically does the same - as long as you DO NOT use color-correction in any way in APG!

best, Klaus

PostPosted: Mon Dec 17, 2012 4:19 pm
by kimaldis
Thanks for the help Martin,

As it turns out the panoramas themselves don't quite match up either, there's a slightly different curve to the horizontals on each one so it's kind of academic anyway. In the end I hacked Photoshop Merge to HDR Pro script and built a batch tool to do the merging for me in Photoshop before bringing the slices back into Autopano for stitching.

I don't know if this is normal but I'm finding HDR in Autopano unpredictable and I'm getting much better results by keeping the HDR in Photoshop.

PostPosted: Mon Dec 17, 2012 4:23 pm
by kimaldis
Thanks but I'm having problems letting Autopano merge the exposures, Klaus. I can't get the same amount of highlight detail as I can with a PS merger and if I let APG deal with the HDR it becomes unpredictable.

PostPosted: Mon Dec 17, 2012 4:28 pm
by marzipano
kimaldis wrote:Thanks but I'm having problems letting Autopano merge the exposures, Klaus. I can't get the same amount of highlight detail as I can with a PS merger and if I let APG deal with the HDR it becomes unpredictable.

If it's any consolation I found exactly the same problem although I always suspected I was using an incorrect option somewhere along the line. The problem is there is a myriad of different choices and no clear documentation to give you a leg-up to start

I bit the bullet and got Photomatix so I now use APG to stitch the pre-fused images and happy with the results !

best
Martin

PostPosted: Mon Dec 17, 2012 4:37 pm
by kimaldis
glad I'm not the only one. I'm putting down to unfamiliarity, to a degree, but I'm usually pretty good at figuring stuff out and when a render looks vastly different to your preview or sliding sliders has no effect, you begin to wonder.

I think it worked out OK in the end though: http://www.flickr.com/photos/kimaldis/8281540200/

PostPosted: Tue Dec 18, 2012 12:25 am
by marzipano
Hello Kim

That certainly looks to have to have turned out well after your perseverence

What workflow did you end up using and what APG settings gave you that result ?

PostPosted: Tue Dec 18, 2012 10:02 am
by kimaldis
It's a test really, I need to do some work on the post, it's a bit overcooked for my taste. However ... I'm only stitching in APG. Maybe it's just me being unfamiliar with the process but I'm really not getting on with the HDR workflow in APG, there's something really not right about it and I'm not convinced its exposure merging is that great. (sorry guys)

I'm using Photoshop to do the merge to HDR, giving it 16 bit TIFs and having it output 32 bit EXRs for AGP. There's 20 images, 3 exposures each. I didn't want to merge each of those by hand and I'm going to be doing more of these and I needed a slick pipeline so I built a photoshop script that takes a folder of ordered images - exported from Lightroom as 16 bit TIFs and merges them using photoshop's mergeToHDR. These are passed over to AGP for stitching and the result rendered out to EXR. I then use PS to turn it into 32 bit TIF and import back to Lightroom. This particular shot was posted entirely in Lightroom; sometimes I'll take shots back into PS for detailing but this one was more by way of a test.

The script isn't quite ready for public consumption but I'll have it online when it is. Although it's pretty straightforward.

PostPosted: Tue Dec 18, 2012 10:06 am
by kimaldis
Klaus: I'm curious, why wouldn't it produce consistent results each time with the same set of input images? I don't see off the top of my head why there should be anything stochastic about such a process.

PostPosted: Tue Dec 18, 2012 2:22 pm
by gkaefer
kimaldis wrote:Klaus: I'm curious, why wouldn't it produce consistent results each time with the same set of input images? I don't see off the top of my head why there should be anything stochastic about such a process.

try it with 5 images and you'll see it. the CPs are not pixel exact placed each try... so in result the pano resulting differs too...
Georg

PostPosted: Tue Dec 18, 2012 8:21 pm
by kimaldis
Sure, I get that, but why?

PostPosted: Tue Dec 18, 2012 8:27 pm
by HansKeesom
I agree that logically it follows that the same set of photos should always give the same result. so the remark of George suggest the use of some random generator being used which sounds a bit crazy, althought there might be a reason behind this,

However, if you stitch each layer seperately you are not using the same data each of the 3 stiches, therefor it is logicall that the three layers dont match. If the photos where taking on a tripod you migh fuse each stack first and then load the into .