Hi Clarence,
I think it is common to have better success with fewer images instead of a full 
15-image sequence. Having extra frames is probably good practice since you 
might not get another chance to capture that scene, but you shouldn't need that 
many for the actual HDR creation. You mentioned trying images 6, 9, 10, and 11 
as example. The important thing would be to select images that capture as much 
of the low, middle, and high exposure range as possible - particularly 
important for a new camera response function. Let's say your image exposures 
are in order from darkest to brightest. Your selected sequence should hopefully 
include at least one of the first few under-exposed images - or whichever one 
is dark enough to have some information but no fully white pixels. You should 
also include at least one of the last few over-exposed images - hopefully one 
with image information but no fully black pixels. Then pick a few of the 
exposures in between. You might find better success with four or five images 
instead of 15.

You might also try changing some other settings regarding automatic alignment, 
ghost removal, lens flare removal, etc. if these or any other default settings 
might adversely affect your specific image sequence.

-Chris

From: Clarence Wang [mailto:[email protected]]
Sent: Monday, October 29, 2018 10:32 PM
To: High Dynamic Range Imaging <[email protected]>
Subject: [External] [HDRI] hdrgen problems

Hi Greg,

I've subscribed the HDRI mailing list and the request seemed to be approved.

For the questions mentioned before, I got failure to derive the camera response 
function when running a whole set of images (1-13) in Photosphere. Thus, I just 
selected 9 images in sequence (1-9) for rendering HDR image.

However, it was confusing that I still failed to derive the camera response 
function using the other sequence of images (3-11) which was successfully run 
using 'hdrgen' tool and was to generate a new camera response function (the 
second-order polynomial). Since you mentioned "even the same set of images 
given twice won't necessarily yield the same coefficients, although they should 
be close", which function or rendered HDR image is reliable? Or, can we trust 
the HDR image rendered with different polynomials but well agree with each 
other (almost overlap)?

Also, If both Photosphere and hdrgen tool cannot solve response function and 
fail to render a HDR image using a sequence of 15 exposed images, is it 
reasonable to select less number of images (say, 6, 9, 10, 11) to yield the HDR 
image and relevant camera response function using Photosphere or hdrgen tool ?

Thanks,

Clarence
 ____________________________________________________________
Electronic mail messages entering and leaving Arup business systems are scanned 
for viruses and acceptability of content.
_______________________________________________
HDRI mailing list
[email protected]
https://www.radiance-online.org/mailman/listinfo/hdri

Reply via email to