Apr 28, 2006

Send us your poor, your tired, your haloed images…

…yearning to blend free.

Source: John Nack on Adobe (John’s Adobe Blog)

We’d like to ask your help in improving HDR (high dynamic range) imaging in Photoshop. The halos produced by many current HDR conversion techniques (see the Flickr HDR pool for some examples) are kind of cool and wonky, but to make HDR more than a fad, we need to produce more reasonable results.

With this in mind we’d like to get sample images–particularly ones with which you’ve gotten better results converting 32->16/8 bits using another package than you have using Photoshop. Photoshop engineer John Peterson writes,

I’m looking for cases where the “other leading brand” is doing a better job than Photoshop. I’d like to get three or four really good cases of this from customers that are (potential) heavy users of Merge to HDR. I’d be interested in JPEG or raw source files, plus the HDR result file from the other application. JPEGs should be generated by the camera, not via Camera Raw. f-stop should be held constant, exposure should differ by two stops or so, and resolutions in the 2-6 MP range would be sufficient.

Check John’s blog entry to get contact info…

HDR composite by Jeff Schewe

2 Responses to “Send us your poor, your tired, your haloed images…”

  1. JIm Scott Says:

    Hi John!

    I do a lot of QTVR work using bracketed images where I can have, as you might expect in a 360 x 180 degree panorama, a huge variation in lighting. I almost laugh at situations where “still” photographers struggle over one image – try 31 images stitched together in your final picture!

    I think the methodology is flawed here.

    Instead of having everone send in their examples – there should be just one set, as in the featured images above (which I could not download) – and then have people work them over so there is an “apples to apples” comparison.

    My favorite HDR “program” is in fact a Photoshop filter – “Optipix” from Reindeer Graphics (

    There are two ways to do HDR in this filter -both are powerful in that they let you continue to add exposures (in whatever combo) from the same sequence into the mix to get the result you are looking for – instead using of a single sequence as it is to output an image.

    For example using Optipix with (7) bracketed images (1 = darkest; 7 = lightest)
    Run all the images through the filter – result a little too light washed out? OK, you can “add” a darker image (say #2) from the sequence in *again* – or as many times as you want. Or add #2 & and then #3 to change the output.

    The problem I have found with most “automated” HDR output is that one size does not fit all. You need to have the option of manual intervention.

    But really, all considered, I still need to use masks to achieve the exact look I want… but Optipix gets me most of the way there (major exception: interior shots with windows featuring bright light outside, i.e. blow hightlights – must be done completely manual in these areas).

    In conclusion:
    The content creator needs to have the ability to manually augment the recipe for the “perfect” HDR image IMO. Total automation will never work in this situation.

  2. Chris Cox Says:

    Jim — no, John asked for sample images so Adobe can check the results of their software and improve their software. Having everyone convert one set of images wouldn’t help Adobe improve their software in any way (and Adobe’s already been doing that with their own test images).

    Also, the only automated part of making HDR images is the merge of exposures – and it should be automatic because there is only one right answer (within a constant multiplier). And Photoshop does give you controls (4 algorithms representing the best available when the feature was written) for converting 32 bit images down to 8 or 16 bit images.

    And what you describe with Optipix isn’t High Dynamic Range – just adding more images together to get the result you want (more like an average). In theory, there is only one perfect HDR image for a scene — one that is proportional to the photons that hit the camera in that scene. Then the user chooses how they want to convert that HDR image to a Low Dynamic Range image (toning). But there are a lot of open questions about how to deal with edge conditions, movement/changes in the scene, and noise — and Adobe wants to make sure they have the best possible answers and can handle the problem images.

    So, please, read what John posted again, and see if you can help improve the software.

    And if you still don’t understand what I’m saying, there is a good book on High Dynamic Range Imaging (in fact, that’s the title) available from Amazon that goes into all the details.

Leave a Reply

You must be logged in to post a comment.