Input to the HDR software - created exposures

[Colin Southern]

Hi Dave,

15 bit (which Photoshop uses for 16 bit mode) - 14 bit, 12 bit, 8 bit etc define the same "perfect black" to "perfect white" range, and only the resolution (or "granularity") of the range varies.

"8 bit" and "JPEG" are of course different beasts though;

- with "8 bits" you can essentially fall victim to compounding rounding errors - made worse after gamma conversion where compression & expansion of ranges result in banding and posterisation.

- with JPEG the emphasis is placed - for a given compression / quality setting - on the smallest possible file size - and to do that, some pretty serious massaging is needed including throwing away what we can't see (bar a small safety margin) - normalising what we can't tell the difference between - compressing bits that end up the same etc - and all of that after the camera and/or software has already converted to gamma 2.2 and applied a reasonable strong contrast curve. In other words JPEG is simply most suited to being a great output format.

Trying to construct an HDR composite from JPEG input files is probably a bit like landing an F18 on the deck of an aircraft carrier at night during a storm in that even if it's possible, it's probably still not going to deliver an ideal result (or perhaps trying to complete a jigsaw puzzle with 15% of the pieces missing) (or trying to bake a different tasting cake using ingredients that have already been baked into another cake) - trying to do HDR from an 8 bit file is a different story though, in which case the result is going to depend on a few factors including how many 8 bit chunks the original HDR scene was carved up into.

Sorry - we've just come off daylight saving and my brain still thinks it's 12:35am - does the above help with your question?

[Dave Humphries]

15 bit (which Photoshop uses for 16 bit mode) - 14 bit, 12 bit, 8 bit etc define the same "perfect black" to "perfect white" range, and only the resolution (or "granularity") of the range varies.

I take that as a "Yes" to it being tone mapped then

"8 bit" and "JPEG" are of course different beasts though;

- with "8 bits" you can essentially fall victim to compounding rounding errors - made worse after gamma conversion where compression & expansion of ranges result in banding and posterisation.

- with JPEG the emphasis is placed - for a given compression / quality setting - on the smallest possible file size - and to do that, some pretty serious massaging is needed including throwing away what we can't see (bar a small safety margin) - normalising what we can't tell the difference between - compressing bits that end up the same etc - and all of that after the camera and/or software has already converted to gamma 2.2 and applied a reasonable strong contrast curve. In other words JPEG is simply most suited to being a great output format.

True, that's me being sloppy

As you say, HDR with jpg would be daft, Tif, even 8-bit, would be much better. In my defence, I have never tried an HDR program, but I have done a manual (area dependent) blend of two alternate exposure processed files opened in Elements from ACR. It was a while ago though, possibly before I know as much as I do now.

Sorry - we've just come off daylight saving and my brain still thinks it's 12:35am - does the above help with your question?

Ah, ok, we did that a couple of weekends ago.

Thanks,

[BarryTX]

It's not getting "more data" per sec because you can't extract information that isn't there to start with - but - it can certainly alter the result, and quite possibly "for the better" if "better" is defined as "more pleasing to the creator".

Good point - maybe a better way for me to say what I meant is that it appears that Photomatix uses more data from 3 files, even if 2 of them were created from the third. I'm not sure this is a true statement, but it seems like it based on my results. If it is true then I suppose a better way to describe it than 'artifact' would be that the processing algorithm used for a single input file differs significantly from that used when 3 files are available. Thanks for all the comments, I find them all helpful.

[Colin Southern]

I take that as a "Yes" to it being tone mapped then

Thinking back to my assembly language programming days, I would think that all that would be required is a simple right-shift "X" number of bits.

[Colin Southern]

Good point - maybe a better way for me to say what I meant is that it appears that Photomatix uses more data from 3 files, even if 2 of them were created from the third. I'm not sure this is a true statement, but it seems like it based on my results.

Hi Barry,

"Dunno." Possibly not so much a case of "using mode" as "giving more weight to portions further from the noise floor" - just a guess though.

BTW, I'm not trying to poo-poo the concept entirely - I occasionally use it myself as it makes it easier to transition between 2 areas - rather than having to built / feather masks.