CeroNoice - HDR tool inspired from ZeroNoise

[garry23]

Alex

Just to reassure you I am not one to 'just' jump in. I read and re-read this thread and others to find a solution, but, because of my poor coding skills I was flummoxed, ie I didn't understand the error messages.

I also honestly thought I had 1.4 running. I checked after reading your last post and, yes, I had 1.3 running.

1.4 running now and conversion to 32 Tiff works fine.

The Tiff initially has a strong green tint, but I can remove this in post.

Thanks for you response: I'm up and running!

Garry

[garry23]

Realtime feedback here: so don't moan at all my posts :-o).

1. lots of lag in LR (64 bit) in processing 32 bit Tif. In fact I had to click on and off images to force updates. Could 'just' be size of Tif, ie about 250Mb.
2. No problems in Ps-CC (64 bit)
3. Tif starts with green tint, easily removed via WB tool
4. Real problem is that the tif is very noisy in the shadow regions, despite one of the brackets being taken for those zones

These are early observations, but the noise still being there worries me. I seems that the noise is being carried over from some of the images that are taken for the highlights.

[a1ex]

Did you pass the bracketed shots in the correct order? (see the help)

[garry23]

Alex

The brackets are in order of increasing exposure, from lightest to darkest. I can't see any reference to order.

Here is the script:
CeroNoice.exe 1.cr2 2.cr2
dng_validate.exe -3 out out.dng
exiftool.exe -tagsfromfile 1.cr2 out.dng
pause

BTW I still get error messages about exposure time <=0 and padding.

[a1ex]

Usage: CeroNoice DARK.CR2 BRIGHTER.CR2 EVEN_BRIGHTER.CR2 ...

[garry23]

Alex

Thank you!! I simply could not find that reference.

Looks much better now!!! Will carry on experimenting.

BTW I still get exposure <=0 warning and padding messages.

Once again: thanks for your patience with me.

[engardeknave]

A1ex, will you make this work with full paths & file names for the input and output files yet? I'd like to script this to run along side my current workflow so I can compare the results. Would be tested on a ton of files.

[garry23]

I have just posted a workflow, but it is awaiting moderator action.

[garry23]

Engardeknave

BTW here is some workflow scripting I did.

I find this greatly speeds up my processing from within LR.

http://www.magiclantern.fm/forum/index.php?topic=11508.0

Cheers

Garry

[engardeknave]

Garry, my existing scripts consist of several thousand lines of vbscript that automatically process and sort hundreds of bracket sets from several shoots at once. I'm asking A1ex to make Ceronoice work with paths so I don't have to employ weird hacks or copy gigabytes of temp files during this process.

[garry23]

Got it. I thought I was in a different league to you :-o)

[garry23]

Alex (anyone else)

I have tried to use CeroNoice with .cr2 files from my S95 and G11: but no luck, even if I convert to .dng first.

I get RAW info inject failed and no camera model error.

As I've converted the 12-bit .cr2 to .dng, I thought this might be OK. But it appears it isn't.

Bottom line: I'm guessing CeroNoice doesn't run and runs with 14-bit .cr2s? Or am I missing a trick?

[a1ex]

engardeknave: will do (sorry, forgot about it).

garry23: nice workflow, would be interesting to see more samples. Maybe you can compare the results with whatever other HDR software you were using? (I don't really know them, I use enfuse for tonemapping and that's pretty much all my experience with HDR software).

I also have in mind some more updates, basically updating the exposure matching algorithm from the latest cr2hdr, Kalman weighting (to reduce noise even more), white balance matching (some samples from engardeknave require it) and soft-film curve (to squeeze the HDR data in a 16-bit DNG container). Not sure what to do about motion artifacts though.

Can you upload some raw files from S95 and G11?

[garry23]

Alex/Engardeknave

The 'issues' I have seen, having spent a weekend playing with CeroNoice, is that it only works with my 14-bit images, ie not 12-bit. BTW how do I send you bracketed images from my S95? I have a dropbox but I think I need your email address to share. I'm sorry, I'm not up to speed on dropboxes etc.

The other issue I have seen is that some images have a very fine, ie a pitch of 4 pixels, slightly dark grid imprinted on the processed image. Clearly visible.

I also always get padding warnings and exposure <= 0 warning. But as these don't appeared to interfere with the processing I ignore them!

I have also tweaked by workflow by making my .bat an .exe. This way I simply export the images from in LR and by .exe does the rest in the background, ie the user sees no command window and the CeroNoice processed image 'magically' appears in a LR watched folder. I also added in some error handling, ie checking for the right input (.cr2 or .dng) and returning an error jpeg to LR instead of a 32-bit TIFF. My .exe also returns the TIFF with the original file name plus a "_32bit" appended for identification.

I will compare images with other 32-bit merging routes asap.

Cheers

[garry23]

Alex

Try this link:

https://www.dropbox.com/sh/5i1by5uw86hz8nc/yZhcwsHQaF

[garry23]

Alex

Here is a TIFF that exhibits the matrix-lije artifact:

https://www.dropbox.com/sh/2z5k0ylrh642s0z/CVBp6urTtI

[a1ex]

Can you upload the CR2 files for the matrix artifact?

Tip: you can also use the edit button

[garry23]

Dropbox updated with .cr2s.

If it works: here is a CeroNoice processed image:



Not sure if there is a size or downsizing engine working here?

[garry23]

Image clearly didn't post. OK I'm dumb!!!

[garry23]

Anyway, here's in the processed image in Dropbox:

https://www.dropbox.com/s/7gsb9hp7p17nnh9/Backbone.jpg

[a1ex]

How are the artifacts looking like? I don't know what to look for...

[garry23]

Alex

In the tiff if you zoom in to the pixel level you can see that there is an every fourth pixel pattern in  the x and y directions.

The effect is clearer to see on some images than others, and not on every one.

[a1ex]

Found it by merging the raw files; looks like a black level issue in my algorithm (had it with dual iso). Need to update the code.

[garry23]

Glad you found it!  I was beginning to worry it was me!

BTW can you explain why I get the exposure and padding warnings?

Cheers

[a1ex]

Because I didn't fill them in the DNG