Author Topic: Dual ISO - massive dynamic range improvement (dual_iso.mo)  (Read 2121222 times)

reddeercity

  • Contributor
  • Hero Member
  • *****
  • Posts: 2303
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2750 on: March 20, 2016, 05:54:10 AM »
@dfort Here you go  , 5D2 MLV Dual ISO 100/800 1856x1044 23.976p 693MB file (about 300 frame)
https://www.dropbox.com/s/utb5b9hvfzuxr52/M19-2200.MLV?dl=0
I check on my PC , mlvfs & BarracudaGUI for cr2hdr  , everything looks correct .

Anyone else that wants to play around with the files , be my guest
I'll keep the link up for 2 Weeks  :)

dfort

  • Guest
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2751 on: March 20, 2016, 06:43:50 AM »
Ok, here's what I'm getting on the Mac.

I thought I was on to something when several of the frames in reddeercity's MLV were corrupt when run through mlv_dump followed by the testing version of cr2hdr but the same frames were not corrupt when run through MLVFS without the dual iso setting then through the testing cr2hdr. (Well not 100%--only one frame was currupted.)

In other words I thought I had it narrowed down to a problem with mlv_dump but going through the same workflow on DeafEyeJedi's sample MLV came up with the same errors with MLVFS as it did with mlv_dump.

One workflow that is working fine with these 5D2 files is using MLVS to make the DNG's and convert the Dual ISO.

I wouldn't rule out an issue with my mlv_dump and cr2hdr builds--I recently rebuilt the development environment on my MacBook Pro, but the compile seemed to complete fine. The MLVFS I'm using was also compiled on my system though with a different compiler. In addition, it looks like the Windows binaries I cross-compiled on the Mac are working fine. Go figure.

Maybe it is because I'm using gcc 5.3.0? Should I roll back to a previous version?

reddeercity

  • Contributor
  • Hero Member
  • *****
  • Posts: 2303
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2752 on: March 20, 2016, 08:05:47 AM »
Seems to be a problem also on mac  just tried to use mlv_dump and getting errors with my 5d2 dual iso.
mlvfs is find thou .

dfort

  • Guest
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2753 on: March 21, 2016, 12:17:28 PM »
@reddeercity

Where did you get the cr2hdr experimental binary? Is that the one I compiled?

reddeercity

  • Contributor
  • Hero Member
  • *****
  • Posts: 2303
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2754 on: March 21, 2016, 09:00:14 PM »
@reddeercity
Where did you get the cr2hdr experimental binary? Is that the one I compiled?
Yes Sir got for Here

Edit: just a note , the images I posted came from MLV Viewer to view .mlv (and .raw) files on windows (C#) as Dng's then process with your Win build of cr2hdr thought BarracudaGUI (Very fast)

Ottoga

  • Member
  • ***
  • Posts: 203
  • The Ox is Slow but the Earth is Patient
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2755 on: March 22, 2016, 06:21:05 AM »
@dfort @reddeercity
Quote
I have built and re-built cr2hdr from the cr2hdr branch on more than one Mac and verified that it keeps tripping up on the same frames from DeafEyeJedi's sample 5D2 files. Has anyone tried to convert these files on Windows or Linux?

DNG's -- https://mega.nz/#F!qgFDVIYY!Qe6XZc8FyVnbM6C4_OtdsA

MLV's -- https://mega.nz/#F!Wo8SDLaR!KD98tGfnzJMHlLy-cllB1A

Does anyone have another short Dual ISO test shot on a 5D2 they can post to see if this is a recurring issue or maybe just something with that one sample?

My test results processing the MLV's in a W10 environment. MLV file names were M14-1732 and M14-1733. A third MLV file from reddeercity m19-2200 was also tested.

Initial observation is that mlv file m14-1733 is not a Dual ISO mlv file. A visual inspection of the individual DNGs does not show the tell tale bright / dark horizontal lines of the alternate ISO's. Also, none of the apps used for testing recognised the DNGs as Dual ISO. If it was filmed in Dual ISO, then I'd suggest that some other setting in camera has overridden the alternate ISO value.

For files M14-1732 and m19-2200:

MLVP with Dual ISO enabled
  • Was able to export a video file. Playback showed that that the individual frames had been processed for dual ISO successfully. In M14-1732 there did appear to be a bit of fringing? around the window frame.
  • Exported cDNGs were not processed for dual ISO 

MLVFS Dual ISO enabled
  • From the virtual folders, I was able to successfully view dual ISO processed DNGs and successfully extract them to local HD directory.

MLVP with Dual ISO disabled and cr2hdr
  • Dragging and dropping dngs from the dokan virtual drive directly onto cr2hdr resulted in cr2hdr going through the motions but it was unable to overwrite the original dng with the dual ISO processed image.
  • Dragging and dropping dngs directly from the dokan virtual drive into a local directory on my HD and then processing them through cr2hdr successfully processed them as dual ISO images. Note that this resulted in the source dngs being overwritten with the dual ISO processed dngs.
  • In both of the above bullet points the warning "Exiftool - not recognised as an internal or external command, operable program or batch file", "Exiftool couldn't update dng data" was display. Exiftool was in the same directory as cr2hdr so in light of a better explanation, I assume that this is because the source file was a dng that was being overwritten by the process.

Barracuda
  • No joy here as this app only accepts .cr2 files as input

Edit:  Correction. I was able to get this app to process the dng frames after extracting the individual frames to a local HD directory and renaming the file extensions from .dng to .cr2
Edit2: Correction. If I run this app via the desktop shortcut created at installation time vs from its installation directory, then dragging and dropping dngs into the app works fine.

MLRawViewer v1.4.3
  • Allows preview of video but cannot process dual ISO.
  • Allows export to dng but cannot process dual ISO.
  • The following error was logged for reddeercity's file M19-2200.mlv

    MlRawViewer v1.4.3
    (c) Andrew Baldwin & contributors 2013-2014
    Using GLFW
    Loading 'D:\\MLVs\\M19-2200.MLV'
    Set indexed. Frames missing: 2
    M19-2200.WAV D:\MLVs\M19-2200.WAV
    FAILED TO FIND FRAME AFTER SCAN 212
    FAILED TO FIND FRAME AFTER SCAN 213

MLV Converter 1.9.2
  • Extracts the dng frames 1st, then processes the extracted dngs for dual ISO (overwriting them). A very slow process (5 seconds per frame to convert to dual ISO on my notebook).
    There is a restricted parameter set for the dual ISO conversions.
  • Can produce a proxy mov video using temporary tiff files generated from the dual ISO processed dng files. There was a noticeable colour shift in the proxy video vs the source dng images.

Hope this helps.
EOS 7D.203, EFS 55-250mm, EF 75-300 III, Tamron 16-300 DiII VC PZD Macro, SpeedLite 580EX II.

reddeercity

  • Contributor
  • Hero Member
  • *****
  • Posts: 2303
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2756 on: March 22, 2016, 07:42:46 AM »
Thanks for the test
Barracuda
  • No joy here as this app only accepts .cr2 files as input

No , It take Cdng,s & DNG's Trust me ! Just drag & drop on the batch list that's how I  process all of  my Dual ISO on Windows.

MLRawViewer v1.4.3
  • Allows preview of video but cannot process dual ISO.
  • Allows export to dng but cannot process dual ISO.
  • The following error was logged for reddeercity's file M19-2200.mlv
    MlRawViewer v1.4.3
    (c) Andrew Baldwin & contributors 2013-2014
    Using GLFW
    Loading 'D:\\MLVs\\M19-2200.MLV'
    Set indexed. Frames missing: 2
    M19-2200.WAV D:\MLVs\M19-2200.WAV
    FAILED TO FIND FRAME AFTER SCAN 212
    FAILED TO FIND FRAME AFTER SCAN 213
Funny I never got any error  , All DNG's, are accounted for no missing frames but I only ! use MLVFS On Mac & Windows
It the Only convertor that up to date . All the other convertor have not been updated.
MLRawviewer v.1.43 is obsolete now Last updated 2014/12/12 , in fact most are at least with the new dual ISO for 5D2.
I guess the only other up to date app is MLVProducer but the dev. wrote his own code for extracting dng's

So I guess I'll kept using MLVFS so not to have any problems until the other convertor catches up on updates at least with dual iso

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7405
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2757 on: March 22, 2016, 09:00:58 AM »
Regarding file M19-2200.MLV from reddeercity https://www.dropbox.com/s/utb5b9hvfzuxr52/M19-2200.MLV?dl=0

mlv_dump metadata says 214 frames but the files contains only 211. Don,t know if that has anything at all to do with the problem here.

    Size        : 0x00000034
    Ver         : v2.0
    GUID        : 2648616435936133618
    FPS         : 23.976000
    File        : 0 / 1
    Frames Video: 214
    Frames Audio: 46

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7405
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2758 on: March 22, 2016, 09:12:36 AM »
The flicker from the experimental cr2hdr is caused by a faulty black level number. I just lowered the number from Black Level 459008 to Black Level 7172 with exiftool and the file was fine again.
Here,s a problematic sample https://drive.google.com/file/d/0B4tCJMlOYfirTXVBUVkzYXJzNHc/view?usp=sharing

Coming from this file  https://www.dropbox.com/s/utb5b9hvfzuxr52/M19-2200.MLV?dl=0

jpegmasterjesse

  • Member
  • ***
  • Posts: 100
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2759 on: March 22, 2016, 09:33:47 AM »
I've had lots of success with MLP.  It works and swiftly for dual-iso on the 5d2.  I have also had some issues.  On a few sequences - even though the still-interlaced preview indicates a healthy image - conversion through MLP renders totally blank, empty black frames. 

I wasn't able to successfully export converted Dual-Iso DNGS with MLVFS.  This may have just been me not being sure how. (Not sure when/where output is modified due to in-browser toggles.)

Is there a GUI version of the experimental CR2HDR for Mac floating around anywhere?

Here is a file - Converted with MLP on Mac - to test for anyone curious what the finished output looks like.
https://drive.google.com/file/d/0B1lCd952bWxQaV91NmtKZnU2Nnc/view?usp=sharing
(To save space I only included some of the frames in the sequence. Hopefully this doesn't contribute to another issue like the one @Danne pointed out.


kgv5

  • Hero Member
  • *****
  • Posts: 643
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2760 on: March 22, 2016, 10:18:31 AM »
I tried the newest MLVProducer (PC win7 64) but couldnt export processed dualiso files as CDNGs despite preview which looked ok and processed. CDNGs still have stripes after export. Does this feature works in build 2241.intel ?
www.pilotmovies.pl   5D Mark III, 6D, 550D

Ottoga

  • Member
  • ***
  • Posts: 203
  • The Ox is Slow but the Earth is Patient
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2761 on: March 22, 2016, 02:23:50 PM »
@kgv5
Quote
I tried the newest MLVProducer (PC win7 64) but couldnt export processed dualiso files as CDNGs despite preview which looked ok and processed. CDNGs still have stripes after export. Does this feature works in build 2241.intel ?

MLVP is a work in progress. At this stage it will process dual ISO MLVs when exporting/rendering as a video (MOV format) only.

For now I suggest that you use MLVFS, cr2hdr or Barracuda_GUI.
EOS 7D.203, EFS 55-250mm, EF 75-300 III, Tamron 16-300 DiII VC PZD Macro, SpeedLite 580EX II.

Ottoga

  • Member
  • ***
  • Posts: 203
  • The Ox is Slow but the Earth is Patient
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2762 on: March 22, 2016, 02:37:46 PM »
@reddeercity
Quote
No , It take Cdng,s & DNG's Trust me ! Just drag & drop on the batch list that's how I  process all of  my Dual ISO on Windows.

Your right, tried it on another PC and it works fine. May need to reinstall Barracuda_GUI on my main notebook.

Quote
Funny I never got any error  , All DNG's, are accounted for no missing
MLRawViewer didn't display any error messages. The error identified was written into a log file that is located in:
     C:\Users\[username]\.mlrawviewer - @danne has identified a metadata mismatch which was the cause of the
                                                                      error message.
EOS 7D.203, EFS 55-250mm, EF 75-300 III, Tamron 16-300 DiII VC PZD Macro, SpeedLite 580EX II.

a1ex

  • Administrator
  • Hero Member
  • *****
  • Posts: 12564
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2763 on: March 22, 2016, 04:29:42 PM »
Here,s a problematic sample https://drive.google.com/file/d/0B4tCJMlOYfirTXVBUVkzYXJzNHc/view?usp=sharing

Do you mind extracting me just the unprocessed frame from the test video?

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7405

DeafEyeJedi

  • Hero Member
  • *****
  • Posts: 3411
  • 5D3 | M1 | 7D | 70D | SL1 | M2 | 50D
Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2765 on: March 22, 2016, 06:26:15 PM »
You guys have been magically amazing on this journey in the past several days ... Absolutely stunning, I dig the collaborations and excellent team effort from all of you!

The flicker from the experimental cr2hdr is caused by a faulty black level number. I just lowered the number from Black Level 459008 to Black Level 7172 with exiftool and the file was fine again.
Here,s a problematic sample https://drive.google.com/file/d/0B4tCJMlOYfirTXVBUVkzYXJzNHc/view?usp=sharing

Coming from this file  https://www.dropbox.com/s/utb5b9hvfzuxr52/M19-2200.MLV?dl=0

Ah, ha ... Nice work Danne and now I'm gonna try this trick of yours with a few more new Dual-ISO test files from the co-worker's 5D2 that I saved.

Has this fix be implemented into your MLP app or not yet? Manually adjusting the Blacklevels via exiftool within Terminal is the only way atm, correct?

Thanks again, D!
5D3.113 | 5D3.123 | EOSM.203 | 7D.203 | 70D.112 | 100D.101 | EOSM2.* | 50D.109

dfort

  • Guest
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2766 on: March 22, 2016, 09:44:01 PM »
@dfort @reddeercity
My test results processing the MLV's in a W10 environment.

Wow, great report.

FYI--M14-1733 was not shot in Dual ISO. It was shot so we can compare it to M14-1732 which is a Dual ISO shot on the 5D2.

I've been trying all sorts of ways to compile cr2hdr experimental on the Mac and I keep getting those pesky frames. I thought it was just a Mac issue and possibly something I did wrong when compiling but I just tried it on Windows with the same results.

Thanks Danne for pointing out the Black Level issue. Here's the output (on Windows) from that MLV reddeercity posted.

First frame 26 which always processes fine:

Code: [Select]
Input file      : M19-2200_frame_000026.dng
Camera          : Canon EOS 5D Mark II
Full size       : 1856 x 1044
Active area     : 1856 x 1044
Black borders   : N/A
Black level     : 1793
ISO pattern     : ddBB RGGB
White levels    : 14176 14817
Noise levels    : 8.00 8.00 8.00 8.00 (14-bit)
ISO difference  : 3.14 EV (882)
Black delta     : -29.52
Dynamic range   : 10.60 (+) 10.67 => 13.81 EV (in theory)

Next, frame 27 which is the one Danne pointed out:

Code: [Select]
Input file      : M19-2200_frame_000027.dng
Camera          : Canon EOS 5D Mark II
Full size       : 1856 x 1044
Active area     : 1856 x 1044
Black borders   : N/A
Black level     : 1793
ISO pattern     : BBdd RGGB
White levels    : 15917 14819
Noise levels    : 8.00 8.00 8.00 8.00 (14-bit)
ISO difference  : 3.14 EV (883)
Black delta     : -28.92
Black level     : 1793 is too high, using 114752
Dynamic range   : 10.79 (+) 10.67 => 13.81 EV (in theory)

reddeercity

  • Contributor
  • Hero Member
  • *****
  • Posts: 2303
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2767 on: March 24, 2016, 06:45:29 AM »
Doing more testing with the new cr2hdr Win7 platform, now I'm getting random frame with Wrong black level/No viewable image  on 5d2 dual ISO video , no preview.
I would say at least 30% are bad ,  with exiftool I set the black level with this command line string on  Win7 PC
Code: [Select]
exiftool -BlackLevel=7172 *.dng
damage dng exiftool metadata
Code: [Select]
Black Level                     : 459008
White Level                     : 64876
Default Crop Origin             : 0 0
Default Crop Size               : 1856 928
Active Area                     : 0 0 928 1856
Opcode List 1                   : (Binary data 28 bytes, use -b option to extrac
t)
XMP Toolkit                     : Image::ExifTool 10.02
Subject                         : Dual-ISO
Exposure Time                   : 1/53
F Number                        : 16.0
ISO                             : 200

Frame before the damage dual ISO frame
Code: [Select]
Black Level                     : 7172
White Level                     : 64876
Default Crop Origin             : 0 0
Default Crop Size               : 1856 928
Active Area                     : 0 0 928 1856
Opcode List 1                   : (Binary data 28 bytes, use -b option to extrac
t)
XMP Toolkit                     : Image::ExifTool 10.02
Subject                         : Dual-ISO
Exposure Time                   : 1/53
F Number                        : 16.0
ISO                             : 200

After repairing the black level , the image still doesn't look correct , black level is ok but looks really pixelated
I have put in my Dropbox the frame before the damage one , plus the un-process dual ISO of the damage one by cr2hdr
there are name by what did or not do
https://www.dropbox.com/s/goai3ghcr2uxki3/CR2HDR_M23-1414_000189.DNG?dl=0
https://www.dropbox.com/s/nfcc6u9efkrrbhv/CR2HDR_damage_M23-1414_000190.DNG?dl=0
https://www.dropbox.com/s/nqeahldxngfv0rn/Exiftool_7172_black_level_fixM23-1414_000190.DNG?dl=0
https://www.dropbox.com/s/kk8y7v8gz2cyf63/Original_M23-1414_000189.dng?dl=0
https://www.dropbox.com/s/wejv7rzzikfqtzv/Original_M23-1414_000190.dng?dl=0

Not sure if the Barracuda_GUI is the cause of the problem , didn't see any problems with MLVFS dual ISO but being it's
slow at conversion I which to Barracuda_GUI for more speed.

Edit: I used 200/1600 ISO , and one last thing in the Barracuda_GUI as it was processing it reported the dual ISO as 0/800
Last time I used dual ISO I tested it at 100/800 , maybe a problem ? more testing needed.

Good news , Dual ISO (5D2) works in 3Xcrop successfully record MLV raw at 2048x1024 23.976p +audio for about 20 seconds.

reddeercity

  • Contributor
  • Hero Member
  • *****
  • Posts: 2303
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2768 on: March 25, 2016, 03:09:44 AM »
Found the problem , cr2hdr doesn't like 200/1600 ISO for same reason from 5d2 re-tested with 100/800 ISO and no problem ,
 all dual ISO dng's converted with proper levels .
So is this a limitation of this experimental build of cr2hdr or a problem with the 5d2 Dual ISO module in video mode ?
Still Testing ......     

Tai.Fighter

  • New to the forum
  • *
  • Posts: 31
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2769 on: March 26, 2016, 10:36:40 PM »
[Post deleted by author]

Ottoga

  • Member
  • ***
  • Posts: 203
  • The Ox is Slow but the Earth is Patient
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2770 on: April 01, 2016, 08:30:29 AM »
@dfort   Correction @dmilligan - a new headache for you  :)

I was using reddeercity's dual_iso  mlv clip to perform some time trials of mlvfs vs other dual_iso processing utilities when I stumbled across a potential issue with mlvfs.

My hardware ASUS s550c notebook, intel i5 processors (4 cores), Intel HD graphics, 6gb RAM, OS w10_Pro

Dual_ISO mlv clip used from this link: https://www.dropbox.com/s/utb5b9hvfzuxr52/M19-2200.MLV?dl=0

The following links are for the same frame extracted from the mlv with hot/bad pixel fixing and chroma smoothing enabled only from mlvfs_x64, mlv_dump and raw2cdng respectively. Note the mlvfs images were just dragged and dropped from the virtual dokan folder to a local HD folder.

from mlvfs_x64:     https://onedrive.live.com/redir?resid=C1A12E009BA4F636!11635&authkey=!AJoTpvZKJdKgdxs&ithint=file%2cdng
from mlv_dump:     https://onedrive.live.com/redir?resid=C1A12E009BA4F636!11658&authkey=!AGfXcTSg0lAm7DU&ithint=file%2cdng
from raw2cdng:      https://onedrive.live.com/redir?resid=C1A12E009BA4F636!11654&authkey=!AP3YXZLggtyjtLQ&ithint=file%2cdng

If you compare the images you will see that in the mlvfs extracted frame, there is a band of horizontal discolouration in the lower half of the image that does not occur in the other two images.

I tried various combinations of bad pixel fix, vertical stipe fix and chroma smoothing in mlvfs to see if it would correct the issue to no avail.

Next in mlvfs, I enable the Full 20it dual_iso with Amaze interpolation, Alias map ON and Fullres blending ON. The following is a link to the processed image:
https://onedrive.live.com/redir?resid=C1A12E009BA4F636!11607&authkey=!AHl0M4oRKUS65fQ&ithint=file%2cdng

When inspecting this image you will se that the horizontal discolouration is still visible in the image. I tried again with combinations of Alias Map and Fullres Blending ON/OFF to no avail.
Lastly, I switched from Amaze interpolation to Mean32 and voilĂ ! the frame was processed cleanly as per the following example.
https://onedrive.live.com/redir?resid=C1A12E009BA4F636!11619&authkey=!AJnOkHpyiK5v_fI&ithint=file%2cdng

Is this an issue? Clearly, not a major one as I was ultimately able to successfully process the dual_iso mlv using mlvfs. However, based on my observations during this testing exercise, a couple of questions are raised my mind:

  • For MLVFS, what is different in the logic that converts the individual frames to dngs vs mlv_dump and raw2cdng?
    It seems to me that this is where the problem lies.
  • It is my understanding that Amaze is a superior interpolation algorithm/process than mean32. So why didn't it work?

Some links to additional frames extracted from the mlv clip and demonstrating the above.

MLVFS unprocessed dngs:           https://onedrive.live.com/redir?resid=C1A12E009BA4F636!11591&authkey=!AJnfDU17yD92grI&ithint=folder%2cdng
mlv_dump unprocessed dngs:     https://onedrive.live.com/redir?resid=C1A12E009BA4F636!11657&authkey=!AOuzH6UPxPb-0R8&ithint=folder%2cdng
raw2cdng unprocessed dngs:      https://onedrive.live.com/redir?resid=C1A12E009BA4F636!11592&authkey=!ABIMtCaTULWJie8&ithint=folder%2cdng
mlvfs dual_iso using Amaze:       https://onedrive.live.com/redir?resid=C1A12E009BA4F636!11589&authkey=!AEWf0uZEcrkT0P0&ithint=folder%2cdng
mlvfs dual_iso using Mean32:     https://onedrive.live.com/redir?resid=C1A12E009BA4F636!11590&authkey=!AHf8g2U5lutfPTQ&ithint=folder%2cdng
EOS 7D.203, EFS 55-250mm, EF 75-300 III, Tamron 16-300 DiII VC PZD Macro, SpeedLite 580EX II.

dfort

  • Guest
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2771 on: April 01, 2016, 10:32:15 PM »
@dfort - a new headache for you  :)

Not my headache. MLVFS is dmilligan's project and the MLV file where those files you are working with originated on a 5D2 (shot by reddeercity) using an experimental build that enables dual iso video on that camera. I believe it was shot at 200/1600 ISO which is another problem reported by reddeercity. I tried the MLV file with MLVFS on my Mac and it works fine. No strange green stripes like you are seeing. So not only don't I have a Windows system that can run MLVFS, I also don't have a 5D2 I can use to test. No headaches!

However, you bring up a good point--Amaze shouldn't be showing that strange artifact. It must be an issue with MLVFS on the PC and it should be reported on that topic.

What I do find interesting is that your mlv_dump processed DNG looks very good. Everyone else seems to be having issues--probably because they are using the mlv_dump that I built and posted on my bitbucket download area. Where did you get your mlv_dump binary?

Finally, people were reporting issues with raw2cdng combined with dual iso. Have you tried processing those raw2cdng processed dual iso DNG files in cr2hdr?

Ottoga

  • Member
  • ***
  • Posts: 203
  • The Ox is Slow but the Earth is Patient
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2772 on: April 02, 2016, 04:17:03 AM »
@dfort

 I pm'd you.

Quote
Have you tried processing those raw2cdng processed dual iso DNG files in cr2hdr?

No, but I can give it a go and report back.
EOS 7D.203, EFS 55-250mm, EF 75-300 III, Tamron 16-300 DiII VC PZD Macro, SpeedLite 580EX II.

dfort

  • Guest
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2773 on: April 02, 2016, 05:41:58 AM »
Just to make sure we're on the same page, we're using reddeercity's 5D2 dual iso MLV file for this test:

https://www.dropbox.com/s/utb5b9hvfzuxr52/M19-2200.MLV?dl=0

So far it looks like we're getting pretty much the same result with mlv_dump and cr2hdr binaries compiled on different systems and on different operating systems.

@Ottoga -- I have not encountered that artifact you pointed out with MLVFS on the Mac.

Ottoga

  • Member
  • ***
  • Posts: 203
  • The Ox is Slow but the Earth is Patient
Re: Dual ISO - massive dynamic range improvement (dual_iso.mo)
« Reply #2774 on: April 02, 2016, 07:25:54 AM »
@dfort
Quote
Just to make sure we're on the same page, we're using reddeercity's 5D2 dual iso MLV file for this test:

https://www.dropbox.com/s/utb5b9hvfzuxr52/M19-2200.MLV?dl=0
  Correct

Quote
I have not encountered that artifact you pointed out with MLVFS on the Mac.

It may be a Windows or w10 specific issue. I don't have a MAC or W7 machines so I can't do any parallel testing.

Quote
Have you tried processing those raw2cdng processed dual iso DNG files in cr2hdr?

Here are the results.

For those reading, this test result is specifically referring to dual_iso dng frames extracted from a dual_iso video.

mlvfs frame processed through cr2hdr

The same band of horizontal discolouration in the lower half of the image that is present in the original extracted frame is retained in the processed frame.
No errors logged by cr2hdr

Original unprocessed frame:  https://onedrive.live.com/redir?resid=C1A12E009BA4F636!11735&authkey=!AAdkihGIrgX66lA&ithint=file%2cdng
cr2hdr processed frame:       https://onedrive.live.com/redir?resid=C1A12E009BA4F636!11635&authkey=!AJoTpvZKJdKgdxs&ithint=file%2cDNG
cr2hdr run time text:            https://onedrive.live.com/redir?resid=C1A12E009BA4F636!11737&authkey=!AISReBXRWPvt5Gc&ithint=file%2ctxt

mlv_dump frame processed through cr2hdr
Successfully processed with no errors logged by cr2hdr.

Original unprocessed frame:  https://onedrive.live.com/redir?resid=C1A12E009BA4F636!11718&authkey=!ABhj6JPqDUwNQrI&ithint=file%2cdng
cr2hdr processed frame:       https://onedrive.live.com/redir?resid=C1A12E009BA4F636!11658&authkey=!AGfXcTSg0lAm7DU&ithint=file%2cDNG
cr2hdr run time text:            https://onedrive.live.com/redir?resid=C1A12E009BA4F636!11739&authkey=!AFJZ5yUIETuwZUA&ithint=file%2ctxt

raw2cdng_v1.6.5 and 1.7.8 frame processed through cr2hdr
cr2hdr didn't like the frames extracted by either version of raw2cdng displaying a number of 'offset too large" messages during processing. The output files generated are essentially a discoloured distortion of the source dng. They visually still look like dual-iso's but are no longer recognised as such by cr2hdr. The examples below are from v1.7.8.

Original unprocessed frame:  https://onedrive.live.com/redir?resid=C1A12E009BA4F636!11758&authkey=!ACLei2iSHagG5mQ&ithint=file%2cdng
cr2hdr processed frame:       https://onedrive.live.com/redir?resid=C1A12E009BA4F636!11695&authkey=!AIhe-SmbWq1W29Y&ithint=file%2cDNG
cr2hdr run time text:            https://onedrive.live.com/redir?resid=C1A12E009BA4F636!11738&authkey=!AEPQG2ghgiDWvqM&ithint=file%2ctxt

Post processing conclusions of dual_iso from the sample video in my w10 environment......
  • mlvfs has a potential problem with the initial frame extract that that causes corruption within the pre-processed dual_iso frame. post processing of the frame using amaze interpolation fails to correct this however mean23 interpolation did. Note that @dfort has advised that he is not seeing this issue in his Apple MAC environment
  • mlv_dump coupled with cr2hdr resulted in a clean extract and successful post processing of the dual-iso frame
  • raw2cdng coupled cr2hdr: a visual inspection of the extracted frames a successful extraction however cr2hdr encounters offset problems and fails to process the extracted frame correctly

Remember that these test results from one source video with capture parameters that may or may not have contributed to the results. If you are into dual_iso video then I urge you to perform similar tests and give feedback to the devs. Given good feedback and examples will enable them the tools available and provide a consistent post processing experience for all supported platforms.

 
EOS 7D.203, EFS 55-250mm, EF 75-300 III, Tamron 16-300 DiII VC PZD Macro, SpeedLite 580EX II.