MLV App - All in one MLV Raw Video Processing App [Windows, Mac and Linux]

[FBongcam]

Hmm... and where are those artifacts in the clip posted above? Cleanest image ever, IMHO

Do you not get these results in your preview? I get the same result when exported as well.

[masc]

Do you not get these results in your preview? I get the same result when exported as well.

No, no such artifacts also for me. What version of MLVApp did you use, and on which OS?


EDIT: STOP! I get the same result you have using v0.16. This version did not completely auto correct wrong black level of your file. The blacklevel of your file is over 5000... but it should be 2048. After correcting black level all is perfect (latest repos revision does it automatically).

[FBongcam]

No, no such artifacts also for me. What version of MLVApp did you use, and on which OS?

EDIT: STOP! I get the same result you have using v0.16. This version did not completely auto correct wrong black level of your file. The blacklevel of your file is over 5000... but it should be 2048. After correcting black level all is perfect (latest repos revision does it automatically).

Yes I'm using v0.16 on mac os. Looking forward to next release then! 

[Danne]

Why wait. Double click this and build your own updated version:
https://bitbucket.org/Dannephoto/mlv_app_compiler/downloads/mlv_app_compiler.dmg

https://bitbucket.org/Dannephoto/mlv_app_compiler/src

[masc]

Why wait. Double click this and build your own updated version:
https://bitbucket.org/Dannephoto/mlv_app_compiler/downloads/mlv_app_compiler.dmg

https://bitbucket.org/Dannephoto/mlv_app_compiler/src

Correct! It is so easy! @Danne: I linked it now in our github readme

[FBongcam]

Why wait. Double click this and build your own updated version:
https://bitbucket.org/Dannephoto/mlv_app_compiler/downloads/mlv_app_compiler.dmg

https://bitbucket.org/Dannephoto/mlv_app_compiler/src

Correct! It is so easy! @Danne: I linked it now in our github readme

Thanks guys! 

[Danne]

Correct! It is so easy! @Danne: I linked it now in our github readme

Awesome. Thank you

[dfort]

@Danne -- You asked me to do a test comparing the antialiasing feature in MLV App against a Mosaic Engineering VAF filter on my 700D. Here is my first attempt shooting one of our favorite subjects, our computer's keyboard. This is shot at 1736x976 with 3x3 sampling. That setting tends to show aliasing.

Processed through MLVFS and the DNGs converted to TIFFs using dcraw. I like doing it this way because it doesn't smooth out the problems like Adobe Camera Raw does.

Not much difference if you look at the whole shot on a computer screen.

No VAF


With VAF


If you look closely you can see a difference. Both of these were magnified 1600% in Photoshop.

No VAF


With VAF


I tried to pick frames that showed the D in focus. The VAF will soften the image slightly, though it is hardly noticeable if you look at the full frame. What does show up is that the edges of the shot without the VAF are some sawtooth looking edges and colored artifacts that are sometimes mistaken for focus pixels but they only appear on high contrast edges.

Now let's use these export settings so we can compare to the TIFF created in dcraw.



No VAF processed through MLV App with the above settings.


Once again the original for comparison:


Not exactly the best test--I should probably go outside and shoot a bicycle wheel. However, it does show marked improvement. Most import, I believe, is that the colored specks on the edges are gone.

[Danne]

Nice one! Did you try the "last resort" setting as well?
If you could run it through the tiff_cleaning.command script too it would be nice to see what that looks like as well.
Thanks for sharing.

[dfort]

Ok--here's what the "last resort" setting looks like.



I also selected the "Post Export: TIF_CLEAN.command" so it continued processing after MLV App said it was done and when the post export was done it deleted the TIFF image files and created a Prores 422 movie file. That's a little strange. Here's what the movie look like--enlarged 1600%.



It does look like this is a good option when you don't have VAF. I need to find something like your bicycle wheel that shows some serious aliasing issues and do further testing.

[Danne]

It does look very good.
Yes, tiff only acts as intermediate in the tif cleaning script.
Thanks again for posting results.

Edit: the better option seems the ffmpeg one since it won't affect moving subjects the way tif_cleaning.command is(thus not an issue in this test). Also faster converting.

Edit 2: also note sharpening effects on the first and 2nd ffmpeg tests which are not included in the one with vaf filter

[bouncyball]

I get the same result you have using v0.16. This version did not completely auto correct wrong black level of your file. The blacklevel of your file is over 5000... but it should be 2048. After correcting black level all is perfect (latest repos revision does it automatically).

Yes right! Black Level = 5111

@FBongcam: That MLV is cut and exported by "MLV App version 0.15 alpha". Can you share the original .MLV? (or .RAW)

I can't believe that "raw2dng" put 5111 black level into the header for 5D3. Not possible! (black = 2048, white = 15000, 5D3 defaults)

Edit: well, it's possible if original .RAW had this wrong black in the footer (I would look at that .RAW)

[FBongcam]

Yes right! Black Level = 5111

@FBongcam: That MLV is cut and exported by "MLV App version 0.15 alpha". Can you share the original .MLV? (or .RAW)

I can't believe that "raw2dng" put 5111 black level into the header for 5D3. Not possible! (black = 2048, white = 15000, 5D3 defaults)

Edit: well, it's possible if original .RAW had this wrong black in the footer (I would look at that .RAW)

Yes this is from some of the first builds when raw was introduced. It worked out fine with the newer version of MLVApp though, that I compiled.

Edit: Would you like to have the original .raw file anyway?

[50mm1200s]

@dford would be nice a shot from a concrete wall,  with lots of detail

A question for the people working on this: since the algorithm seems to work by blending, wouldn't be better to denoise it first? I see some chroma noise on dford's shots that might affect the precision of the final blend...

[dfort]

@50mm1200s - My name is Fort as in strong in French not Ford like the car. Been dealing with that all my life. 

Lots of concrete where I live here in the Los Angeles area but a better subject would probably be bricks like this example a1ex uses:



Not many buildings like that in this concrete jungle.

Funny that it is a challenge to get aliasing artifacts to show up. I know that it is more pronounced on mv720 (5x3 sampling) than mv1080 (3x3 sampling) and pretty much absent on zoom mode and the other high resolution formats from the crop_rec module (1x1 sampling). Maybe shoot a worse case scenario in mv720 mode?

This is the first time I tried the VAF on the 700D. It does seem to do a very good job but it is rather expensive and not available for all ML supported cameras -- the EOSM for example. Doing it in software is a very attractive option even if it takes some extra post processing.

The two biggest issues when shooting video with DSLRs are aliasing artifacts and rolling shutter. Here's a challenge for this app, something to reduce the "Jello" effect. That should be doable, though trying to fix propellers, that's a bit more challenging.



Click on the image to read why this happens.

[mesebar2]

Dfort, I live in Venice, Italy, and we have nothing but canals (highlights) and bricks (aliasing).
I can confirm that on my EOSM mv 5x3 is almost unusable here, while crop_rec produces excellent results!

Please, let me know if you need any footage
You guys rock, and so does the work you do!

[50mm1200s]

@50mm1200s - My name is Fort as in strong in French not Ford like the car. Been dealing with that all my life. 

Oh, sorry about that.

The two biggest issues when shooting video with DSLRs are aliasing artifacts and rolling shutter. Here's a challenge for this app, something to reduce the "Jello" effect. That should be doable, though trying to fix propellers, that's a bit more challenging.

I think the VAF does a decent job on aliasing. About the shutter, this solution is neat:

[bouncyball]

Edit: Would you like to have the original .raw file anyway?

Yeah would be nice to take a quick look at it.

[dfort]

@50mm1200s - Well a global shutter like described in the video would be the best solution but there are software solutions that can clean up some of the minimal Jello effects of the rolling shutter we've got to deal with.

@mesebar2 - Upload a sample of an "unusable" shot of Venice and let's see what we can do with it. Preferably a very short (or trimmed) MLV.

[FBongcam]

Yeah would be nice to take a quick look at it.

Original .RAW file: https://www.dropbox.com/s/lbolcjtyexvvhg2/M01-2016.RAW?dl=0

[mesebar2]

@mesebar2 - Upload a sample of an "unusable" shot of Venice and let's see what we can do with it. Preferably a very short (or trimmed) MLV.

Will do. I think I have some shots I took around Feb/Mar, before I opted to use only crop_rec with the EOSM. Are those okay or have there been changes in the 3x5? In that case, I could shoot new ones later/tomorrow 

[dfort]

@mesebar2 - Whatever you have should be fine. We're looking at fixing aliasing so it doesn't matter what version of ML you used as long as it works with MLV App.

[bouncyball]

@FBongcam: Thank you, appreciate this.

Edit: it indeed has wrong black level in the footer

[mesebar2]

Link to two examples: https://we.tl/QRiXYw8l8z
I realised I shot these back in October, time flies! I had just purchased this second-hand EosM and was exploring its potential. I believe I was recording in 3x5, but I would have to ask you guys to make sure I didn't just screw up!

Hope it helps

[masc]

Link to two examples: https://we.tl/QRiXYw8l8z
I realised I shot these back in October, time flies! I had just purchased this second-hand EosM and was exploring its potential. I believe I was recording in 3x5, but I would have to ask you guys to make sure I didn't just screw up!

Hope it helps

Whohooo... thanks, that is indeed extreme! And here we also see the biggest enemy of my ffmpeg based idea: a tripod. Without movement in the picture that does not work very good.