Author Topic: Danne's crop_rec_4k experiments for EOS M  (Read 183931 times)

norek

  • New to the forum
  • *
  • Posts: 8
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #425 on: February 15, 2019, 06:18:44 PM »
Danne!

No words... incredible...  :o testing all weekend...

henricusmaria

  • Freshman
  • **
  • Posts: 70
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #426 on: February 15, 2019, 06:45:24 PM »
Nice!

Does this also mean less aliasing and moiré?

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7293
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #427 on: February 15, 2019, 07:00:59 PM »
Please do tests, compare and post your results.

kayman1021

  • New to the forum
  • *
  • Posts: 21
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #428 on: February 15, 2019, 08:28:29 PM »
Apart from the center focus pixels in red and blue, i have noticed some repetetive pattern in both greens as well. Under the cat's head. Very faintly i see it in red and blue as well.


EOS 100D, Xiaomi A1

masc

  • Contributor
  • Hero Member
  • *****
  • Posts: 1979
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #429 on: February 15, 2019, 09:16:31 PM »
Hi Danne, something is strange here when trying your todays build: using mv1080p MCM rewire preset I get a resolution of 1672x564 for 16:9... is that expected? Do I have to stretch it in height like the mv720p files? And I only get 29.973fps, not 23.x.

Edit... no... nothing to stretch. it is 1672x564 and needs no stretch. But ratio in liveview is not correct here too.
5D3.113 | EOSM.202

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7293
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #430 on: February 15, 2019, 09:26:29 PM »
You should get 1736x976 mv1080p. Try canon menu button a few times if nothing works. Also set resolution to max under Raw video menu.

masc

  • Contributor
  • Hero Member
  • *****
  • Posts: 1979
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #431 on: February 15, 2019, 09:36:45 PM »
Hm... max resolution is set to 5xxx pixels, pressed Menu button so often now, but nothing changes. I get only this:


Edit... AAHHH! Movie Crop Mode has to be switched ON? (see it in your screenshot now)
5D3.113 | EOSM.202

masc

  • Contributor
  • Hero Member
  • *****
  • Posts: 1979
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #432 on: February 15, 2019, 09:47:34 PM »
Japp... that's it. Forgot to switch on Crop Mode! Looks fantastic now! ;) Great work Danne!
5D3.113 | EOSM.202

dfort

  • Developer
  • Hero Member
  • *****
  • Posts: 3751
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #433 on: February 16, 2019, 06:14:14 AM »
Made a pass at creating a focus pixel map file for this but it looks like those subtle focus pixels are stubborn.
5D3.* 7D.206 700D.115 EOSM.203 EOSM2.103 M50.102

masc

  • Contributor
  • Hero Member
  • *****
  • Posts: 1979
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #434 on: February 16, 2019, 12:06:50 PM »
The sun is shining, so I was outside with the new reworked mv1080p mode. Handling is fantastic - I can focus fine now! Yay. In the green grass I saw that the bottom line is red. Maybe you know which register has to be get a little hit to make it the right color :)
5D3.113 | EOSM.202

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7293
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #435 on: February 16, 2019, 12:22:20 PM »
New build:
https://www.magiclantern.fm/forum/index.php?topic=9741.msg208959#msg208959

Great @masc. Thanks for pixelpeeping. Is version above without the bottom red line?

masc

  • Contributor
  • Hero Member
  • *****
  • Posts: 1979
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #436 on: February 16, 2019, 01:09:12 PM »
Thanks Danne. The red line is away. But did you change more? In my test clip I got so much moiree and "stair effects" at diagonal lines... way more than with yesterdays build (but don't have it anymore to compare). MLVApp 100% zoom ( f/2.8 ):



Edit: compared to the "old" mv1080p mode, no stairs to see there using same settings. So something is strange with new version.
5D3.113 | EOSM.202

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7293
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #437 on: February 16, 2019, 02:22:32 PM »
Funky. Deleted the build. Will check into it this evening.

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7293
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #438 on: February 16, 2019, 02:56:17 PM »
By the way. Could you test and compare like this:
In Crop mode sub menu do:
Skip_top 2
Skip_bottom -2
This is the old settings.

masc

  • Contributor
  • Hero Member
  • *****
  • Posts: 1979
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #439 on: February 16, 2019, 04:03:29 PM »
Yes, it is better with the change of these two variables. Extreme moiree is gone again. But I don't see the red line... find no green to film for test here.

Edit: yes... red line is back again, moiree is gone with +2, -2. It is not just moiree, there is much more noise and very red focus pixels.
5D3.113 | EOSM.202

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7293
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #440 on: February 16, 2019, 04:15:58 PM »
You could test maybe +1 and -1 too? Will check over here as well.

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7293
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #441 on: February 16, 2019, 04:39:47 PM »
New build:
https://www.magiclantern.fm/forum/index.php?topic=9741.msg208959#msg208959

@masc, please test.
Confirmed getting sharper focus pixels and more moiré before. Why, I have no idea.

masc

  • Contributor
  • Hero Member
  • *****
  • Posts: 1979
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #442 on: February 16, 2019, 04:43:38 PM »
What's different in this build?
I tested:
+0-0: moiree and no line
+1-1: less moiree and line
+2-2: have to search moiree and line
+3-3: extreme moiree and noise and 2 lines

Maybe -1+1 is better? :D Edit: no.  :P
5D3.113 | EOSM.202

dfort

  • Developer
  • Hero Member
  • *****
  • Posts: 3751
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #443 on: February 16, 2019, 04:53:30 PM »
There are some cases where the edges might show a red line or a green line or just a change in density depending on what app you use (assuming MLV App in your case).

Here's an extreme example that ACR shows but dcraw doesn't:



If you pixel peep the bottom of Danne's cat picture there's also something going on:



Not sure how debayering works but depending on the algorithm it seems that things can get messed up at the very edges. I pointed this out to Dave Coffin, dcraw developer, but never got a reply from him or a clue how this could be fixed. My hunch is that solving this would also help when removing focus pixels from dual iso footage.
5D3.* 7D.206 700D.115 EOSM.203 EOSM2.103 M50.102

masc

  • Contributor
  • Hero Member
  • *****
  • Posts: 1979
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #444 on: February 16, 2019, 04:59:23 PM »
Tested new build... no red line, but very colorful moiree (I think more then in yesterdays build)

Picture @ f2.8 (full open)
5D3.113 | EOSM.202

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7293
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #445 on: February 16, 2019, 05:02:01 PM »
I have the issue right now I can´t reproduce. @masc. What are your settings? fps, iso.

masc

  • Contributor
  • Hero Member
  • *****
  • Posts: 1979
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #446 on: February 16, 2019, 05:07:33 PM »

1/97... ups... hahaha
5D3.113 | EOSM.202

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7293
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #447 on: February 16, 2019, 05:18:23 PM »
From the latest upload I tested a static scene and ran:
0, 0
+1, -1
+2, -2
+3, -3
+4, -4
then:
-1, +1
+2, -2

All files were about the same regarding aliasing and focus pixels. Hmm.

Could you do a static scene with exact same settings and produce two examples which clearly shows the difference imagewise?

masc

  • Contributor
  • Hero Member
  • *****
  • Posts: 1979
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #448 on: February 16, 2019, 05:32:09 PM »
Have to do it tomorrow. It is getting dark and the best view to reproduce was outside with fences and trees. Inside I don't get moiree.
5D3.113 | EOSM.202

masc

  • Contributor
  • Hero Member
  • *****
  • Posts: 1979
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #449 on: February 16, 2019, 05:52:00 PM »
All I see with indoor tests is:
+5-5; +6-6: white line and a red line at the bottom
+3-3; +4-4: red line at the bottom
+1-1; +2-2; 0 0; -1+1; -2+2: no line at the bottom
5D3.113 | EOSM.202