Menu

Show posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Show posts Menu

Messages - Yoshiyuki Blade

#1
Quote from: romainmenke on March 23, 2014, 09:19:43 PMif you look closely at the top connectors of the grip you can see a difference in sharpness

That's sample "c." You have it in reverse order for the full shots :P. The one I have the most difficulty telling apart is the building shot.
#2
Interesting, so I got 2/3 correct on the test. I'm surprised that I guessed sample "a" wrong. Even after comparing the "side-by-side" shots, my eyes are still biased toward the 1856 shot  :o.
#3
@romainmenke

I'll give it a shot: a1,b2,c2
They are indeed extremely difficult to tell apart and my choices are without much confidence. I didn't do any zooming though, but had to look closely. The false color has a much greater negative effect than any loss in resolution.
#4
On the Windows version, I just hit Ctrl-A to select all. I'm sure there's a similar command like that on Mac. Also, I hit Ctrl-V to "Paste" a profile I copied from a source.
#5
Just tested audio recording. Ran a recording till it filled a 16GB card (~4 mins). Audio seemed in sync throughout the video, but cut off in the last 10 seconds or so. It's probably because I let it run till it force stopped.
#6
I've been playing with the MLV dump compression options and I have problems when using the -b parameter. When outputting files to DNG, the images are distorted in one form or another (I can't even recognize what's in the image). Most of the time it's all white with some random bits of noise. Tried playing around with different values, but with no luck. This parameter gives just a little more compression than without it, but even that's a nice thing to have. The files were recorded on a 5D2. All the other parameters (-e -l -z) appear to work fine in the short tests I've done.

edit: spoke a little too soon. There are a few artifacts coming from somewhere else, though not as severe.
edit2: Recompressed it and it looks fine now o_O I think having the MLV Player running while its compressing isn't a good idea XD Outputting the frames crashes after a while though.
#7
@delrosestephen

Nice short! I enjoyed it and was compelled to watch it to completion. That dancing baby was creepy as hell.  ;D

On a related note, I noticed that most videos I've seen (related to RAW) have a odd levels (non-black shadows and nonwhite highlights). Is this the style people are going for these days or are they accidentally outputting to TV levels instead of PC? It lowers the contrast by quite a bit.
#8
Quote from: a1ex on July 14, 2013, 10:50:45 AM
Easy, with that theory that almost nobody believed: ISO 1600 is less noisy than ISO 100.

;)

Something along these lines: http://www.guillermoluijk.com/article/nonoise/index_en.htm

Dang, I wish I learned about this earlier  ;D I've been dabbling with enfuse by pushing an image in different directions and merging them together and thought, "Damn, it's great that I can see lots of shadow+highlight detail in one exposure now, but man it's noisy in the shadows." Now I gotta grab some new footage while it's still day outside at a high ISO and play with it later.  :o
#9
Edit: oops, double post
#10
Yeah, I prefer temporal denoising over spatial since it works kinda like multi-sampling for scanning pictures (you make more than 1 scan to cancel out the random noise that occurs). For video, previous and future frames to the same job. The only difference here is that video involves moving pictures so it's trickier. Good temporal denoisers are usually motion-compensated though. I use MVtools in Avisynth for that.
#11
@a1ex

Here are the results from a Sandisk Extreme Pro 600x 16GB:


Using a.d.'s build from commit b77a85f7ac68 (6-27-2013).
#12
Do you guys do FPS override 24.000 fps when the LV mode is set to 24 FPS mode or 30 FPS? I found that I can get a shutter angle closer to 180 degrees in 30p mode and I *think* write speeds are very slightly higher (not sure if placebo). Are there any drawbacks with doing 30 FPS mode overridden to 24.000 FPS versus overriding the default 24 FPS mode?
#13
The new hack option puts my 600x Sandisk 16GB card just into continuous levels at 1880x854! Write speed seemed to stick around 64.7 MB/s. Gonna inspect the frames to see if I get anything like ted ramasola is seeing. I'll also have to account for the fact that I left the unofficial 1880 option on as well...

edit: It was probably a bad idea to record my LCD screen for this test due to the moire lol. I don't see anything wrong, but I'll test it again later.

Whoa, I pushed the res a little higher to see where the write speeds will top off and it seems to be at around 67.5 MB/s @ 1880x940 24.000 FPS. It started off at 68.5 MB/s for a good 10 seconds or so, but dropped. We've come a long way since the early days where I couldn't get more than 53 MB/s. :D
#14
I generally convert the RGB frames to the Rec. 709 Reference Display profile (http://www.color.org/rec709.xalter), import it to RGB24 video using Avisynth and finally convert it to the YV12 (4:2:0) space with Rec. 709 coefficients. For x264 encoding, you have to set a few flags to "bt709" for proper playback.

The most accurate colors I was able to reproduce (images look almost like they do in a color managed app like photoshop) was by using the MPC-HC video player with color management enabled and gamma set to 2.4. The gamma thing may be explained here: http://en.wikipedia.org/wiki/Rec._709#Transfer_characteristics

BTW, this was all guesswork and I have no clue if it's the right way to do things.
#15
@ted ramasola & duncanidaho25: Nice, I got a shot of the supermoon a couple days ago too! I just have a cheapo Canon 300mm lens though. Lots of atmospheric distortions were visible. I compared it to a shot I took a couple years ago of a regular full moon and it was indeed a little bigger!

edit: Btw, the moon was also a yellow/light orange in appearance at the time of the shot and it turns out that way in post when I use the same setup I have for regular daylight. What do you guys think about white balancing it to make it gray?
#16
I just noticed that reducing the frame rate with FPS override also increases the card's write speed slightly. So not only does it reduce the inherent write speed requirements versus the standard 24/23.976, it also increases card performance, which may make the difference between being able to record at a certain resolution or not. Anyone been naughty enough to record at lower rates and interpolate the frames in post?  ;D

At 24 FPS, my 600x Sandisk card normally writes at around 63-64 MB/s out of the required ~80, so the buffer fills up quickly. Lowering it to 19 FPS reduces the speed requirement to 63 MB/s and also increases my write speeds to 66 MB/s. 20 FPS is just beyond the point of intersection for continuous shooting though; write requirement jumps to 66 MB/s and write speed drops to about 65 MB/s . Though I don't think we'll get much beyond 21-22 FPS with a faster card, it's nice to see that one can squeeze out a little more performance than expected from the raw resolution/fps calculations.
#17
Quote from: cheguevara on June 11, 2013, 11:42:56 AM
i got it working somehow :)))
but i find few quircks... average speed of h.264 is 50-54mbs and raw doesnt fo furhter than 35mbs  ???
how to record 1280x720 without crop 
Don't forget that RAW unit of measurement for bitrate is in MB/s (megabytes per second) while h.264 is in Mb/s (megabits per second). 1 byte = 8 bits.

50-54 Mb/s video is only about 6.5 MB/s. RAW video pushes the bitrates on the 5D2 up to 60-70 MB/s!
#18
Quote from: Audionut on June 07, 2013, 05:40:37 PM
I had good results with this back in the day.

If you're handy with avisynth, you can try changing the resizing algorithms in the script for better results.

Apply a good sharpener at the end of your script to break the laws of physics :P

Be prepared to significantly increase your processing time.  But hey, quality comes at a cost ;)

I think I fooled around with some AVS AA filters for this kind of video, but I don't recall getting anything satisfactory at the end. I probably didn't use them correctly though. I wonder if an implementation of Nvidia's FXAA could do anything about it. It's a post-AA so it doesn't require real-time rendering to work. I searched for info about it a while ago on Doom9's forum with not much luck.
#20
Hehe, perhaps there's no silver bullet. :o Seems like quite a predicament here!
#21
Both look much better. I'll give the edge to raw2dng_w1.exe.

Processed JPEGs:
https://dl.dropboxusercontent.com/u/63799907/000350_raw2dng_w1.jpg
https://dl.dropboxusercontent.com/u/63799907/000350_raw2dng_wlog.jpg

DNGs:
https://dl.dropboxusercontent.com/u/63799907/000350_raw2dng_w1.dng
https://dl.dropboxusercontent.com/u/63799907/000350_raw2dng_wlog.dng

I can just barely notice some vertical stripes on the blurry leaves to the left of the rose with the wlog version.
#22
Hmm, not here. I'm noticing stripes in both Raw Therapee and Photoshop, particularly noticeable by the wooden fence in the back.
#23
Quote from: a1ex on June 07, 2013, 03:44:01 PM
Yep, it's the same as mine. The conversion log is whatever it prints on the terminal.

Can you upload the source raw? (or better, the first few MB of it)

Ah, I'm not sure how to copy text from the Windows command prompt so I can't provide that info atm. I'm guessing the relevant info is the vertical stripes correction? It reads like so:
1.000 1.000 1.028 1.025 1.064 1.059 1.028 1.027

Never tried splitting before so I don't know if HJsplit is appropriate. Here's the first 30 MB it produced: https://dl.dropboxusercontent.com/u/63799907/M06-1624-merge.RAW.001

The original is like 14.5 GB  :o so at that point, I might as well have made quick random recording instead.
#24
MD5: eec7a999eece55305222ae455003445c

I don't know where the conversion log is. If it's at the end of the conversion... well I'll have to wait for ~6k frames to extract :P. I'll update in a couple mins.
#25
Ok. Well, here's the DNG from the raw2dng.exe you posted ~30 mins ago: https://dl.dropboxusercontent.com/u/63799907/000350-4.dng