After seeing this difference myself, I can see a difference between Haali's YUY2 and RGB32. The difference is extremely subtle however. I wonder if it is possible to write a shader for mpc-hc that corrects this. I know of a thread on doom9 where a similar issue (only yv12 to YUY2) was discussed and a shader was written to correct it (thread link).
-
AnyStream is having some DRM issues currently, Netflix is not available in HD for the time being.
Situations like this will always happen with AnyStream: streaming providers are continuously improving their countermeasures while we try to catch up, it's an ongoing cat-and-mouse game. Please be patient and don't flood our support or forum with requests, we are working on it 24/7 to get it resolved. Thank you.
What Renderer to use with Reclock?
- Thread starter jmone
- Start date
I found a shader that might correct this subtle leeperry mentioned. Check this thread out for the code.
yeah, some poor rounding.
well, I'm colorblind so we don't see colors the same way
to me, one looks OK, the other one looks B&W'ish and undersaturated
besides it ruins the whole idea of doing gamut conversion afterwards.
PS scripts don't work with Haali (yet).
just like the difference between RGB32HQ in ffdshow, and ConvertToRGB32() in the Avisynth filter of ffdshow.
the later creates 10% larger PNG files, it's a good guess it makes better roundings :
ffd :
AVS :
Last edited:
I can't see any difference between those images last two images.
Your colourblind comment is interesting.
I sit next to a guy at work who can get green and white mixed up (a green traffic light can look almost white to him), but blue is very prominent to him.
Your colourblind comment is interesting.
I sit next to a guy at work who can get green and white mixed up (a green traffic light can look almost white to him), but blue is very prominent to him.
Last edited:
hehe, I don't have bionic eyes
I also don't see a diff between converttorgb32/rgb32hq above.....yet, the PNG size is always 10% bigger with convert which leads me to believe it's more accurate.
there's many levels of colorblindness, some even see almost in B&W.
the army uses colorblind ppl to see camouflages regular ppl can't see.
so we see colors, just differently
personally I can't tell the diff between light yellow/light green, violet/navy blue, dark brown/dark green and such.
but getting older, my brain has learned to improve what my eyes can't see
I watch my movies on D65/gamut converted displays, and I can tell you they look WAY better this way :
I also don't see a diff between converttorgb32/rgb32hq above.....yet, the PNG size is always 10% bigger with convert which leads me to believe it's more accurate.
there's many levels of colorblindness, some even see almost in B&W.
the army uses colorblind ppl to see camouflages regular ppl can't see.
so we see colors, just differently
personally I can't tell the diff between light yellow/light green, violet/navy blue, dark brown/dark green and such.
but getting older, my brain has learned to improve what my eyes can't see
I watch my movies on D65/gamut converted displays, and I can tell you they look WAY better this way :
Doh! I forgot about that.
Well, there is probably a difference between ffdshow rgb and avisynth rgb because ffdshow contains lots of SIMD enhanced code while avisynth is mostly pure c code. SIMD extensions (like sse/mmx/3dnow...) sometimes produce inaccurate/unstable results ([1]).
I watch my movies on D65/gamut converted displays, and I can tell you they look WAY better this way :
I see what you have done - you have cropped your gamut inside the SMPTE-C (REC709?) boundary?
How do you overlay the SMPTE-C gamut on your measure gamut?
I've got HFCR open, but can't figure that bit out..
here's in what order I did this :
-calibrate to D65 with the built-in test patterns within Color.HCFR
-input the xyz coordinates in this PS script :
http://www.avsforum.com/avs-vb/showthread.php?t=912720
-measure the end result with the official test DVD
-open the 2 .chc files, and check "use as reference" in the original file
the PS script is using a 3D lut to make your gamut fit within any reference gamut you'd like(SMPTE-C/EBU/HDTV)
in the same order, they are called REC601/PAL/REC709 in Color.HCFR
I must be missing soemthing with this whole Haali thing.
I thought I would have another play with Haali, to see how my jitter varied over a prolonged period. I put on a 720p HD mkv TV show and I had to stop after a few minutes. The picture quality upscaling to 1080p was simply horrible. Ghastly aliasing visible clearly on text but also on the picture too.
Below is a comparison. It's just very poor.
What am I doing wrong?! OK I could use ffdshow as I do for SD, but it really should not be needed for 720p.
I thought I would have another play with Haali, to see how my jitter varied over a prolonged period. I put on a 720p HD mkv TV show and I had to stop after a few minutes. The picture quality upscaling to 1080p was simply horrible. Ghastly aliasing visible clearly on text but also on the picture too.
Below is a comparison. It's just very poor.
What am I doing wrong?! OK I could use ffdshow as I do for SD, but it really should not be needed for 720p.
Attachments
That looks like a deinterlacing issue.
It's doing BOB, instead of something more sophisticated.
It's probably your video decoder is deinterlacing in hardware when you use VMR9, but not with Haali.
1. Don't watch 720p stuff
2. Don't watch video material, just film
It's doing BOB, instead of something more sophisticated.
It's probably your video decoder is deinterlacing in hardware when you use VMR9, but not with Haali.
1. Don't watch 720p stuff
2. Don't watch video material, just film
I guess it's time to start reading on Doom9.
I either resize DVD (very rare) to 1920x1080 with ffdshow, or watch 1920x1080 HD, so I've never noticed.
Haali is the only Renderer which allows me to retain video levels, rather than expanding to PC levels. VMR9 expands for HD with an ATi card.
I either resize DVD (very rare) to 1920x1080 with ffdshow, or watch 1920x1080 HD, so I've never noticed.
Haali is the only Renderer which allows me to retain video levels, rather than expanding to PC levels. VMR9 expands for HD with an ATi card.
it's a bug in the ATI drivers, same occurs with the KMPlayer VMR9/EVR scaler.
they goofed up in the PS coeffs, and don't care about fixing them.
anyway I prefer to upscale with Spline in ffdshow, than with bicubic.
downscaling is done through the Aniso. Filtering.
when you press ALT+O, you get HR's OSD, then you get 2 letters......one is the horizontal resizing, the other vertical
N = no resizing
A = aniso. filtering(downscale)
S = buggy upscaling on the ATI
as long as you don't get any S, you're good to go
Haali wrote major parts of VMR9, he knows what he's doing
Last edited:
Well it does not occur with VMR9 in either hardware or software mode with TheaterTek or MPC-HC. So to me that says Haali is just driving the driver wrong. You sure do love Haali. He can do no wrong in your eyes!
Can't speak about KMPlayer.