Video Enhancer next release

[Dee Mon]

As every conveyor it's limited by speed of the slowest part. If some filter/decoder/encoder can handle 10 frames per second and other filters can do 100 FPS, overall speed will be 10 fps, and other cores will not be fully loaded. Writing to hard drive can also be such a bottleneck.

[Sunny666]

When i load the same script in MEGUI it uses all cores so its about VE.

[Dee Mon]

Is it in Preview mode? In VE preview is up to realtime, if computations take less time than 1/FPS then some time will be spent resting and again the load will be less than 100%.

[Sunny666]

It is processing mode. Got another question Usually files imported to VE are avs scripts made with MeGUI and when I choose them VE asks me:

So far I have solved this by adding function call in avs script - ConvertToRGB32() so VE understands file format. How could I use new fast YV12 scheme introduced in 1.9.4 version or am I locked to previous slow but with more quality RGB32 SR mode?
Thank you

[Manoman455]

You can try installing the yv12 codec, try this one:
http://www.free-codecs.com/YV12_QuickTime_Codec_download.htm
It should work because this is what I have installed.

[Dee Mon]

We're changing right now how VE handles YV12 video and YV12-RGB32 conversion (to get rid of that checkerboard pattern), also will try to fix this issue.

[Dee Mon]

Fixed problems with YV12 input and conversions. Now can open .avs files with YV12 output and there is no checkerboard when mixing YV12 SR and RGB32 VD filters. Just download and install again.

[Dee Mon]

Here is a detailed explanation of problems and solutions with YV12 in version 1.9.4:
http://www.infognition.com/blog/2010/yv12_conversion.html
1.9.5 is out!

[themaster1]

Interesting, hopefully this time there is no convertion YV12>RGB because the v1.9.4 was painfully slow

On a side note do you know a lossless codec that use YV12 (not YUY2), the less convertions the better

[Dee Mon]

My choice of a lossless codec is Lagarith:
http://lags.leetcode.net/codec.html

In neighbor thread Ut Video Codec was mentioned working in YV12. It's very fast but compressed files tend to be larger.

[themaster1]

Dee mon i just came accross a nice document that would be worth reading, it's about super resolution & deblocking maybe you're aware of it, it's a pdf: http://staff.utia.cas.cz/sroubekf/papers/TIP_07.pdf
I found it discussed in this doom9 forum's topic

If somehow you can implement this method that'd be great

[Maccara]

My choice of a lossless codec is Lagarith:
http://lags.leetcode.net/codec.html

In neighbor thread Ut Video Codec was mentioned working in YV12. It's very fast but compressed files tend to be larger.

I've used both. One noteworthy benefit about Ut Video Codec: it decodes fast, so nice for editing. But yes, compression seems a little worse usually (still better than huffyuv).

[Dee Mon]

themaster1,
thanks for the link to an SR article and discussion. A few thoughts:
1) The article has nothing to do with deblocking.
2) The approach is pretty standard for these kinds of articles: formulate an equation of how low-res pictures are made from hi-res real world and then search for a solution to an inverse problem using some regularization. There are hundreds of such articles published and most of them demonstrate good results on synthetic examples and even some real ones, but what is usually left unsaid is that they usually require 2-20 minutes to process one 170x90 picture. These methods are very computationally expensive.
3) Algorithm described in this paper is patented and sold to HP, as mentioned in discussion.

[themaster1]

Ah sorry, i didn't read it in depth i confess.