Example of H.264 direct grading break down?

[shaocaholica]

I've heard time and time again that editing/grading H.264 directly can lead to poor results compared to something like a ProRes intermediate of the same H.264 footage with all else being the same. Can someone please demonstrate this for me as I don't have all the equipment to test myself at the moment. I truely don't believe this is really the case. Once an application as decoded input footage, the limitations of the input codec are irrelevant. A 10bit 4:4:4 intermediate based on a 8bit 4:2:0 input doesn't have any more information in it. Yes, it might be a faster codec but thats not what I'm talking about. This is strictly about image quality at the end of the pipe.

If I open a poorly compressed jpeg in photoshop, change the color mode to 16bit, apply some curves and save out as a 16bit Tiff, I should get the exact same image even if I used a 16bit tiff intermediate based on the original poorly compressed jpeg.

 

[morgan_moore]

A jpg on level one will be 'blocky'

Upsize it save it a 16bit whatever you cant make it 'unblocky'

By moving to that 16bit workflow the best you can do is stop degrading it further

but you cant undgrade it - ever

the bottom line is the canon codec is weak - shoot something on tungsten - in daylight - and try to correct it - you will see the minimal lattutde for correction

shoot it well and you can get a good result of course

S

 

[ROCKMORE]

A jpg on level one will be 'blocky'

Upsize it save it a 16bit whatever you cant make it 'unblocky'


S

This is true with still shots in Photoshop. I get various quality images from clients sometimes and have to upgrade the picture from what they send me. I usually take an average to poor supplied original say under 600 x 800 pixel jpeg and scale it up to 6000 x 8000 tiff then work on that scale. There is no more original information there but a lot more refined version of "all that I got from the client". It works well in a pinch.
I think the same principle applies to upscaling the H-264 files to an uncompressed domain. In theory having 12 gigs per minute of data as a apposed to say 8 gigs to 24 minutes to work with is similar to blowing my jpeg up 10 times. Not sure but my experience has been positive. Having said that I finally got a 550D in the house today and looking forward to a closer look at the files.

 

[shaocaholica]

Upsize it save it a 16bit whatever you cant make it 'unblocky'

By moving to that 16bit workflow the best you can do is stop degrading it further

Well I'm not really talking about "upsizing". No one is going to upsize 1080p to 4K and then back down to 1080p for delivery.

Also, if you convert to 16bit, you don't have to save the file. If you open a 8bit file and change to 16bit mode, you've already converted to 16bit. Just like in After Effects. If I drop 8bit 4:2:0 footage into a 32bit comp, it converts and works with that footage in real time in 32bit. How does having an intermediate at that depth help with image quality?