OK - now I understand the issue with the compression
but what do I do with the fact that the scanned 600 dpi image looks way better than scanned 300 dpi counterpart. I want to start with the highest quality possible -- is downrezing in - say Photoshop--- the only or even the best course of action for maintaining the iamge quality that I see in the original 600 dpi scan?

thanks fultro

I'd disagree with that based on my experience but the gain is fairly small.
I've worked with 100s of still on the T/L at around 3000x2000 res and even on my dual Xeon system it's very slow, probably need more RAM but that's pretty expensive at the moment.

Be aware also that HiRes stills and video can be a problematic experience due to aliasing. I get around this by running the project at 1080, adding 0.001 vertical guassion blur and encode from that to mpeg-2 SD. Results look pretty stunning.

Bob.

Thanks Bob - a lot of food for thought here...
" I'd disagree with that based on my experience but the gain is fairly small." - disagree with what?

"I get around this by running the project at 1080" - so you set your project properties to 1080 and render SD (720-480 in my case) ? if that is right - could you explain how this helps and doesn't this REALLY slow things down? Also - would there be much to be gained from this in the case of a scan of a half tone print of an etching? -- it sounds like you are working with high quality photos

fultro

There is NOTHING to be gained using ulta high resolution images IF your finished project is of a lessor resolution EXPECT if you plan on lots of panning or zooming in which case it helps. It boils down to simple math. If your making a DVD in NTSC land your project is 720x480. That's 345,600 pixels. Resolution is a odd word which can mean different things when talking about graphics. When you refer to DPI (dots per inch) that's really a reference to dot density in a image printed to paper. Taking the same DVD and playing on a TY how sharp the image will remain depends to some extent on the size of the TV screen. Since a larger screen has more pixels, your DVD player will need to blow it up more. Viewed on a smaller screen the image may look brighter and sharper since less stretching is taking place.

I woul try setting the file res (in pixels) to be the same as (or, allowing for the PAR factor, compatible with) the project res. This should work well as long as you don't need to zoom in on the stills. If you make the resolution higher`than the project settings, you might see moire effects and twinklies because you are effectively viewing a high res image through a lower res display device.

Richard Hunter

Sorry,
to be more specific. Assuming you're starting from a hires still, in my case 3,000x2,000 you could do one of two things, downscale the res to native DV in say PS and put that on the T/L OR put the still at native res on the T/L.
Despite what conventional wisdom would suggest putting the still at native res on the T/L looks a bit better but then you run into potentially serious problems with artifacts depending on the content of the still. Even with the res downscale in PS you can still hit this porblem.

So, the best way I could find to get rid of the artifacts (things like having part of the frame blink every few seconds) was to apply the GB at 0.001in the vertical direction. Except doing that in a SD project introduces more GB than necessary. So that's why I used a HD project setting, Vegas scales the GB based on the project size. So by dropping my stills at native res into a HD project and apply the GB in that and then encoding direct from the T/L I got a very stunning result. Also bear in mind the stills were being cropped to full frame 16:9 so actually the image size should have been 1049x576 (for PAL).

These weren't my stills BTW, taken for my client by someone else. The really good ones on DVD look as good as anything can on a DVD, I doubt you'd get better from HDCAM. I've also worked stills from a large format digital camera, around 13 MB per image, Vegas sure choked on them (they were TIFFs, so I converted to png but kept the res) results were just amazing, they were of hair models and you could count the hairs.

I'm not so convinced about this resolution thing. When PS downscales it doesn't just throw data away and it works in CIE, it uses some rather complex maths to determine the best possible value for each pixel by sampling more than one pixel from the image that it's starting from. If it didn't the results would look truly horrid, all sorts of aliasing would happen.
Now Vegas presents the best possible frame it can to the mpeg encoder and although that uses 4:2:0 sampling they use a different matrix to DV.
So when Vegas downscale from CIE to SD for the encoder I suspect by cutting out an intermediate step a slightly better image is the result.
This is NOT the same as saying you can replace something that has been lost! This is about using ALL the available data that was captured to create the best possible lower definition image, color images are represented in a 3D space and the maths eludes me totally so I have to rely on experimental means. Please don't take my word for any of this, try it, if you think I'm nuts it will not have cost you anything although you might need to have something like a component feed from your DVD player to your TV to see the difference.

There's another angle to why I persisted with the longer render time and a sluggish Vegas. I can easily go back to this project and render to WMV 720p for the client as a demo reeel.

Bob.

Hi Bob - bear with me as I am close to understanding you here and this is just the discussion I have needed .
you say "........you could do one of two things, downscale the res to native DV in say PS and put that on the T/L OR put the still at native res on the T/L." and "So by dropping my stills at native res into a HD projec... "
------ By native - do you mean - the project res?
"So when Vegas downscale from CIE to SD for the encoder ...." - what does CIE refer to?

And, just to be clear, you are saying that this process you describe results in a better image when finally rendered at SD for a standard TV set (with the possible need for component input to it) ?

Many thanks fultro

There are 2 main issues to watch out for. The first is moire fringing when your still is higher resolution than the display device. Preparing the stills in advance to project resolution is the best way to avoid this, BUT, if you are using an interlaced display, you will then face the second issue which is thin lines twinkling because they only appear on alternate fields.

If you are using stills at project resolution, and view on a progressive display, you should be able to get very good results, without artifacts.

If the display is interlaced, then you need to find a way to reduce the amount of vertical detail in the final output, such as what Bob describes.

Richard

------ By native - do you mean - the project res?
No I mean leave them at the original res.
And yes aliasing is a huge problem, way beyond my typing skills and knowledge to explain all that there, suffice to say the article I read last night says there's at least 6 different kind of aliasing issues involved in downsampling video.

The article did bear out what I've seen, this article was about film DIs so bear with me on the terminology:

"2K is 2K is 2K, right". Well not exactly, what ends up in that 2K DI depends on many things, by scanning at 6K res and downsampling to 2K res they can get a better result than a native 2K scan.

CIE, sorry forgotten what it stand for, but it's a color space that can represent all of the possible colors. This is a much bigger space than video uses.

Bob,