DVD Upconversion...

[Peter]

All those algorithms may be ranging from poor to excellent ... but never forget the #1 basic principle of data processing: You can't make stuff up out of thin air. /Every/ step of data processing, no matter what it is, is inherently /losing/ information - unless you're not actually processing but just format converting. In that case nothing is lost or gained, you've just changed how the same data are represented.

Thus: "Upconverting" is just blowing things up and filling the gaps. The upscaled picture does not contain more information (more "detail") than the original did - simply because the data for that are /not/there/ to begin with.

In other words: Bad upscalers can make things worse - but good ones can't make the picture better than it was in its recorded resolution.

Now, what upscaler to use? Two-edged sword. Theoretically, the LCD screen's integrated scaler knows how to make the best of the LCD screen hardware - but then, the rendering software (the DVD player application in this case) knows best how to handle the original source material.

Typically (but not necessarily) you'll get closest to 100% original quality by using a /good/ scaler that's near the source, and feeding the output device /its/ native resolution so it doesn't interfere.

ffdshow is a good choice - if your CPU has the horsepower for it.

 

[Fallen Kell]

Originally posted by: Peter
All those algorithms may be ranging from poor to excellent ... but never forget the #1 basic principle of data processing: You can't make stuff up out of thin air. /Every/ step of data processing, no matter what it is, is inherently /losing/ information - unless you're not actually processing but just format converting. In that case nothing is lost or gained, you've just changed how the same data are represented.

Thus: "Upconverting" is just blowing things up and filling the gaps. The upscaled picture does not contain more information (more "detail") than the original did - simply because the data for that are /not/there/ to begin with.

In other words: Bad upscalers can make things worse - but good ones can't make the picture better than it was in its recorded resolution.

I disagree almost 100% with that. Just take a look at http://www.htpcnews.com/index.php?option=com_content&task=view&id=132&Itemid=29
and tell me that the before and after pictures show that the image isn't better then in its recorded resolution. Now is this a perfect example? No, since we are also dealing with the screen capture, and conversion to jpeg image, etc., etc., all in the way, but in general, that is an accurate representation of the difference I personally see on my setup.

Now, what upscaler to use? Two-edged sword. Theoretically, the LCD screen's integrated scaler knows how to make the best of the LCD screen hardware - but then, the rendering software (the DVD player application in this case) knows best how to handle the original source material.

Yeah, in theory, and in theory alone, does the LCD screen's integrated scaler know what to do. The problem is the "in theory" version cost $800 or more to do, so it wasn't done, and instead you have the $30 solution in its place. Unless you are using $2000+ professional computer monitor or a $20,000+ professional display/TV, then you certainly do not have the "in theory" scaler integrated in your device.

 

[Steve]

Originally posted by: Fallen Kell

Originally posted by: Peter
All those algorithms may be ranging from poor to excellent ... but never forget the #1 basic principle of data processing: You can't make stuff up out of thin air. /Every/ step of data processing, no matter what it is, is inherently /losing/ information - unless you're not actually processing but just format converting. In that case nothing is lost or gained, you've just changed how the same data are represented.

Thus: "Upconverting" is just blowing things up and filling the gaps. The upscaled picture does not contain more information (more "detail") than the original did - simply because the data for that are /not/there/ to begin with.

In other words: Bad upscalers can make things worse - but good ones can't make the picture better than it was in its recorded resolution.

I disagree almost 100% with that. Just take a look at http://www.htpcnews.com/index.php?option=com_content&task=view&id=132&Itemid=29
and tell me that the before and after pictures show that the image isn't better then in its recorded resolution. Now is this a perfect example? No, since we are also dealing with the screen capture, and conversion to jpeg image, etc., etc., all in the way, but in general, that is an accurate representation of the difference I personally see on my setup.

All that's happened in the 'After' pictures is that the existing images got sharpened, to make up for the blurriness inherent when enlarging an image. Extra image data (more detail) wasn't added because it wasn't there to begin with, and the software can't guess what it would be.

Now, what upscaler to use? Two-edged sword. Theoretically, the LCD screen's integrated scaler knows how to make the best of the LCD screen hardware - but then, the rendering software (the DVD player application in this case) knows best how to handle the original source material.

Yeah, in theory, and in theory alone, does the LCD screen's integrated scaler know what to do. The problem is the "in theory" version cost $800 or more to do, so it wasn't done, and instead you have the $30 solution in its place. Unless you are using $2000+ professional computer monitor or a $20,000+ professional display/TV, then you certainly do not have the "in theory" scaler integrated in your device.
[/quote]

Huh?

 

[Slammy1]

I think the key point is where the extra pixels come from. Taking a source from 720x480 to 1920x1080 means a lot more pixels than data would grant (6x as many). How the image scales can range from simple to complex conversion. Programs like ffdshow give control over how the extra data is interpolated from the original, and with a fair amount of tweaking you can bring out detail from an image (though in actuality, you'll never obtain a better resolution than the source). Sorta like a photo editing software, you can sharpen, lighten, and do various things to smooth out an image. The fact that it's done on every frame is why the program is so CPU intensive, but you can run on pretty slow CPUs (you're just limited on how much you can do). ffdshow is all about filter control.

 

[Auric]

Ja, edge enhancement and/or sharpness in not necessarily desireble. If intending to view mucho olde timey low-res then better to use a CRT -especially if at such a small area anyway. A 24" LCD just cannot compare in any way to a CRT of equal size, nevermind a 34" direct view.