Stunning Resolution vs Spectacular Clarity
 

Michael, I am still learning, but in Photoshop or Premiere when they de-interlace video for still photos I thought they threw out upper or lower fields (you choose which) of interlaced video frame and replaced missing fields with their interpretation. So, when you de-interlaced the 60i, weren't you throwing out half the fields and relying on software to fill in the missing fields?? On a non-moving shot like the res chart do you need to de-interlace?? I only de-interlace when trying to get a "still photo" from interlaced video in which the subject race car moves enough in the 1/60 second between fields that I get offset fields and hence "blur."

This does get at all the posts I have seen both ways regarding resolution or maybe more the "look" one gets from Normal Movie Mode (NMM) and Frame Movie Mode (FMM) with GL2 when viewed as video and not as still photo. In GL2 brochure they may touch on this when they refer to "STUNNING RESOLUTION AND COLOR REPRODUCTION" when using NMM and "SPECTAULAR CLARITY OF SUBJECT" in FMM mode (my caps just to highlight what Canon says in brochure). Maybe for some reason the resolution is better in NMM, but "less blurring" or better clarity occurs in FMM because there is no offset of fields for rapidly moving objects?? GL2 brochure does seem to talk about FMM more in context of still photos, video for website contents and "cinematic like appearance" and not in context of normal video.

I have seen posts saying you get better vertical resolution from FMM vs NMM, and other posts saying you get a more "film look" from FMM (aside from strobing I have also seen referenced). I thought the better vertical resolution could be due to capturing 60 fields at once vs 30 fields 1/60 of a second apart. Maybe that "film look" comes from lower resolution (less sharpness) in FMM vs NMM??

It seems that people never agree on this. I like frame mode much better always. I've tried de-interlacing 60i footage (mainly with my HDTV which deinterlaces ALL 60i input), and there is definitely a palpable difference. The frame mode looks much more like 24p film. And, on my own resolution comparison I found the frame mode to look just as good.

I am a novice, but I have tinkered with both Frame and Normal on my GL2. I will say that the Frame does give a much more film like look than de-interlacing later on. The de-interlacing after the fact tends to leave more artifacts as some intelligence is used to interopalte the fields into frames. I see this more on footage with fast motion. I also like the very crisp detail and resoltuion of normal. SO I see uses for both. I actually liek normal for nature shots and stuff where I want extreme detail. Frame is good for that "elegant" film like look where the nit details are not as important. It also takes more practice with Frame to avoid struttering images when panning and such. The great news is that cameras like the GL2 work great in both and give you the flexibility to do either!

I agree the temporal flow of frame mode is better. But the facts are frame mode discards around a 100 lines of res. Takes a 480 full frame image to around 380 or a little less on my charts. If you turn down sharpness as you should it gets lower.

I wrote the first article on frame mode two months before the XL1 came out back in 1997. I was its biggest fan because I had been trying to get companies since 1988 to install adjustable frame rates in video cameras and post technology before it was really known or hip to have 24P/30P. I was pretty happy about FM's arrival . In the early days it was a nightmare trying to explain frame rates and its affects on the brain to people. I would have to drag people I knew over to Showscan in Culver City ( 60FPS 65mm format ) 12+ years ago to show them that this film running at higher frame rates looks like the best hd video you ever seen. The only clincher was, it was not HD, but film that was shot a 60fps and projected at 60fps. I would show people video copies of that stuff and they would swear on the mothers that it was video. It was then that I started to slowly convert the people I knew in their ways of thinking. Doing this back in the late eighties when no one in LA knew what the "F" I was talking about was not easy. I can think of two people maybe three that knew in LA. Rebo of Rebo HD, Douglas Trumbull and the guys at FilmLook. Outside of that, people thought you were nuts to want to change video frame rates and believed temporal res was horse shit. Now it's common talk...

So I agree that frame mode has a great look, but if your going to uprez to HD or go to film stay at 60i and keep all the rez you can. I showed Chris Hurd a while ago some footage at Bob Jones house ( The skycrane guy) that I shot with the VX1000 and converted to 30P. It looked great. I think he can attest for that. De-interlace requires more then blending fields but proper blurring of vertical data , Etc. Experiment, and come up with a formula that works for you. Otherwise then just shoot frame mode, it's your creative decision in the end. Good Luck!

Michael Pappas
Arrfilms@hotmail.com

I have no reason to doubt you and I ahve not tried the Magi Bullet software that is gaining favourable reviews, but I did extensive experiments with DVFilmaker and exchanged e-mails with it's makers.

I found with the XL1s (Pal) the frame mode came out on top 100% of the time. No matter how I tried to deinterlace the normal 50i footage I got very unacceptable jaggies and artifacts. Sure it looked like film but the resolution suffered more than with frame mode.

No matter what people say about the loss of resolution between frame mode and 50i if you setup your camera with a boradcast monitor and switch between frame and normal the 'percieved' loss does not seem to be 25% and is very minimal indeed.

I have shot commercials and video with frame mode professionally and both clients and boradcasters seem to be happy - I was looking for a way to preserve even more rez and achieve the same 'film' look with software -so far the results have been unbroadcastable!

Regards.

Michael,
I want to ask you about the resolution loss you're talking about. You might be right that the CCDs only use 75% of their resolution when in frame mode.

But, Canon says their CCDs are 410,000 pixels. That gives about 640 x 640 resolution of just the CCDs. If you take off 25% of that 640, you get (nor surprisingly) 480, the exact vertical resolution of MiniDV, and the exact resolution of the resulting digital film frames.

Isn't the proper calculation how much frame mode reduces the resolution of the CCD, and not of the MiniDV image? And if I'm right, then reducing the CCD resolution to 640 x 480 is only minimally harmful because it creates images in the exact output resolution. If anything, it's better.

So I guess my question to you is are you certain that in your 75% calculation, that's 75% of the resolution AFTER the CCD images are converted to 480 lines, or before?

It is important to distinguish between pixel count and resolving power.

The SD TV image will always be the same , i.e. 720 x 486 NTSC and 720 x 576 PAL, but the resolving power of the CCDs (i.e. the ability to resolve individual lines) will change depending on the mode etc. Many pro camera data sheets specify that the chips can resolve 800 lines +, but the DV25 standard on tape or in a PC can only resolve a little over 500 lines. So when people talk of losing 100 lines rez that is not the same as losing 100 actual lines from 486 or 576 vertical TV lines. Of course starting with that much greater image resolution at the camera head in the first place certainly pays dividends.

This why I believe the frame mode does not show a straight subjective drop of 25% vertical resolution.

It is important also to bear in mind that many filters like diffusion, Pro-mist etc. will have an effect on resolution, possibly in some instances greater than that produced by frame mode, and yet these filters are regularly used on Digi Beta shoots.

The fact that it makes frames out of the raw pixel shift data helps. But aliasing is increased as soon as you switch to FM. I did extensive tests with the GL2 and it put out a nice image but it couldn't hold up in the uprez process. The GL2 is a very good camera for its price. It's image is sharper then that of the XL1S. But FM and the way it emulates progressive is not the best way. It was explained to me by Canon that a field is captured ( not a video field but raw ccd data ) and the offset pixel shift data mixed in there completes it as a full frame. The look of frame mode still has a film feel and is more pleasing then 60i look. By the way I can't speak for Pal cameras since I'm in ntsc land.

If you really want to see what frame mode does to the quality vs regular. Film a 1956 resolution chart. Set up a 800+ line monitor. Switch between the two modes. That's speaks more the all the typed words I can do.


If you want to see what true Progressive DV looks like. Check out John Beales latest tests. Scroll down to the last few tests. There you will see what 60i looks like vs true Progressive. Even on the DVX100 it's 60i test can't compare when it's switched to 24p mode.

Here is the link:

http://www.bealecorner.com/trv900/cats/cats.html

Michael Pappas

Canon Tech Service Explanation of FMM
 

Michael, I don't know the people you know at Canon, but I called Canon tech service and the person (after considerable clicking on his computer and talking to other people there) said " In Progressive Movie Mode/Frame Movie Mode every other frame has 3/4 of the lines of resolution and the next frame has all the lines of resolution which is why it can look choppy." He said I could get the best quality still photo(vs interlace mode) by using Frame Movie Mode and selecting the frame that had all of the lines to print out. I am just repeating what he said and I don't know what this means, but the "every other frame has 3/4 of the lines" does tie in to earlier comments about losing 1/4 of the resolution. The every other frame thing could also possibly tie into other comments about the strobing effect you get when using Frame Movie Mode (if every other frame was slightly different). Does any of that make sense??

"If you really want to see what frame mode does to the quality vs regular. Film a 1956 resolution chart. Set up a 800+ line monitor. Switch between the two modes. That's speaks more the all the typed words I can do."

I too , as I have already mentioned, looked at both modes with high rez broadcast monitors (PAL only). I have 20:20 vision; have never worn glasses; have worked in film and broadcast TV for 20 years and I simply do not see the dramatic drop in quality that you imply when using frame mode as opposed to normal interlaced. Of course there is a drop in quality (and let's be frank we are not comparing XL1s frame mode to Digi Beta progressive) but I do not see a 'subjective' drop of 25%, and as I mentioned filters can have a major impact upon this matter too.

We have to be careful when we use comparative resolution tests because much creative film and video manipulation is by its very nature damaging to resolution yet the end result remains pleasing.

There's another way to tell that frame mode is not nearly as bad as some might think. Just compare it to 16 x 9 mode, where a good part of the image resolution is without a doubt lost. Compare a 16 x 9 mode interlaced video to a 4 x 3 progressive video and you'll see that the 16 x 9 mode has MUCH less resolution than the 4 x 3 frame mode. Probably because the 16 x 9 mode actually takes the 480 line video after processing and cuts it off into 16 x 9, as opposed to using the native CCD pixels, which is what it should.

I'm a novice to this whole thing but i spoke to a couple guys who knew more than me and the best advice they gave me was "trust your eyes". Whatever looks better to you and you like the look of more, go with that.

I was all excited about frame mode when i got my GL2 but after using it on a day of shooting some mountain bike footage and then working with it in premiere, I realized that i prefer
de-interlacing alot more. I think it looks better and I can get very smooth slo-motion when shooting in Normal.

Resolution charts and prints of kittens only tell part of the story. If all I did was shoot resolution charts and posters it would be easy to tell which camera is best. Take a look at how Popular Photography does it's test. Yes, they do all the prerequisite charts. However, subjective user data is also used to determine which image looks best. Similar online reviews are found at http://dpreview.com/ and http://www.steves-digicams.com/default.htm subjective reviewer results are always included.

Resolution charts are fairly outdated as a means of evaluating total performance. Zeiss lenses in the '80's didn't always have the highest resolution. Yet their images were generally regarded as the best. Zeiss took the forefront in championing the use of Modular Transfer Function as a means of evaluating optical performance. Highest resolution doesn't always produce the best image.

Jeff

All this talk but no movies....

I'd like to see the GL2 perform these operations:

Frame mode: of a typical action shot either panning or not
Normal mode: Same shot as above

16:9 Frame mode: of above shot
16:9 Normal mode: for above shot

In full resolution and mpeg2 compression maybe a 3-4 sec clip each....

Does anyone have these already available?

I'll be purchasing a camera in the next month or two and don't have access to and pro-sumer cameras without going 3 hours round trip out of my way to test them.

Any help would be greatly appreciated.

<<<-- Originally posted by Michael Pappas : ....A few weeks ago I took Canons GL2 and filmed a 1956 res testing chart and macbeth color chart. ....

Michael Pappas -->>>


Michael

Do you still have the frame grab of the 1956 chart that you shot? If so can you send it to me at krisp9@yahoo.com

Kind Regards

Jayvid