The Ambiguously Progressive Vixias

Andy_L wrote on 2/1/2010, 9:13 AM
I keep seeing questions as to whether or not Canon's Vixia HD Camcorders are progressive or not. I thought they were, then I thought they weren't, then I thought they were again, and now I think (with some confidence) that they are not.

It looks like what is really happening when you shoot in "progressive" mode with a Vixia is that the camcorder takes a 60i field scan from the sensor, uses some sort of software interpolation/blending to create a 30p frame, and then divides that field in two and stores both in a 60i stream.

The 60i stream can be recombined in Vegas to produce a "progressive" output, but the critical difference is the original scan was interlaced, so you're getting no more resolution than if you just shot 60i and discarded every other field yourself.

I have verified this by shifting between "interpolate" and "none" in Vegas on a Vixia-progressive clip and looking for differences. I see none. Do this same experiment with real progressive footage, and you can see a clear loss in resolution when you shift to interpolated.

I also got similar information from Canon's service department, though I wouldn't take that as absolute proof.

Assuming I'm right about all this, the alarming thing is that Canon is billing its Vixias as progressive cameras--especially the upcoming models. Which makes me wonder if other manufacturers are doing/going to do something similar.

"Progressive" may not mean "Progressive Scan" when it comes to product features in consumer designs, so be aware.

Comments

Coursedesign wrote on 2/1/2010, 9:57 AM
Words are cheap for manufacturers today.

I was really pissed off recently when I found that some LCD TV manufacturers change panels without changing the model number.

This is especially egregious when the change is from a wide-angle 10-bit panel to a cheesy 8-bit panel where you need a head clamp to assure full color and brightness. You have no way of knowing which you are buying without taking it out and starting it up (in some cases you can check avsforum for manufacturing dates).

Where are all these infamous "greedy trial lawyers" when we need them?

Maybe they are too busy with all the other consumer fraud that is filling our society today.

Also on my sh*t list today is a certain administration for accepting a $45B TARP money payback from Bank of America that had major conditions attached, while immediately giving the bank the money back as an ultra-low interest loan with no conditions attached...

We are getting screwed royally. The Big O (hint: it's not Oprah) is genetically related to six or seven past presidents. Whose genes are acting up here?
FilmingPhotoGuy wrote on 2/1/2010, 11:11 AM
I find this very interesting. Vegas reports progressivly shot material from my Canon HF10 as 50i even I know I shot in progresive. The setting gives 2 options, 50i NORMAL or PF25. Why would they give you the 2 options if they were captured 50i anyway?

BTW what is CINE MODE?
jabloomf1230 wrote on 2/1/2010, 11:48 AM
Why is this issue being raised again? The Canon consumer camcorders store everything in 60i, no matter what mode you have it set to. In 60i mode, the upper and lower fields of each frame are displaced in time. In 30p mode, the upper and lower fields are at the exact same time. In 24p mode, you have a hybrid of the two with a fixed pattern of some progressive and some interlaced frames. The earlier VIXIAs had no flags for software to figure out the 24p cadence. The most recent models have flags, so that the cadence is established from the first frame.

The 30p mode is true progressive. It's just stored as two separate fields.
Coursedesign wrote on 2/1/2010, 11:59 AM
Canon has a 10+ year history of offering "24P" that was really "24F" aka Frame Mode.

Not quite interlaced, not quite progressive.

I believe the most recent models are real 24P, but I certainly wouldn't bet my life on it.
Andy_L wrote on 2/1/2010, 4:58 PM
jabloomf1230,

The 30p mode is not true progressive. Take another look at what I wrote.
BudWzr wrote on 2/1/2010, 9:24 PM
The latest Funkyzeit is that Canon is "out" and Lumix is "in".
FilmingPhotoGuy wrote on 2/2/2010, 12:39 AM
Bud, Bud, Bud What the hell is a funkyzeit and what "reviews" has it ever written on anything? Their superhero is Bruno or Borat?

C'mon give us something with meat on it.
Rob Franks wrote on 2/2/2010, 3:10 AM
Canon has a nasty habit of "over-advertising" (if you will) with their consume line of cams. Well, they all do.... but Canon tends to be the worst offender. I remember when the HV10 first came out. they actually advertised as 1920x1080 image. Of course they left it at that and didn't quite go on to explain that you don't get 1920x1080 out onto the tape. When the 24p came out on the HV20, they didn't quite explain that is was in a 60i wrapper....WITHOUT pull-down flags. Now out comes the 30p and here we are pretty much in the same boat again.

The interesting thing is that their "over-advertising" seems to work because they most definitely are sucking a lot of people into this.... rather unique (choosing words carefully) progressive style. But what amazes me about all of this to no end is that the "24p look" that these cams put out don't even come CLOSE to the 24p appearance of film... yet it just keeps getting gobbled up.
Xander wrote on 2/2/2010, 4:44 AM
Looking at the onscreen options on my HF11, the three recording modes are:
60i
PF30
PF24

This is repeated in the manual and the manual also states:
"Recordings made with the [PF24], [PF30] frame rate are converted and recorded on the memory as 60i".

I am not sure where this confusion is coming from. I would say that the only annoying thing is that the stream is not flagged as PF24 or PF30.
FilmingPhotoGuy wrote on 2/2/2010, 5:03 AM
This is one of the things that puzzled me. The professional Canon XH-A1 only captures at 1440x1080 and cost more than double than that of the Canon HF10 which captures at 1920x1080.

I'm still happy with the quality of the Canon HF10.
Rob Franks wrote on 2/2/2010, 5:23 AM
The A1 is older technology (tape/HDV) which is 1440 basically brought on by a restriction in the old style hardware. The hard drive/flash cams don't have this hardware restriction so they can go beyond the HDV spec.

Not withstanding, although the HDV technology is still being sold (and will be for a while yet), it is not really being advanced anymore.
BudWzr wrote on 2/2/2010, 7:06 AM
"Funkyzeit" is just a parody of the "in/out" list, where the new becomes old, and the old becomes new. Where today's truth becomes tomorrows lie. The "vortex", if you will.

Canon has become "tiresome", haughty, and fails to listen to customers.

===============================
Bud, Bud, Bud What the hell is a funkyzeit
Andy_L wrote on 2/2/2010, 7:48 AM
Xander,

I'll try this again. Your PF30 mode is actually taking pairs of interlaced fields (not frames!) captured at 60i, discarding one, and interpolating a progressive frame in-camera, which is then split again and stored as 60i. Yes, you can recombine the 60i stream in Vegas to get "progressive" 30 fps footage, but this is identical to shooting in 60i and using Vegas to interpolate to 30p.

In other words, it's almost impossible to explain the difference, but a Vixia's 30p is not what most of us want when we say "gimmie 30p".

disclaimer: this is my current best guess as to what is happening, based on conversations with Canon service personnel as well as my own experiments
david_f_knight wrote on 2/23/2010, 8:45 PM
Your hypothesis is false. Your test is inappropriate. What you are claiming is that footage shot with a Canon Vixia in the PF30 mode has 540 lines of resolution. You should test that directly, and by doing so your hypothesis is easily disproved: grab a frame of something with a fine diagonal line, such as a single strand of hair, shot in PF30 mode, and look at an enlarged view of it in any photo editor. You will see (at least, I see with footage taken with my Vixia HF200) that it does indeed have 1080 lines of resolution (and none of them is interpolated). Q.E.D.
PerroneFord wrote on 2/23/2010, 11:50 PM
You're wasting your time.

The only true progressive cameras Canon make are the DSLRs. They make no PsF 60i cameras that I am aware of.
david_f_knight wrote on 2/24/2010, 10:33 AM
Not at all. The Canon Vixia PF30 mode shoots true progressive frames. Period. The defining characteristic (from the image quality point of view) of a progressive frame is how it was exposed and captured, not how it is stored on the recording media. The Canon Vixia PF30 mode exposes and captures the entire frame at one instant, so that makes it a true progressive frame in every sense regarding image quality. That is, there are NO interlacing artifacts, and the resolution is EXACTLY as advertised.

If you want to make an issue about how progressive frames are stored on the recording media, well that's fairly silly because that does NOT affect image quality in any sense. Period. The only significance to the way a progressive frame is stored on the media is that it dictates how it should be READ from that media. If it weren't for that one fact, Canon would not even need to disclose how they store PF30 mode frames on the recording media. If it doesn't affect image quality, then NOBODY has any legitimate reason* to care how it's stored (other than to know which deinterlace setting to use for the media in their editor). Canon has not lied or mislead anyone about the claims they have made regarding the PF30 feature.

* I get the impression that many people believe they have been cheated or some other evil by Canon by storing PF30 progressive frames in an interlaced format. I believe the true reason Canon chose to do so is for backwards compatibility. That is, older or cheaper TV equipment that can't display non-interlaced video would be unable to show video shot with progressive frames unless it is stored in an interlaced format. So, by storing it in an interlaced format, Canon Vixia's PF30 mode can play on both interlaced and non-interlaced TV equipment. Is that really such a bad thing? Universal compatibility plus the highest image quality. Clearly a win-win situation.
CClub wrote on 2/24/2010, 12:53 PM
In addition to David's information above, I believe it's incorrect to state that the HV20 doesn't produce true 24p. There's a difference between 24p and 24f.

I recall researching this when looking at an HV20 a few years ago, and I saved a quote from David Newman (Cineform software, where their product actually lists the HV20 as an example of 24p recorded in a 60i wrapper): when asked whether or not the HV20 is actually 24p:

"Yes, it is 24p. It seems that there have been a bunch of forum threads attacking Canon, saying that this awesome little camera doesn't really shoot 24p. Not that misinformation is unusual for the internet, however, these posts often quote me or CineForm as backing this position. Neither myself nor CineForm support these posts or claims. The problem arose when I did state that there can be a subtle issue for chroma keying when using any 4:2:0 24p signal encoded into 60i. Some users took that and ran with it. I had seen some of this in customer footage, nothing I have shot. Basically, the 24p signal is good, and the CineForm pulldown from 1080i60 HDV tape works perfectly. That is my position."

All you need to extract the flags from the Canon 24p footage is the right software (Cineform, etc.).
Melachrino wrote on 2/24/2010, 2:42 PM
David is quite correct in most respects. The characteristics of an image are strictly determined by how it is captured and how it is reproduced (rebuilt), assuming that the storage function does not introduce artifacts.

A very simple test of whether a camera is exposing in progressive mode (the whole picture at once) is to shoot a clip in 30p setting with some horizontal motion and use a frame grab, or save a still in Vegas without interpolation, from the clip in question. If the exposure was progressive, then Field 1 and Field 2 were made from the same time instant and there will be no serrations on moving edges. (Different from blur). If, however, the exposure was made in interlaced mode, then, since Field 1 is exposed and captured at a different time than Field 2, there should be serrations on moving edges.
Notice that I say exposure (at the CCD).

I concurr that compatibility is precious, therefore Canon's choice is good in several respects.

24p is an entirely different matter on which I pass.
jabloomf1230 wrote on 2/24/2010, 6:15 PM
The HV20's 24p within 60i has no "flags". What Cineform (and a few other pieces of software do (like TMPEGenc XPress) do, is to establish the cadence of the interlaced and progressive fields. Unless the clip is very, short, this technique works like a charm.
drmathprog wrote on 2/25/2010, 3:47 AM
Does any of this mean that video from a Canon HV20 or S10 will cause problems in the latest Vegas Pro?
Dach wrote on 2/25/2010, 5:19 AM
I have had no problems with footage from my HV20 and Vegas Pro. When respecting the fact that all things come down to personal preference I only film in 60i.

There's never really been a discussion if 60i... is 60i. Also, when I was first introduced to the "jagged" effect when panning with the A1 in 24p mode I lost interest.

The film "look" involves a lot more effort than just chooosing 24 frames. My opinion any "consumer" camcorder... is a point and shoot device no matter what features it has. Leave it to the marketing department to produce a dog and pony message. (That applies to every manufacturer.)

Just for fun... Why the magic number 24 frames, is it because theres 24 hours in a day? Is it because 24 looks so appealing... did they even experiement with 27 framse. Is it because they were curious back in the day in how they could get rid of the "flick" effect... perhaps an engineer said lets try 20 fps.... 21 fps... hey look we got rid of it at 24 fps.

It seems camcorders and politics have one thing in common... they both try to be progressive and the public still wonders what is truly going on.

Chad
Andy_L wrote on 2/25/2010, 6:25 AM
David,

Perhaps you should tell this to Canon. As I wrote, Canon themselves confirmed that the cameras do not take a progressive scan.

As for other comments on looking at screen shots, if the Vixia is constructing a "progressive" frame by interpolating from an interlaced field, using horizontal motion as a test would give a false negative.

I'm really not sure how the Vixias work, but it seems to me there is good reason to suspect (as Canon themselves told me!) that they are not taking a true progressive scan.
Melachrino wrote on 2/25/2010, 10:37 AM
Dach:

You are right. Historically, 24 fps was chosen, eventually, by the film industry many years ago, before TV, as the compromise between cost of film, editing and other handling nightmares, (keep the number of frames as small as possible) and large area flicker.
This choice of low frame rate was helped at the time by problems in economically getting bright enough projectors. In addition, and hand in hand with the former, it was acceptable to darken the projection room as much as possible to obtain better results.

Along came TV and the engineering choice in the US was 60 fields per second so that the much brighter images, viewable in normal lighted living rooms, would not show large area flicker. There is a well established formula that relates large area flicker to image brightness and repetition rate. (Fink TV Engineering Handbook). Thus, the choice was not arbitrary but scientifically and engineering grounded.

Our modern camcorders emulate more the TV video methodology rather than film since the recording is going to be viewed at home, on TV's and in brightly lit living rooms. But, it is possible to provide some form of "film" look for those in need.

Melachrino wrote on 2/25/2010, 11:03 AM
Andy L:

I do not know of any method in existance today for interpolation from interlace to progressive, where motion is involved, that provides a perfect result such that you cannot tell the characterisitics of the source. This is very different from "acceptable", "good enough", and there is nothing wrong with such results. But a perfect interpolation would be the holy grail and none of us would be talking about this subject if it existed.

There are always artifacts and compromises needed when interpolating - just the choice of the term itself tells you: "aproximation" via calculation - from interlace to progressive in motion areas. If one knows the algorithm used, then one knows what the artifacts and compromises were and what the side effects are also. As I pointed earlier, IF you run the test I suggest, you will be able to determine for sure whether the EXPOSURE was interlaced or progressive. It does not matter what process or interpolation the camera does AFTERWARDS. You can tell.

By the way, I think this question and subject should be raised more properly to Canon video design engineers.