PDA

View Full Version : Analysis of Footage



DavidNJ
04-26-2009, 02:21 AM
It is interesting...anything over about 200lp/ph is pretty much gone in all the clips. However under 200lp/ph there is a uniform excellent contrast with no edge sharpening. Blowing up sections you can see a fairly large (mabye 25 pixel) section of a leaf with the absolute same shade of green. However, the overall picture has very good image sharpness.

The texture detail is what's missing. It is obvious in the spectral highlights (last place you think you'd find texture detail) where they seem muted.

The low light in the night shots seemed better than EX1 clips but not as good as 5DM2, as expected.

Overall, except for the bokeh and much better white balance the shots seemed awfully close to a Sanyo FH1. When Lis was talking about 'compression algorithms' was this what she meant?

Reference, one bird is a GH-1, the other a Sanyo FH1.

http://www.dvxuser6.com/uploaded/113/1240723622.jpg

I'm not positive yet, but I think the smoothing takes place before HDMI output, if it has it. Or rather, if it does on the Sanyo it is probably a good bet—at least 50-50—that the same algorithm, firmware or even chip was used on the GH-1.

The good news is that it always produces very pleasing images. The bad news is, if that is what is happening, it won't match the depth, warmth, and feeling of 5DM2 images.

dcloud
04-26-2009, 03:09 AM
is the footage of the bird from the youtube video? then its probably compressed. Try the one with barry's footage

not that you can compare a pretty woman to a bird :P

squig
04-26-2009, 03:29 AM
yeah it's compressed H.264 12.5mbps.

I haven't been able to confirm if it is AVCHD or mjpeg.

LizaWitz
04-26-2009, 03:47 AM
The camera's default movie mode is full auto with image correction. It looks like the footage we're getting was shot in that mode, and that's what David is seeing. Looking for compression artifacts in YouTube video is like looking for water in the ocean. It is silly to assume this camera shares a processor and firmware with a camera from another company that has a completely different design. But I am not surprised that you compared the GH1 to the FH1 (why because the names are similar?) and concluded that the 5D mk II won.

jamesmallonuk
04-26-2009, 07:14 AM
Of course its early days BUT everything I have seen looks like perfectly stable, great sounding!, VIDEOOOO, its screams Video. I gave up any interest in the camera for creating a more organic film look which is what I personally am wanting. Then I saw this clip http://www.vimeo.com/4337192 with a Nikkor attached and I again the Jury is still out for me, it has a nice bokeh and looks really nice, I guess I underestimated the influence of the glass in front of the camera. Now just need to see how a wider lens looks, at least a 14mm to get 28mm equivalent. Hopeful...maybe.

John Caballero
04-26-2009, 07:34 AM
This cam is gonna shoot some pretty sweet footage, there is no doubt about it.

DavidNJ
04-26-2009, 09:01 AM
Of course its early days BUT everything I have seen looks like perfectly stable, great sounding!, VIDEOOOO, its screams Video. I gave up any interest in the camera for creating a more organic film look which is what I personally am wanting. Then I saw this clip http://www.vimeo.com/4337192 with a Nikkor attached and I again the Jury is still out for me, it has a nice bokeh and looks really nice, I guess I underestimated the influence of the glass in front of the camera. Now just need to see how a wider lens looks, at least a 14mm to get 28mm equivalent. Hopeful...maybe.

Looking at that video it was uploaded as a 720p30 image at 30Mb/s. So some stuff was done to it. However, the detail is still about 180lp/ph. However the lighting, that he used good glass and kept it open, and the crispness at some edges which is the best yet displayed (of course it was a metal edge light silver next to deep black) implied this shooter knew what he was doing and had some opportunity to setup.

The bird was from YouTube and was hammered. However, the detail was about the same as in Barry's woman so I thought it was fair. The bird detail was really very similar.

The Sanyo (which has full manual controls) can take some very good images. Just like the GH-1 it seems to have used image processing as a way to manipulate low-bandwidth high-compression. In that decision they decided to remove high frequency detail which would not be accurately compressed yet use compression bandwidth. The result appears to be a very crisp image without edge sharpening that maintains its crispness to the lower range of texture detail.

In many ways it is an intelligent decision. Subjectively, the images always look good. They have less of a video look than many images that may be sharper. However, they never have the texture detail that makes 5DM2 pop.

There are two T1i clips out there, on 1080p20 and the other 720p30, both shot outside and including some motion. This is a camera without manual controls costing 3/5ths the price. A better comparison. http://www.imaging-resource.com/PRODS/T1I/T1IMOVIES.HTM

I was going to PM Liz to ask if this was the compression algorithm she talked about. However, judging from her post she is still working off a Saturday night hangover. Maybe later she will give a technical analysis other than comparing GH-1 to 5D Mark 2.

DavidNJ
04-26-2009, 09:18 AM
There are two clips from Japan my the same person on youtube. One (http://www.youtube.com/watch?v=OEt0q0vU2_A) using a 50mm Canon FD lens which shows bokeh but also the loss of texture detail (look at the leaves). The second (http://www.youtube.com/watch?v=rHR62kKyjok) on a motorcycle that shows jello. However, had he shot 720p60 and converted to 1080p24 that would have probably been reduced or eliminated with no effective loss in image quality.

I remember a comment on one of the PMA videos where the reviewer, looking at the Panasonic screen at the display said: 'It looks like TV'.

dcloud
04-26-2009, 09:28 AM
no one in their right mind would hard mount a cam and shoot on a motorcycle. even the 5dmkII would jello in that situation.

http://www.youtube.com/watch?v=afHZoLK_2xM
heres a nice footage.

seriously, comparing youtube clips for details? maybe motion.. but details? i dont even get close to a TV to look at details for christ sake.

LizaWitz
04-26-2009, 09:30 AM
The compression algorithm I was talking about was the suite of feature extraction options that are used with H.264. What you're describing is image processing that's done for consumers. It will be interesting to see what the video looks like with this kind of "foolproof enhancement" turned off.

H.264 is a compression method that looks for similarities between frames and then encodes just the differences to maintain high resolution at lower bitrates. There are a number of different algorithms you can use when encoding to look for these features, and the H.264 standard specifies how you encode the resulting features-- not how you look for them. Thus how many frames forward and backward you search, what mechanism you use to look at the data to extract features, and what range of levels of detail you search for are all up to the person designing the encoder. If you have a lot of horsepower at your disposal, you can use more methods in your encoding to better handle more edge cases and produce higher quality images per bit. If you have less horsepower you are restricted in the degree of feature extraction you have time for when encoding, and thus to get the same level of detail you'll have to have a higher bitrate.

Maybe Panasonic isn't putting much horsepower into their encoding and the images will suffer at a 17Mbps bitrate. Or maybe their vaunted dual CPUs allow for significantly improved encoding quality in realtime and thus the 17Mbps bit rate is more than sufficient.

At this point, to do a proper cinema test we need to get a camera, turn off the "tourist mode" and put it into a cinema mode where the images are not "enhanced" and see what kind of video the camera can put out.

Like every other camera, it will take some time to figure out what exactly it is capable of.

mico
04-26-2009, 09:51 AM
no one in their right mind would hard mount a cam and shoot on a motorcycle. even the 5dmkII would jello in that situation.

http://www.youtube.com/watch?v=afHZoLK_2xM
heres a nice footage.

seriously, comparing youtube clips for details? maybe motion.. but details? i dont even get close to a TV to look at details for christ sake.

Other than hard mounting what are the other options? Handholding?

Everyone hard mounts to motorcycles if thats what you need to do. The question is how did he hard mount. It makes a difference. To say no one in their right mind would hard mount a camera and shoot on a motorcycle is wrong. If the guys mount was solid you should have said its ridiculous to hard mount the GH1 to a motorcycle because that looked HV20 bad.

Kholi
04-26-2009, 10:17 AM
This entire thread is invalid until David actually gets a unit in his hand.

Or hell, until someone who actually has time with it and cares about the video gets one in their hands.

This man is scrutinizing a 720/30P MJPEG sample off of youTube. Gimme a break.

DavidNJ
04-26-2009, 10:30 AM
no one in their right mind would hard mount a cam and shoot on a motorcycle. even the 5dmkII would jello in that situation.

http://www.youtube.com/watch?v=afHZoLK_2xM
heres a nice footage.

seriously, comparing youtube clips for details? maybe motion.. but details? i dont even get close to a TV to look at details for christ sake.


That clip has about the same 180-200 lp/ph resolution as the others. For comparison a very good DV camera has a theoretical maximum of 240 lp/ph. My DV camera is very sharp (much shaper than a DVX100 was, never compared a and b versions) and not much different than the clips I've seen although it starts as 60i.

What is disappointing is that the sensor has to be seeing the image we (I?) want. Encoding like Canon does required very little power (They do it a $600 SX1, $350 SX200, $280 SD780). If they had just done that...

Isaac_Brody
04-26-2009, 10:50 AM
Or rather, if it does on the Sanyo it is probably a good bet—at least 50-50—that the same algorithm, firmware or even chip was used on the GH-1.

David, I'm not sure what the purpose of your thread is. To assume that two different cameras from competing companies are sharing the same algorithm, firmware or chip just makes me question your purpose in singling out this camera. I feel the same way about people who single out the 5DII and nitpick it to death. People like the image that produces, and so far people like what the GH1 has put out there. People are very aware of the limitations of these new cameras, we can see it with our own eyes. What we don't need are assumptions or people pixel counting youtube footage which is further compressing the original material. It doesn't help, it just confuses and adds some FUD (fear, uncertainty, doubt) to the fire. Enough of this...