Sanyo HD2 footages

[John Markus]

Don't bother translating this site, just take a look at the samlpe movies to get an idea of how this camcorder will perform.

First the review site (page 1):

http://arena.nikkeibp.co.jp/rev/2007...1105/?from=RSS

Sample clips (page2) compares HD1 to HD2:

http://arena.nikkeibp.co.jp/rev/2007...5/index2.shtml



John,

BTW, I'll be posting another link soon that has a ton of HD2 clips taken in dark locations inside and out..... you won't be happy.

[John Markus]

As promosied, here is a bunch of dark HD2 footage, both video and stills.

http://fujiiphoto.com/2ch/HD2/

I have to say that I was really pulling for the HD2 to be a big leap above the HD1 but as you can see from both sites (other site posted before) the quality of the video is marginally better. Now I understand why Sanyo Marketing is not posting any real samples here in North America (That post on Youtube is pathetic), there's no way Sanyo will be able to compete with the Canon TX1 (Lots of samples available now on the net) at it's current MSRP price point.

Take a look at the Panasonic HDC-SD1, it's MSPR has already dropped before introduction in the States as well as Canada (It's not even widely avail in Canada yet). The Hybrid market is heating up. Yes, I know that Panasonic is way more expensive, but take a look at it's IQ, you'll be a convert as well ( I wish there was editing software available for AVCHD).

Please post your impressions of the samples.

John

[Peter Solmssen]

Thanks to John for the samples. The direct comparison of the HD1a and the HD2 is informative; there is clearly some improvement.
The dark samples don't tell us much (even after auto translation of the captions) except that none of the little cameras are much good in low light.
You can't compare them to the Panasonic, which is a different animal -- larger, more expensive and using the more troublesome AVCHD format. If Apple brings out models with hardware rendering of H.264 (as rumored) that might make a difference.
As between the Sanyo and the Canon, we haven't seen enough to be really decisive. The Canon sample from the PMA ballroom is lousy (pity the poor Canon rep who is depicted in it and seen around the world) whereas the Akihabara sample in daylight looks pretty good.
Since I am pretty much commited to the pocketable HD concept that also includes good quality stills, I am leaving both the Sanyo and the Canon on pre-order for the moment.
I have about a dozen quite satisfying videos for viewing on my HDTV that I shot with the Sanyo HD1 (video and stills combined in Quicktime HD), I'm glad that I didn't wait for something better.

[Wayne Morellini]

Thanks John. Just downloading pages now, did they try that hypersensitivity button?

I have been informed that the Panasonics HDMI output is compressed.

[Wayne Morellini]

From the images, these cameras are set up differently, different colourisation and saturation, and different edge enhancement. We cannot tell much, unless the cameras are setup manually to reflect their best performance.

The HD1a picture is wider (in girl at desk shot) and possibly has more latitude, which would be very disappointing if it is true, but then again could have something to do with how the camera is responding to handling at the moment.

[Wayne Morellini]

The closeups show a lot of difference between the two cameras in fine detail and all that compression artifacting. It definitely needs an faster bitrate.

An lot more investigation of these images needs to be done before we can say for certain what is happening.

[Wayne Morellini]

Good news gentlemen, in the review, snapshot and blowup the seagulls down the bottom for each camera. By the looks of it we have got rid of the dreaded jaggies encoding error. The seagulls now look normal, not like they are from the x-files. Unfortunately, it is hard to tell, because the HD2 footage lacks a lot of resolution, maybe because it is in an soft detail mode. If you look at the ladder in the blowup you will notice the vertical feature to the right of it is virtually missing in the HD2 footage, and if you look at the diagonal beams on the bridge there is hardly any detail there in the HD2 shots. Also, an train is goign through the HD1a clip (which would suck up some data bandwidth) and not in the HD2 footage. Water (and train) still looks blocky, still need better data rate.

The night lamp shots don't tell me much, and I can't download 126-14? MByte monster files from the reviews easily. But the lamp shots (which I assume is an paper lamp with an candle inside, don't prove much, street y street lighting, or room by baked 60/40/15 watt bulb light does (preferably without bulb in the frame, to allow optimal iris and gain). And none of the lamp shots were purple as happened previously, which is good news.

I am currently downloading night alley footage. It mentions hd1 down the bottom, does anybody know which of those is HD1 and HD2?

Altogether, I might have to wait for 18mb/s H264 camera unfortunately. 18mb/s 60fps 720 should be better than this, let alone 30fps 720p. But as with Sanyo, the other shots we have seen show better latitude and color, so maybe a few surprises in the handling and setup are waiting for us.

[Wayne Morellini]

Have found out Sanyo has just announced an new H264 camera:

http://www.dvinfo.net/conf/showthread.php?t=88841


Have looked at picture and read translation of review. Found my monitor set to 9300K. changed to other settings which helped with the bluish ting through the image.

Noise is less, and the article seems to indicate compression noise is less (as would be expected without noise) but hard to verify, as it to notes the lack of detail. It seems to indicate that this is an pre-production camera, and the situation may improve. I would imagine that the lack of detail is merely too a soft an setting, or the 9 pixel binning been done is such a way to average out neighbouring pixels. Some handling of the camera should prove wherever this can be adjusted.


Definitely not the high data rate (modern storage can handle it) high latitude, low noise, 50fps h264 camera we would prefer, but it is a good little camera anyway. Hopefully it handles well fro the consumer too.

http://translate.google.com/translat...language_tools

[Wayne Morellini]

At least one of the clip sources indicates that this is an pre-production camera, so some of the apparent problems might be related to this, and be resolved in the finale product. This is everything the HD1 should have been (less higher bitrate and 25/50fps (50fps being something desirable in sports television recording)). This is based on the existing clips I have downloaded, and as much as I have been able to downloaded, it is representative of those clips, but may not be of other clips on production cameras.

Jaggies, and horizontal details
Yippee the diagonal jaggies block issue appears to be gone, but there seems to be less detail in the horizontal direction which makes it seem to have another problem on the diagonal, and details submerging in the horizontal direction, examine the angular supports, and the ladder, on the clip with the bridge.

Pixel flashing/Interpolation
HD2 still has extreme pixel flashing when it comes to areas with high contrast, white on black areas, moving (the girls black and white stripes, the bright bush branch at the end of the flower scene clip. But it is less in the horizontal direction, where there seems to be some sort of detail loss in the clips.

Color
Not too saturated, but maybe a little under, with an blueish tint (the aim is for optimal saturation and color balance because I don't think color adjustment in post is the best option for this).

Noise/Codec artifacts
Primary colours still show noise, and even in the well lit in store shot, much noise was evident (the enemy of coding efficiency) but noise in general much less, especially in good lighting (see outside store stationary clip). Could this lack of noise be caused by the apparent lack in horizontal detail?

Extensive codec/noise in an border around edges, but better than HD1a examples (see the girls faces in the in house shot, and the edge of the wall).

The in camera noise removal/noise improvement, is important to get the most out of the codec in the data to start with.

There is extensive analogue noise in some of the images, analogue patterns and mioring (see the clip with the girls).

Plain area crawling blocking
That blocking/crawling blocking, on plain surfaces seem to be largely gone, some little stuff is viewable on the beige/pink bottom panel of the fridge, seen briefly at the start of the interior of shop clip). But I would like to see static dark bitumen road surface, and deep blue sky shoots, where this is much more evident on the HD1a clips I have seen.

The mystery of the white sky (clouds or overexposure?)
Blue skies seem white in places (clouds or overexposure?) unless near dark intervening objects on some shots, the static still of the street exterior, it can be seen around the overhead cables, one corner of the sky also gets blue.

Latitude
The latitude may not be improved, the snow skying shot, from Youtube, seems to show much better handling, but this might be from Neutral Density filter and internal camera processing. I am suspect that internal camera processing is causing the white skies, maybe the noise removal, it gets blue around the cables, even charge spilling over in the CCD pixels to surrounding pixels. There is another possibility, as slim as it maybe, that the dark feature is blocking some light entering the lens, bringing it just under over-saturation, not something I strongly regard, but I should mention it.

Motion handling
Motion still stuffs this camera around, looking at many clips on the Fuji site, the motion just trashes a good image, it is desperate need of more data rate, and higher frame rate would help. Movement seems to make it block more (but need to verify with more footage). As can be seen in the clips from Fuji site, the camera can suffer from an lot of handling shake. I think that the lower noise helps, but it works best on an tripod, stationary.

Waves
On the water/bridge/train video, waves, might, look better on the HD2, but the HD1a footage has an train moving through it sucking data up.

TV is moving to Cinema angle of view (100-200 inch TV's)

Used as an cinema camera, viewed at the distance of an good cinema seat, picture looks bland, resolution is not enough, even video like. Sitting further back in the theatre it should look better. 1080p would look better, with four times the data-rate, from an good seat 1/2 to 2/3rd way down to the front. This is an bottom end HD camera for Indie (I mean bottom end). I prefer it's picture to the compression in HD broadcasts I get here. I think the problem is resolution and detail, requiring extra data-rate, but with processing might give you an particular "look". Motion is the problem as far as restoration of detail and processing is concerned, which would require an certain handling.

Good news, is that all this should be able to be adjusted out in camera or post processed out, to attain a suitable professional, though flat, quality.

I will go quickly through the options, because I am not too familiar with them myself:

De-blocking, de-noise/grain, resolution upscaling, image restoration:

Color correction and de-noising using after effects de-grain, neatvideo, and noiseNinja in photoshop:
http://www.dvinfo.net/conf/showpost....09&postcount=3
http://www.dvinfo.net/conf/showpost....96&postcount=6
http://www.dvinfo.net/conf/showpost....14&postcount=7

Some threads on picture enhancement and de-artifacting/de-noising:
http://www.dvinfo.net/conf/showthread.php?t=73983
http://www.dvinfo.net/conf/showthread.php?t=74760

Somethings we discussed in the Elphel camera thread to treat MJpeg footage (convert to whatever high quality format that can take the enhancements and apply similar techniques):
http://www.dvinfo.net/conf/showthrea...t=63677&page=4
http://www.dvinfo.net/conf/showthrea...t=63677&page=5
http://www.dvinfo.net/conf/showthrea...ock#post532176

http://avisynth.org/

Image restoration, I have had an theory about restoring images from details in adjacent images and calculation, and resolution upscaling for years, there are now firms that do these things with massive computers (and Intel apparently has software). Image upscaling software is also available. Transfer to film companies have their own procedures.

Another image enhancement product is scaling software, on consumer equipment they use it to upscale DVD video to HD screens. The only free one I remember is:
http://sourceforge.net/projects/deinterlace/

There is probably much better alternatives now days, as many HD TV's and an number of DVD's use these techniques (some are much better than others, apparently). Can you get resolution and detail where it wasn't, Intel (and whoever else) had some software a few years back, but be prepared to use an nice GPU acceleration (if available) or run it on an PS3. Why, to make it smoother at an good distance.


In another forum, an user talks about using MPEG Streamclip with Vegas to convert VGA Mpeg4 to DV compatible file.
http://www.stevesforums.com/forums/v...25&forum_id=27


It would be tempting to do an short film with this to see how well it can be done. An h264 version would be interesting, I know there are other h264 cameras coming, so we will have to see. But 18mb/s Mpeg4 may have some advantages over 9mb/s H264.

[Wayne Morellini]

I thought I would add this frame posted to the Elphel security camera thread for comparison. It was shot at 2000*850 pixel frame, at 85% Jpeg quality.

http://www.dvinfo.net/conf/showpost....&postcount=544
http://www.buysmartpc.com/333/333framecc.jpg

Amazing what can be achieved with even an cheap sensor.

[Robert Batta]

http://www.sanyo-dsc.com/english/pro...ips/index.html

[Peter Solmssen]

So far, Sanyo is giving photography lessons rather than posting full resolution samples. Also, the reviewers who have been given Canon TX1's to test (e.g. "Small Temple Trust", Akihabara) have not been given Sanyo HD2's. However, some individuals do have them somehow and have posted clips here and there.

[Bill Busby]

Originally Posted by Wayne Morellini

... High Signal to Noise with an latitude extension scheme, and large sensor (at least 4/3rds).

4/3rds? I'm no mathmetician, but.... ;)

Bill

[Wayne Morellini]

Bill, it is an camera format, and if you think that is confusing, you see the actual size of an 1/2inch chip, at least 4/3rds makes some sense compared to 2/3rds. 4/3rds is an desirable solution I think (not as good as 35mm, but better than 2/3rds). The present HD1 sensor is around the size of some super 8mm formats. These Mpixel CCD's, just don't have the technology (range extension=latitude) or pixel pad size (latitude and low noise).

[Bill Busby]

Thanks Wayne. I'll take your word for it I guess. I'm no photo guy by any stretch of the imagination. But to satisfy my curiosity I Googled 4/3 format & came up with some thread from a Olympus forum that basically is just talking about a 4:3 (4x3 as well as several others) format camera. So it seems they're just talking about aspect ratio &/or dimensions.

I still don't get it though. If you're talking about CCD's, or a CMOS or whatever else kind of image sensor... Let's say there's a given length, no matter how small or large & it's divided into 3. How can it be measured 4/3rds? 3/3rds would be the max. Where does the other 3rd come from.

Arghh! Sorry... It's late & I'm losing it. :) I don't like late night edit sessions :-\

Bill