Hi all,

Has anyone done resolution testing with charts for the HVX200? If so, please post some still frames.

For those users that have the camera, but don't have a chart, feel free to use my Resolution Chart Generator located at http://www.goofers.org/reztest/. This is a general purpose line pair generator that can be used with any imaging system. For HD, you may want to try using the recommended settings indicated above the form, although printing on 8.5x11 may work. Make sure to print B/W only, no color.

I'd be interested to see how the contrast pans out using various resolutions and shooting modes (720p, 1080p, 1080i, etc). Also, notate the camera settings for each image (F-stop, gain, gamma, etc).

This test may clue us into the resolution of the CCDs, and give us an general idea of how this camera resolves images. Of course, there are other factors that play into the image quality, but this type of test will allow us to measure one important aspect of this camera.

Note: If my server bandwidth goes down to nothing, I will remove the link and place the generator program on a different server and let you know where that is located (when I can find one).

Have fun,
Bob

Thanks Robert,

Shoot me an email. I'm happy to make DV Info Net's server space and bandwidth available to you.

Bob, typically res charts have the line pairs at several angles to avoid aliasing issues.

you should look up what aliasing is if you don't know how that can influence the results.

Pathological worst case example: I take a photo of your chart with a digital camera, and my cameras ccd sensor lines up looking at each black to white transition ( one to one ) . I get a totally gray frame with no image detail. That's because each CCD sees a half white/half black thing ( = gray) . Now I move the camera 1/2 a pixel, so each ccd sensor sees either a black or white bar. Now I get full resolution !! If I turn the camera 2 degrees CCW , I get a wild pattern ( moire pattern ).
But by imaging a pattern with several 'rotations' or a starburst type line pattern, we can from that single image determine what basic resolving power the camera has.
Do you see what I'm getting at ?

-Les

I see what you mean Les. Thanks for bringing up that issue.
I know the chart is not full-proof in this version, but it could give us an objective way to determine approximately the resolving power. We did notice some of the gray values that you're talking about and tried to compensate for it by shifting the camera.

I can add a feature to generate other shapes such as concentric circles. Would this do the the trick?

Thanks,
Bob

Bob, The circles would be great!
Many circles with smaller and smaller spacings would let the observer pick the one where they can no longer see any details.

Then, let's see the data hit the fan, when we can convince the different camera owners to shoot the same exact test chart !

-Les


I see what you mean Les. Thanks for bringing up that issue.
I know the chart is not full-proof in this version, but it could give us an objective way to determine approximately the resolving power. We did notice some of the gray values that you're talking about and tried to compensate for it by shifting the camera.

I can add a feature to generate other shapes such as concentric circles. Would this do the the trick?

Thanks,
Bob

Sounds like fun to me.
I'll get started on that right now.

I'll keep you guys posted when it's ready to go.

One more thing before I get to work on this, do I need multiple converging circles, scattered throughout the frame, or can I have just one set of converging circles centered on the page? Also, is it required to print "gage" numbers near some of the circles in our case?

BTW: The charts are generated procedurally with PHP, so you can make the charts as large as you please and at any resolution.

Thanks,
Bob

Hey Chris,

Thanks for the support. Once I get this thing nailed down the way dvinfo.net users want it, I'd be happy to give you the source code to put on dvinfo.net. It will require PHP, which you should already have, and the GD image library.

Thanks,
Bob

Having many circles , each with different # of rings would be the best IMO.
Numbers would be good, but we can always count over and down to get to the right one ( where details are no longer seen ).

There can be issues with center resolution vs edge resolution. Rubber stamping the whole circle group center and 4 corners would cover that issue as well.

Since not everyone has 11x17 printers, maybe 8.5x11 would keep it consistent.

Just some ideas.
-Les

Hi Les,

I was digging around online for some sample charts that include circles, and I came across this PDF chart.

http://www.graphics.cornell.edu/~westin/misc/ISO_12233-reschart.pdf

It came from this site:
http://www.graphics.cornell.edu/~westin/misc/res-chart.html

It looks like we can use this one freely.
Let me know if this will work for us. If not, I can continue to work on something suited to our needs.

I printed this chart on 8.5x11in plain paper at 1200dpi, it should work if it's printed on photo quality paper, otherwise 11x17 would be fine at this resolution.

Bob

Good find. Always helps to google it ;)

OK, HVX200 people, print it and lets see what you got !
Maybe Mr Hurd can even put up a main thread under affordable HD/HDV to allow
even more users to visit his nice site to see quantifiable results !

Different people can post their results to respective camera threads.
Revolutionary. And Jpgs or LZ tiffs don't suck the bandwidth much either.
( Jpegs at 100% quality, of course.)

The first ever HD-HDV res shootout.
-Les

OMG!!!, I think something else hit the fan!

I just FAT-FINGERED a command that deleted my Generator program! So, don't try using it for now until I can figure out how to "undelete" it!

Looks like the HVX users will need to use the PDF chart listed in my previous post (#9) for now. HAVE AT IT GUYS! :-)

BoB

Raincheck! Special thanks to Dreamweaver's special disk cache, I got my PHP file back! Yay, I'm a happy camper. Alrighty, the web site link is working again!

Bob

Looks like this chart is going to be the winner for now.
http://www.graphics.cornell.edu/~westin/misc/ISO_12233-reschart.pdf

It goes all the way up to 2000 lines per picture height, so this will give us a good indication as to how much the HVX200 can resolve.

HVX200 owners: We'd love to see some unmodified frame grabs of this chart. It'd be best if you can print this on 11x17 paper at a high resolution. It'll be best if it's printed as "b/w only". If it comes out correctly, you should be able to easily discern between the black and white lines at the 20 mark.

QUESTION to anyone: What is the Leica's "sweet spot" on this camera? Whatever that is, we should target that setting when shooting this chart.

Thanks a bunch,
Bob

Please, someone shoot this puppy!

Hey Guys!

I rasterized this PDF chart at 4K (Film), 720p (HD), and 1080p (HD) using Photoshop so we can use this as a comparison for a true full resolution 1920x1080 HD "scan". The TIFFs are 16-bit LZW lossless compression.

1280x720 HD Resolution:
http://netnet.net/~rgraf/ISO_12233_720p.tif

1920x1080 HD resolution:
http://netnet.net/~rgraf/ISO_12233_1080p.tif

4096x2305 Film Resolution:
http://netnet.net/~rgraf/ISO_12233_4K.tif

For those that are interested in the settings for this:
Here was the procedure: I imported the PDF as 137.13pixels/inch, 16bit, Gray-scale, Bounding-box, Antialiased (non-antialiased made the image MUCH worse, so I enabled this) using Photoshop CS2. I cropped the image to the 16:9 marks and did *not* scale the image. The 4K image used the same settings, but I adjusted the DPI to make the cropped image the correct size. Ditto for the 720p.

I doubt that the HVX200 is going to be able to do better than these charts as it has more system components that contribute to a lossier image, as compared to the single anti-aliasing filter in Photoshop. Also consider that the HVX200 doesn't have 1920x1080 CCDs anyway. It'll be interesting to see how it matches up with the listed charts and charts shot with other cameras.

Let's have at it.
Bob

You want to shoot the highest res image you can print out. You don't tailor the test chart to the expected resolution you think you are shooting for. I looked at the 720p tiff, it is horribly aliased and barley suitable for shooting with an SD camera.
Again, print out the maximum quality that your printer can make. Use glossy paper if it's 8.5x11, it can hold more resolution. Let the camera see the best quality image, and let the camera throw away what it can't see.
That is how res testing works.
-Les

Les is correct. Don't print the charts that I downres'd, *use the PDF* instead.

I downres'd those examples as benchmarks for what we might expect to see as pseudo-best test results. I suspect we should find that actual frame grabs will reveal even less detail than the "best" case scenerio samples that I posted.

For example, in the 1920x1080 HD resolution image, notice how the lines break down in the 14 mark area. Because this is a 1920 rez image, it should be able to cope with more lines per picture height, but because of losses in PDF to image rasterization, and antialiasing, 14 is about as much as we can get.

DVCPRO-HD has a max recorded horizontal resolution of 1280, so we'd expect to see noticable gray values near the 12 mark, in an ideal world without more loss.

Bob