Saturday, January 29, 2011

The Secret Of Kells (updated twice)

I like flat-perspective, simple, decorated line art like this. I experimented with similar things back in the nineties, and I'll return to it.
You could say it's the opposite to the trend with more and more 3-dimensional art going on in recent years.

Much of the art of The Secret Of Kells is strongly reminiscent of Samurai Jack, they must share designers.
This film is amazing, I don't think I have ever seen so many outstanding compositions in one film. It's an art feast.






------

Here's an oddity: the pics I take of the TV with the Panasonic GF1 have interference patterns:


But those I take with a tiny Canon Ixus 860, don't:


The GF1 has a bigger sensor and a better lens, so this surprises me, I have no clue what makes for the difference. Shutter speed does not seem to be it.

Update 2:
Andreas said:
... if you shot both pictures with very similar framing the most likely reason simply is the different pixel count of the cameras (about 22% linear). Try framing the shot 20% tighter with the Ixus and you might get moiré as well. The GF1 on the other hand definitely shouldn't show moiré from 20% further away. For moiré to occur the grids of subject and sensor have to match; due to the different pixel pitch of both cameras you'll see moiré for different subject grid frequencies.

You are right. Watch these down-scaled but un-cropped photos with the GF1 at two distances:


16 comments:

Philocalist said...

Weird: the usual suspect IS shutter speed, though you would expect the lines to be horizontal (unless newer screens are now scanning side to side, rather than top to bottom?)
The magic figure was a 30th of a second: a shutter speed of this or faster gives banding, a slower speed avoids it ... its a matter of the shutter being open long enough to allow the TV screen to scan across at least one full time.
Maybe the Canon in auto mode is setting a slower shutter speed?

Eolake Stobblehouse said...

I think it is not a scanning phenomenon, but probably an interference pattern between the pixels of the TV and the pixels of the imaging sensor of the camera.

Philocalist said...

Never had THAT problem with film :-)

Andreas Weber said...

> The GF1 has a bigger sensor and a better lens, so this surprises me, I have no clue what makes for the difference.

Actually, the smaller sensor and softer lens might be the very reason why the Ixus doesn't show it. Try throwing the GF1 slightly out of focus. ;-)
But if you shot both pictures with very similar framing the most likely reason simply is the different pixel count of the cameras (about 22% linear). Try framing the shot 20% tighter with the Ixus and you might get moiré as well. The GF1 on the other hand definitely shouldn't show moiré from 20% further away. For moiré to occur the grids of subject and sensor have to match; due to the different pixel pitch of both cameras you'll see moiré for different subject grid frequencies.

Kentg said...

It seems to be something separate from the image because the image is not distorted. I suspect it is something to do with the screen surface. Try taking the image somewhat off the perpendicular and see if the pattern changes.

Bru said...

On mu phone's camera there is a setting that may have something to do with this. I can set antibanding for off, 50hz, 60hz, or auto.

Alex said...

The 50/60Hz is for artificial lighting. If you are in say the US, they have 60Hz mains electricity. In the UK it's 50Hz. Video can be adjusted so that the recorded frame rate is in phase with lighting so you get even lighting frame to frame.

On some cameras with simplified user interfaces it's embedded in the locale setting. eg English (UK) not only gives you all the missing "u"s back, but also would set you to 50Hz, and English(US) puts your dates in that crazy month day year format and your video aligned to 60Hz.

I still think it looks more like an interference pattern between sensor array and screen pixel.

Eolake Stobblehouse said...

Adreas, good call, it seems to be subject very much to distance and resolution matching or not. Post updated with new pics showing it.

Jeff R. said...

OK, allow me to throw my spanner in the works.

Here is a snap I took a few months ago with my iPhone.

http://www.mendosus.com/jpg/img_0069.jpg

The subject matter is irrelevant - it's a frozen food product in a supermarket freezer - taken through the (vertical) glass window of the freezer cabinet.

The interference pattern is *not* a reflection of the fluorescent lighting - you can see they're at 90 deg to the pattern. The pattern moved and "swirled" in the "viewfinder" of the iPhone, but could not be minimised.

Please mote that this is *not* a photo of a matrix of pixels (e.g. a monitor) - it's Real Life - albeit through a glass door.

I've left the image exactly as taken (needs rotating) and the EXIF is included intact.

Now... why would an iPhone generate an interference pattern like this *without* the benefit of a grid pattern to cause moire patterns? Obviously the fluoro lights - with their 50Hz cycle - but how?

Beats me.

Alex said...

This must be the "rolling shutter" effect I've hear of.

The lines manifest themselves as vertical because of the camera being held 90deg from the sensors upright position.

Indoor no-flash, longer exposure. I'm not sure how the sensor array is captured, but there is possibly one set of ADC per several pixel rows. In the time it takes to sample each row in turn the ambient light has changed, and yielded this result.

I can ask the guys at work if they know what's going on.

Eolake Stobblehouse said...

I have seen some iPhone photos of propellers on airplanes which were profoundly weird because of a shutter effect. It might be related.

Aha:

http://thurly.net/0rok

Alex said...

I just got a reply back from one of the sensor experts (video cameras) at work.

The florescent light that is illuminating the scene is probably very bright, so the rolling shutter is a small window from top to bottom of the image. The frequency of the ballast in the florescent light is much higher frequency than the exposure time of the whole scene but “less” time than the rolling shutter window. So you see the oscillations in the lighting of the scene. Your eye can never see this. It is probably around 120 to 240 Hz.

I keep forgetting that "high speed ballasts" have been added to flourescents. The flicker rate of filament bulbs is in time with line frequency. As for LED, it depends on the power supply (line freq->rectifier +ve only AC->DC->PWM or current limiter) which can be in the kHz

Eolake Stobblehouse said...

What is "ballast" in this context?

Alex said...

Ballast is effectively a transformer, but may also be a capacitor too. In Britain we used to call them "starter motors" but the term is actually "starter".

http://en.wikipedia.org/wiki/Electrical_ballast

Oh, I see, they can also have inductors and resistors in them too.

Alex said...

Electromagnetic ballasts may also cause problems for video recording as there can be a "beat effect" between the periodic reading of a camera's sensor and the fluctuations in intensity of the fluorescent lamp.

From http://en.wikipedia.org/wiki/Fluorescent_lamp#Electrical_aspects_of_operation

Jeff R. said...

Hey, thanks guys!

That makes sense.

...and that propellor photo is wild!