Lytro REVOLUTIONARY Optics a MATCH MADE I HEAVEN for CCTV?

[HiroPro 0]

Hey guys checkout these AMAZING cameras.

 

http://www.lytro.com/

 

That technology would just be INCREDIBLE for CCTV cameras.

 

What do you experts think?

 

Can you imagine what technology exclusives for a CCTV cam manufacturer would do for said company!

 

It really is revolutionary technology.

 

Here is the CEO's dissertation.

 

https://www.lytro.com/renng-thesis.pdf

 

REVOLUTIONARY

Edited May 16, 2012 by Guest

[dustmop 0]

My immediate thought is that wherever you would put it, you're going to need some great lighting. Face shots into places with large lines/queues (subways, shopping malls, casinos, amusement parks) would be neat. Not sure where else it would be truly useful, given that lighting really does matter to this type of device.

 

Also, I bet the recorded stream would be ENORMOUS compared to a typical h.264 stream of a similar resolution. Not to mention the processing power to refocus video in playback as opposed to refocusing just a single frame.

[HiroPro 0]

I think the data for that is stored in meta bro... (it's NOT a series of images at different focal points)

 

That means it's simply ASCII text data dude...

 

They key to processing is the algorithm and the optics to capture the data. The extra data I'm most certain is meta and stored in the EXIF.

 

For sure for analytics it would take a lot more processing power.

 

Still... in high light areas like you're saying it would work wonders.

 

PS I asked here if there was an industry standard for cameras and this kinda thing. An API for IP cam trigger stuff and access for parameter changes etc. Not one dude responded ehh...

 

It's called ONVIF BTW and the ver. 2.0 standards including event handling and analytics meta standards.

 

I don't get why nobody responded to tell me about this when I queried.

[HiroPro 0]

hahah

 

What a great marketing video

 

7babcK2GH3I

[dustmop 0]

In a dream world, I'd LOVE to have a 360 degree recording camera that can do this. I'd like SEVERAL of them, actually. Sadly, that day is likely still 5 to 10 years away.

 

According to this http://support.lytro.com/entries/20718762-what-file-formats-does-the-lytro-support:

 

The light field data, including 3D info about the color, luminosity and direction of 11 million rays, is stored in a format developed by Lytro called the LFP (Light Field Picture) format. When you move your "living pictures" to your computer via the Lytro Desktop software, each LFP file is about 16MB.

 

They don't list a true resolution; they say it's "11 megarays", whatever that means. (You can currently export images as JPEG, at 1,080px X 1,080px) They say it's 16MB for a single image. Given the sheer amount of data they do store in each image, using I/P frame differences like h.264 does is either highly unlikely, or insanely processor intensive.

 

Estimated sizes if we assume NO compression:

1 image: 16MB

1 second: 480MB

1 minute: 28.8GB

1 hour: 1.65TB

1 day: 39.6 TB

 

If they can swing something akin to h.264, then the numbers would obviously change to something a bit more reasonable.

 

Just checked my recording, and an entire day of analog h.264 (1.5Mbps) is between 15GB and 16GB.

 

The biggest hurdle will be price. Digging around for some Raytrix camera pricing (the people that really started the whole lightfield thing), they sell their R5 camera for just under €3000. That camera can do 30fps over GigE, or 90fps over USB3. Aperture is lousy, so lighting matters. But the max 2d image size? 1 megapixel (source: http://www.raytrix.de/tl_files/downloads/R5.pdf). Also, the Raytrix cameras require CUDA GPUs to perform a lot of their processing it seems.

[Birdman Adam 0]

Looks pretty stupid to me. I know very well how to use my nice DSLR, and how to achieve the same effect of that camera, except better with my knowledge of how to change it.

 

Maybe its cool for the average consumer, but I would much rather choose my stops, shutter, etc for myself.

 

I do wonder how it works - how can we record light data at an infinite range of focus? I bet the image files are like 1GB each if one focus level is equal to the one-focus-dimension volume of my DSLR.

[HiroPro 0]

In a dream world, I'd LOVE to have a 360 degree recording camera that can do this. I'd like SEVERAL of them, actually. Sadly, that day is likely still 5 to 10 years away.

 

According to this http://support.lytro.com/entries/20718762-what-file-formats-does-the-lytro-support:

 

The light field data, including 3D info about the color, luminosity and direction of 11 million rays, is stored in a format developed by Lytro called the LFP (Light Field Picture) format. When you move your "living pictures" to your computer via the Lytro Desktop software, each LFP file is about 16MB.

 

They don't list a true resolution; they say it's "11 megarays", whatever that means. (You can currently export images as JPEG, at 1,080px X 1,080px) They say it's 16MB for a single image. Given the sheer amount of data they do store in each image, using I/P frame differences like h.264 does is either highly unlikely, or insanely processor intensive.

 

Estimated sizes if we assume NO compression:

1 image: 16MB

1 second: 480MB

1 minute: 28.8GB

1 hour: 1.65TB

1 day: 39.6 TB

 

If they can swing something akin to h.264, then the numbers would obviously change to something a bit more reasonable.

 

Just checked my recording, and an entire day of analog h.264 (1.5Mbps) is between 15GB and 16GB.

 

The biggest hurdle will be price. Digging around for some Raytrix camera pricing (the people that really started the whole lightfield thing), they sell their R5 camera for just under €3000. That camera can do 30fps over GigE, or 90fps over USB3. Aperture is lousy, so lighting matters. But the max 2d image size? 1 megapixel (source: http://www.raytrix.de/tl_files/downloads/R5.pdf). Also, the Raytrix cameras require CUDA GPUs to perform a lot of their processing it seems.

 

 

Very interesting stuff dustmop and thanks for sharing.

[MaxIcon 0]

Here's a bit of background on this fascinating technology, and why it's currently useless for video.

 

The Lytro uses micro-lenses to capture the extra data on the sensor elements. As a result, the image resolution at any given focal point equivalent is much lower than the sensor resolution, which is why they're cagey about resolution numbers.

 

It also limits the low-light sensitivity, which is pretty important for CCTV. Finally, for a given pixel-equivalent resolution, it requires much higher processing and data bandwidth, since each viewed pixel is accompanied by multiple pixels' worth of data. This all takes time, as well.

 

As far as I know, there are no video capabilities, and no plans for any in the near future. Reviews show that it takes a picture every 1.3 seconds, and processing and downloading each picture to a PC takes over a minute. No chance of h.264 for this puppy with the current state of the art!

 

I haven't been able to find any specs for the shutter speed, and I'm guessing that it's because it's slow. All the demos I've seen have been of static shots rather than action shots. I could be wrong, but specs that aren't shown often tell as much as the ones that are.

 

Here's the technical overview of the science behind it, for the tech geeks:

http://www.lytro.com/renng-thesis.pdf

 

It's definitely not real-time video ready at this point. Whether it will ever be will be clear in the next few years.

 

It's a very cool concept, but not all cool concepts make it to street-level technology.

[Soundy 1]

Here's a bit of background on this fascinating technology, and why it's currently useless for video.

Max, nice breakdown.

 

It's a very cool concept, but not all cool concepts make it to street-level technology.

Agreed... revolutionary, maybe... match made in heaven for CCTV? Not even close.