Video ADC from NTSC Composite RCA Signal

Pheal

Hi guys, I am looking for a video capture device. The ones that are for sale are "closed source" as in I can't modify anything from them or have access to the schematics to make my own. I know I will not really be able to perform a video capture from an NTSC signal without proper compression as the bandwidth required will be too high. I have been told by one of my friends that I will also not be able to test Video ADC chips on a breadboard since the frequency will be too high for most oscillators required and that will create noise because of the breadboard layout.

All the projects I've seen to far uses cameras and don't seem to get good "human readable" resolutions (like 720p coming from a NTSC Composite RCA cable, hell it is probably sent to a lower resolution as I found out might be 480i or 576i, but that is also ok).

I know video ADC and Computer Vision is the kind of stuff that is done as final a project, hopefully I will be able to overcome the lack of knowledge by trying hard enough and learning the bits one by one. I am mainly a programmer looking to learn the hardware side of things, so my electronics skills can only improve at this point haha.

I will try to explain what I am trying to do a bit clearer:

NTSC Composite Signal > Goes to Video Capture / Decoder Board > Builds an image on board > Processes the images with algorithms on board to reduce the bandwidth required to send to PC afterwards by USB (kind of dumbs the image down with an algo I made) > PC does another round of processing on the image

This is VERY close to what I am trying to achieve but it uses a camera instead of a NTSC Composite RCA signal and is slower:

http://forums.parallax.com/showthread.p ... NTSC-Video

How easier does it get if I only need a B/W image instead of color? How would I go to have a higher resolution image and output speed out of a similar circuit?

I don't need a complete solution as this topic is kind of complicated and I know I have lots of learning and searching to do on this, but pointing in the right direction always help .

Thanks for all the input in advance. I will post back when I get more stuff figured out and will keep looking here if anyone has other awesome knowledge to share.

PS: I had already looked at propeller chips before but I tought I had to program using their language... Apparently there are tools to stay in C, which makes them way more appealing now, hence me posting here.

Mike Green

This sort of thing is not really feasible for the Propeller except for a very low resolution B&W image. There's not enough memory available and the Propeller is simply not fast enough unless you do all the code in assembly language and use multiple cogs processing parts of the image in parallel. The C tools mostly make use of a special high-speed interpretive code that's similar to assembly language, but several times slower. This interpretive code is still faster than Spin, so it's useful, but not for the sort of video processing you're talking about. Running large C programs requires using external memory (to the Propeller) which slows things down further although allows larger programs and data areas.

CMUcam4 is one example of video processing with the Propeller, but it uses low resolution images and limited types of processing.

ericball

It used to be for basic video capture an old Pinnacle PCTV Pro PCI board (or similar) was the way to go, typically using HuffYUV for lossless realtime compression, although I'd recommend using FFV1 these days. There's also USB capture devices which can handle uncompressed 640x480.

But the Propeller simply doesn't have the ADC, CPU, RAM & I/O to handle video capture.