2009-11-29 10:15:07     Analyse Video Image - appreciate some pointers!

Document created by Aaronwu Employee on Aug 19, 2013
Version 1Show Document
  • View in full screen mode

2009-11-29 10:15:07     Analyse Video Image - appreciate some pointers!

Owen OH (IRELAND)

Message: 82860   

 

Hello All,

 

I'm hoping to do the following:

 

    Capture an image from an ADV7184 (video decoder)

    Transfer it to memory via the PPI interface

    Analyse the video image

    Repeat

 

I'm aware that V4L seems the way to go to setup the AD7184 but unsure of

 

    How & where to transfer the image to memory.

    Is ffmpeg the way to go?

    Should I simply save the image to a file & then analyse that?

    I've tried building ffmpeg into the kernel but it seems to occupy around 1->2 MB alone. Is it possible to reduce it's size?

 

Any input would gratefully be appreciated,

 

Owen

QuoteReplyEditDelete

 

 

2009-11-29 10:37:06     Re: Analyse Video Image - appreciate some pointers!

Mike Frysinger (UNITED STATES)

Message: 82861   

 

the V4L driver takes care of the sensor and PPI interface.  once those are in place, you tell them where to transfer the frames.

 

whether you analyze things in memory or save to a file first is up to you and what you need to do.

 

if you want ffmpeg trimmed down to the bare min, you'll have to configure and link it yourself.

QuoteReplyEditDelete

 

 

2009-11-29 10:54:06     Re: Analyse Video Image - appreciate some pointers!

Owen OH (IRELAND)

Message: 82862   

 

Mike,

 

Thanks for the quick reply. I've looked through the V4L documentation (www.linuxtv.org) & haven't been able to find commands to control the sensor or placement of images. Do you know where I could find them?

 

Thanks,

 

Owen

QuoteReplyEditDelete

 

 

2009-11-29 11:04:00     Re: Analyse Video Image - appreciate some pointers!

Mike Frysinger (UNITED STATES)

Message: 82863   

 

a driver needs to be written first for the sensor.

 

V4L requires dedicated buffers in memory for capturing frames.  the specific sensor can take care of the allocation or let the common V4L code do it.

QuoteReplyEditDelete

 

 

2009-12-02 08:00:12     Re: Analyse Video Image - appreciate some pointers!

Owen OH (IRELAND)

Message: 82975   

 

Mike,

 

Thanks again. I found an API reference:

 

  v4l2spec.bytesex.org/v4l2spec/v4l2.pdf

 

Is this the best way to go? I'v done some kernel driver work before & I have the ADV7183 driver as a template so I hope it won't be too difficult. Just trying to get my head around how to get the images into memory!

 

Owen

QuoteReplyEditDelete

 

 

2009-12-02 14:08:03     Re: Analyse Video Image - appreciate some pointers!

Mike Frysinger (UNITED STATES)

Message: 82981   

 

that looks reasonably up-to-date

 

V4L generally works on frames, so userspace would open a capture device, configure the parameters it wants via the V4L ioctls, use the V4L ioctls to get the memory addresses the driver is using, and then use the V4L ioctls to initiate a capture.

QuoteReplyEditDelete

 

 

2009-12-04 12:22:23     Re: Analyse Video Image - appreciate some pointers!

Owen OH (IRELAND)

Message: 83094   

 

Thanks,

 

Started churning through the manual.

 

 

 

Owen

Attachments

    Outcomes