Re: Camera preview, thin lines in the frames
- Date: Thu, 12 Nov 2009 17:03:42 +0100
- From: Ryan Raasch <ryan.raasch@xxxxxxxxx>
- Subject: Re: Camera preview, thin lines in the frames
Shun-Yu Chang wrote:
On Thu, Nov 12, 2009 at 9:49 PM, Ryan Raasch <ryan.raasch@xxxxxxxxx
Shun-Yu Chang wrote:
I am new to v4l2. I am working on integrating a usb camera
the beagleboard(Omap3530 dev board).
Now I met a camera preview issue that is there are thin lines
in the frames.
I still have no idea how to describe this exactly. It's like
I have two guesses. One is the jpeg compression ,or is it jpeg? Try
saving in raw RGB or Ycbcr format and viewing with imagemagick.
The raw data I got is yuyv422. I used the mmap way. I tried to
convert raw data to yuv420 and jpg files both. I used ffmpeg to convert
yuv420 files to jpg files and I can still see the lines.
I don't know how to use imagemagick to see yuv files...
Does this happen in live preview? Maybe write data from camera
directly to screen to see if happening there.
It happens in live preview. So I save the frames directly into
files to verify.
But if you save the files, you don't know if the compression is the
problem. You need to visually look on the screen with a magnifying glass
The second would be the FIFO (hadn't worked with omap before) levels
of the data lines to the processor in the kernel.
Look at the system processor (top) usage to see how much cpu% is
being used. USB has high overhead.
the top command shows
PID CPU% S #THR VSS RSS UID Name
1012 56% R 1 55000K 54468K root capture_test
241 40% R 1 0K 0K root pdflush
[....Omit, others are 0%. .....]
After grabing the first 100 frames to files, capture_test exits.
Looks pretty busy to me. Maybe you could kill the pdflush, don't save to
files, and look at the display with a magnifying glass.
I don't quite understand about FIFO levels of the data lines part.
Could give an instruction where I can start to look into?
It probrably has dma priorities. So if other drivers in the system
(network, usb, display, etc.) use a lot of cpu time, the camera dma
might be delayed.
Are you saving the files over network (nfs)?
Thanks for your reply..
I modified capture.c sample to save the frames to picture
files. So in
my guess, the problem is not in userspace. And this is not
happen on my
laptop with the usb camera.
Could anybody give me a clue ? Any one would be thankful.
video4linux-list mailing list