As a bit more detail, because it is probably relevant. I am using the latest Angstrom release as of 2 days ago. All the demonstrations seem to use the Cheese webcam application to leverage the camera cape. This produces a garbage image from the outset and has never worked for the cape. It does work fine, however, for a usb webcam. Go figure.
I am using a program from Tobias Müller (http://www.twam.info/linux/v4l2grab-grabbing-jpegs-from-v4l2-devices) that reads the camera device leveraging Video4Linux. Works fine with the LCD and produces a strangely low resolution, poorly colored image without the LCD in the stack.
Thanx – Ken
one more question:
did you capture the pictures just through the console or terminal (by a minicom connection) or did you use something like a vncviewer?
Thanks for following up Alex. Two different ways. One is using mplayer (mplayer -vo png -frames 7 tv://), and the other using the program v4l2grab (./v4l2grab -o image.jpg) from Tobias that I mentioned. Both read the camera and then store a PNG or a JPG to the filesystem. Both produce the similar results as each other when using the BB/Camera stack (bad image) and using the BB/Camera/LCD stack (good image). [I am capturing 7 frames with mplayer because it seems to take a couple of frames to settle the autogain. By 7 frames the image is stable.]
Also, Cheese is working, sort of. I had initially left the resolution at full res and Cheese failed to produce output. If the resolution is set to 1024x768 or less, it works. Laggy, but it works. I also get out of memory errors if we try to capture full resolution from the command line, but that issue is moot if the image is distorted. This is probably why Cheese fails at higher resolutions. First things first-- good images, then memory management.