Libargus with sony IMX camera binning

Autonomous Machines Jetson & Embedded Systems Jetson Nano
#1

Hi,
This looks like the generic problem of IMX image sensor more than v4l2/argus stack. But I hope you can share your thought/idea. Thank you.
I’m working with IMX485 Sony image sensor and Jetson Nano. The image sensor support binning mode, but the way it support doesn’t make any sense for picture’s brightness.
In binning mode, 4 10bits pixels will be summed into one 12bits pixel, and output is 12bits image. And libargus process this image as a normal 12bits image. So the advantage of binning image regarding low light condition is lost here.

My question is, is there any mechanisms to pre-process the image before libargus starts ISP processing? In this case, multiple the input image with 4 should be used.

#3

hello nvl1109,

may I know what’s the scenario and also what’s your use-case,
the sensor configuration is by default to readout all active pixels and sending to ISP,
since you would like to active sensor’s capability to perform binning mode. you may also contact with your sensor vendor for sensor driver implementation.
thanks

#4

Hi JerryChang,
My scenario is to get camera work in low-light condition. As you known, binning mode will give more bright output compare to full resolution.
The binning mode is already supported in v4l2 driver. But the problem is maybe belong to the sensor.
Instead of giving sum of 4 12bits pixels as output 12bits pixel (set to max 0xfff if over boundary), it gives sum of 4 10bits pixels as a 12bits pixel output. So the ISP treats binning output as a no binning 12bits image. Then no brightness advantage here.

#5

hello nvl1109,

it’s sensor driver to handle the binning process and output 12-bit data to ISP.
please examine your sensor binning output of the v4l2 capture results; please also contact with Jetson Preferred Partners for your camera solutnios.
thanks

1 Like
#6

Thank you JerryChang. I’m contacting with FRAMOS. But beside that channel, I would like to know if there is another way to hook into the pipeline between RAW value from v4l2 and libargus ISP. Before ISP processing, I want to multiple the RAW image with 4 on all pixels, cut out the over range value.
Do you think that is possible?

#7

hello nvl1109,

you may also refer to Applications Using V4L2 IOCTL Directly to use V4L2 IOCTL to verify basic functionality during sensor bring-up.
in theory, the raw file you’d dump by v4l2 is the input of ISP.

so,
you should check with the v4l2 raw files of sensor’s binning results.
thanks

#8

Yes, I did check the result of v4l2 driver. That matches my explanation above regarding binning output. Sensor bring-up is correct compare to datasheet.
The incorrect here is only the way the binned image is processed in sensor image, so have to adjust the image data before ISP processing.

#9

hello nvl1109,

it seems there’re still some works need to be done in sensor driver,
please contact with Jetson Preferred Partners for more details.
thanks

1 Like