problem about the camera image is reddish

HI,nvidia :

We have a custom sensor(a mipi raw10 rggb sensor), and now it can preview and capture use argus_camera.

but the color of the image is reddish,the driver and dtsi we used is based on ov5693.

Can you give some suggestions, thank you!


please refer the attached jpg picture captrured with argus_camera. the real color is white ,but the previewed color in image0000.jpg and image0003.jpg is the more red or more blue (it looks dipends on the brightness of the light)

hello jll2020,

  1. please share the sensor format dump with below commands,
$ v4l2-ctl -d /dev/video0 --list-formats-ext
  1. please refer to Camera Architecture Stack, camera launching with argus_camera would also involve tuning block.
    there’s badge property to match the corresponding the tuning files.
    please review your sensor device tree, you should modify badge property as unique naming.
    for example,
<i>$l4t-r32.2/kernel_src/hardware/nvidia/platform/t18x/common/kernel-dts/t18x-common-modules/tegra186-camera-e3326-a00.dtsi</i>

                    modules {
                        module0 {
<b>                                badge = "e3326_front_P5V27C";</b>
                                position = "rear";
                                orientation = "1";
                        };
  1. there’s an alternative way to access camera sensors, please dump the raw files with v4l2 standard controls,
    for example,
$ v4l2-ctl -d /dev/video0 --set-fmt-video=width=1920,height=1080,pixelformat=RG10 --set-ctrl bypass_mode=0 --stream-mmap --stream-count=5 --stream-to=test.raw

then, you could use the 3rdparty tools, such as 7yuv to analysis the raw files.
thanks

HI,JerryChang:
Thanks for your quick reply.

jin@jin-desktop:~$ v4l2-ctl -d /dev/video0 --list-formats-ext
ioctl: VIDIOC_ENUM_FMT
Index : 0
Type : Video Capture
Pixel Format: ‘RG10’
Name : 10-bit Bayer RGRG/GBGB
Size: Discrete 1920x1080
Interval: Discrete 0.018s (55.000 fps)

2.I have modified the badge property as unique naming like
badge = “cmv2000_front_S156CM”;
but did’t get any improvement ,the color is also not right

3.dump 5 frame and use 7yuv to analysis the raw files
I did’t see any problem, the raw file I dumped please see the attated file
test_raw.rar (6.62 MB)
testraw.rar (1.27 MB)

hello jll2020,

suspect there’s something wrong with ISP post-processing.

  1. may I know did you have any parameter settings to capture reddish images through argus_camera?
    for example, noise reduction setting; edge enhancement setting…etc.

  2. had you perform image tuning process? could you please have a try to disable chroma and luma settings of noiseReduction function for testing.
    thanks

Do you have the correct pixel phase for the CFA on your image sensor? Or, perhaps are you windowing/cropping lines that might accidentally result in a change in the pixel phase?

Hello,

Check that your pixel phase is properly set, as Greg mentioned.

You can also check the color gain, it may be affecting the final image.

Using v4l2-ctl you can also modify some camera properties to improve the image.

Example:

v4l2-ctl -d /dev/video0 --set-fmt-video=width=1280,height=720,pixelformat=RG12 --set-ctrl <b>bypass_mode=0,exposure=2000</b>

The available controls for your camera depend on the driver, please run the following command to check available controls and valid ranges:

v4l2-ctl -d /dev/<YOUR_DEVICE> --all --verbose

Regards,
Greivin F.

Hi JerryChang:

I found before flash the dtb file which changed the badge string , it must delete the bin file (nvcam_cache*.bin)
in the /var/nvidia/nvcam/settings/ directory , otherwise the isp setting seems would not change.

Now, I use the e3326_front_P5V27C badge . but it color is not good yet, the image color is more red when the light is more brighter.
But when I change the YUV format option in argus_camera from ‘nv12’ to ‘p016’ , the image color is quite normal

My question is:

  1. Why YUV format frome ‘nv12’ to ‘p016’ will effect the color ?

  2. We now using the default setting on argus_camera, and we also have tried other setting on argus_camera but did not have goot result, so is there any other way we can do the improve the image color ?

attached file is the dtsi file I’m useing now,
I have comment tegra186-quill-camera-plugin-manager.dtsi and tegra186-quill-camera-modules.dtsi
and include camera-dummy.dtsi

camera-dummy.txt (15.7 KB)

hello jll2020,

you should not use the badge name as e3326_front_P5V27C, since you’re develop your custom sensor.
could you please modify that and also delete those cache bin files for verification.
thanks

HI,JerryChang:

I modified the badge to a unique name and delete those cache bin files

but the image is Fogy and reddish ,is there any other way I can improve the image ?

Is this an HDR sensor?

Hi D3_growe:

the sensor is an lvds sensor and we get the image from a fgpa conversion.

the sensor have HDR mode, please refer the attached datasheet file.
CMV2000_DS000434_1-00.pdf (3.17 MB)

HI JerryChang:

I have another question .
we have two same sensor on our board, one is one serial_a(4lane),another is on serial_c(use v4l2-ctl can get both raw data from the two sensor)
When I use other badge like a unique badge , I can only see Device ID 0 in argus camera

When I use badge name e3326_front_P5V27C, I can see Device ID 0 and 1, but if change to Device 1 the image is device 0
also I runed the muti senssion mode I can see two image ,But the the two image are from serial_a

that means I am not able to make sensor on serial_c work or make two sensor work together.

can you give some advise ?

hello jll2020,

by checking your images in comment #10, you color divergence issue seems related to HDR results.

you might have more tests to narrow down the issue,
could you please using this setup to take another images with color chart for checking.
is it possible to disable HDR mode for image captures?
also, may I know which approach of your sensor configuration you’re used to acquire high dynamic range images?

BTW,
your question in comment #13 was driver issues.
suggest you open another new discussion thread, sharing device tree configurations for better tracking.
thanks

Hi JerryChang

Thank you for your quick reply

I’ll check the sensor configuration and do more test

about comment #13 ,I have opened a new discussion

HI JerryChang:

We used v4l2-ctl -d /dev/video0 --set-fmt-video=width=1920,height=1080,pixelformat=RG10 --stream-mmap --stream-count=1 --stream-to=camera.raw
captured a raw image ,then We used a third part tool(FastRawViewer) and tune the white balance para,like color temperature and tint, the we can get a quite good image.

It seems the problem comes from the ISP, is there any way to adjust the para of ISP ?

hello jll2020,

we don’t support tuning process through forum discussion thread, please contact with Jetson Preferred Partners for further support, thanks