I’m running on a Jetson Orin (not a Nano and not a Super). Running JetPack R63.
I’m trying to capture an image using only v4l2 utilities. But I’m unable to view the image correctly. The image looks fine if I use the nvarguscamerasrc and convert to RGB.
Here are the formats for the camera:
$ v4l2-ctl --list-formats-ext
ioctl: VIDIOC_ENUM_FMT
Type: Video Capture
[0]: 'RG10' (10-bit Bayer RGRG/GBGB)
Size: Discrete 3840x2160
Interval: Discrete 0.033s (30.000 fps)
Size: Discrete 1920x1080
Interval: Discrete 0.017s (60.000 fps)
I’m using v4l2-ctl to capture the frame. I’m skipping the first 10 seconds of data (300 frames) and outputting the following frame to disk.
$ v4l2-ctl -d /dev/video0 --set-fmt-video=width=3840,height=2160,pixelformat=RG10 --stream-mmap --stream-to ctl-frame.bin --stream-count=1 --stream-skip=300
If I then try to render that image using the following in python, I get an image that is very gray and dull, where bright spots are purple.
import numpy as np
import cv2
from pathlib import Path
def main():
file_path = Path("~/data/inspect_raw_frame/frame.bin")
num_frames = 1
with file_path.expanduser().open("rb") as f:
data = np.fromfile(f, dtype=np.uint16).reshape((num_frames, 2160, 3840))
data = data[-1,::-1,:]
# Remove the 4 duplicated MSBs
data = data >> 4
# Truncate the lower 2 LSBs to convert from UInt10 to UInt8
data = data >> 2
data = data.astype(np.uint8)
cdata = cv2.cvtColor(data, cv2.COLOR_BayerBGGR2BGR)
cdata = cv2.resize(cdata, (1920, 1080))
cv2.imshow("image", cdata)
cv2.waitKey(0)
cv2.destroyAllWindows()
if __name__ == "__main__":
main()
I based the Python code off of this post. However, something is still not quite right.
Is that post accurate?
I attempted to validate empirically. The 2 MSBs are indeed repeated in the first 2 bits for this frame. However, the upper most bits of the 16 bits are not zero which seems odd.
I checked this with the following code:
import numpy as np
from pathlib import Path
def main():
file_path = Path("~/data/inspect_raw_frame/frame.bin")
num_frames = 1
with file_path.expanduser().open("rb") as f:
data = np.fromfile(f, dtype=np.uint16).reshape((num_frames, 2160, 3840))
num_equal = 0
num_zeroed = 0
for value in data.flatten().tolist():
binary = f"{value:016b}"
if binary[:2] == "00":
num_zeroed += 1
if binary[2:6] == binary[-4:]:
num_equal += 1
print(np.size(data), num_equal, num_zeroed)
if __name__ == "__main__":
main()
And here’s the output from running it:
8294400 8294400 8266286
Here’s a binary dump showing that indeed the highest 2 bits are not zeroed out.
$ xxd -b -l 0x1ff1e8 frame.bin
001ff1e2: 10010011 01001110 01010001 01000100 11010110 01011010 .NQD.Z
^ ^ ^
Am I misunderstanding the RG10 format? Any idea why my image does not look correct?
