Working with raw image data

Capturing and downloading

Command line, demo script: raw.py

• run:

root@elphel393:~# raw.py

• description:

1. scans all available ports, determines the trigger master
2. reads current frame_number to be used later
3. formats system memory buffers and mmaps to user memory (all available ports)
4. selects position in video memory (0 or 1)
5. triggers copying pixel data from video memory to system memory for the frame_number+1 (waiting is implemented in the videomem driver)
6. prints the first 16 bytes of the image
7. prints raw image info for each port

Command line

• Available options (channels can be 0..3):

# set receiving buffer size in 4kB pages (thus 4096 equals to 16MB)
echo {size} > /sys/devices/soc0/elphel393-mem\@0/buffer_pages_raw_chn2
# start from the N-th frame position in buffer. Currently, 0 or 1
echo 1 > /sys/devices/soc0/elphel393-videomem\@0/video_frame_number
# initiate transfer, wait for X-th frame (X - absolute frame number), 0 = no waiting
echo X > /sys/devices/soc0/elphel393-videomem\@0/membridge_start2
# copy to file
cat /dev/image_raw2 > /tmp/test.raw
# see in hex
hexdump /tmp/test.raw
# copy to PC
scp /tmp/test.raw user@address:/path/

• Minimal:

echo 4096 > /sys/devices/soc0/elphel393-mem\@0/buffer_pages_raw_chn2
echo 0 > /sys/devices/soc0/elphel393-videomem\@0/membridge_start2
cat /dev/image_raw2 > /tmp/test.raw

Displaying

ImageMagick

• Camera settings: 2592x1936 JPEG, actual valid pixels - 2596x1940, pixels read - 2608x1940:

convert -size 2608x1940 -depth 8 -endian MSB -normalize gray:test.raw -compress lzw result.tiff

• Camera settings: 2592x1936 JP4, actual pixels available: 2592x1936

convert -size 2592x1936 -depth 8 -endian MSB -normalize gray:test.raw -compress lzw result.tiff

The result is a grayscale tiff

bayer2rgb program

• Available here.
• Camera settings: 2592x1936 JPEG:

bayer2rgb -w 2608 -v 1940 -b 8 -f GRBG -m AHD -t -i test.raw -o test.tiff

The result is a colored tiff image. Pixel values are linear.

Python

• Camera settings: 2592x1936 JPEG:
  • a)

#!/usr/bin/env python
import cv2
import numpy as np

width = 2608
height = 1940

with open("test.raw", "rb") as rawimg:
   img = np.fromfile(rawimg, np.dtype('u1'), width * height).reshape(height, width)
   img.tofile("test2.raw")
   colimg = cv2.cvtColor(img, cv2.COLOR_BAYER_GB2BGR)
   cv2.imwrite("test.jpeg", colimg)
   #cv2.imshow("color", colimg)
   #cv2.waitKey(0)

• • b) for PC:

#!/usr/bin/python3
from PIL import Image
rawData = open("test.raw", 'rb').read()
imgSize = (2608,1940)# the image size
img = Image.frombytes('L', imgSize, rawData)
img.save("result.jpeg")# can give any format you like .png

The result will be a grayscale image.

Processing

Open the file and process as pixel array RGGB, GRBG, GBRG or BGGR - 8 or 16 bits