peteasa
commented
3 years ago image file is here img.zip
Closed peteasa closed 3 years ago
letmaik
commented
3 years ago What camera is the file from?
peteasa
commented
3 years ago ON Semiconductor - Sensor Part Number = AR1335
I found that I could read and write the file directly with numpy so I am not now using rawpy.. suspect I am loosing some file header information along the way!
def read_write_raw(): with open('template.raw', 'r') as infile: data = np.fromfile(infile, dtype=np.uint16) pixels, = data.shape height = int(3136) width = int(pixels / height) data = np.reshape(data, (height, width)) im = Image.fromarray(data) im.save('template1.png') data.astype(np.uint16).tofile('template1.raw')
kmilos
commented
3 years ago Yep, this is just a really raw bytestream dump straight out of the sensor (i.e. not a raw file format produced by more sophisticated camera software), so reading it in using numpy directly is about the only thing you can do*.
Your file is 26492928 bytes = 4224 x 3136 x 2 bytes, there literally isn't anything else in there.
rawpy/LibRaw can only parse actual "raw" file containers that also come with mandatory metadata helping them to interpret the bytes within.
* If a proper raw file is absolutely necessary, you could wrap your raw bytestream with the metadata you know/record separately in e.g. a DNG file using either libtiff, or something along the lines of the pgm2dng tool.
However, there also seems to be a low-level open_bayer() method in LibRaw, but I'm not sure if it makes sense for rawpy to expose that...
letmaik
commented
3 years ago I agree with @kmilos that wrapping as DNG would make the most sense in order to load it up in rawpy and be able to demosaic it. The new open_bayer() function of LibRaw would theoretically allow to bypass this step but for that I would probably have to expose a lot more of LibRaw in order to set all required metadata options for it do to the right kind of processing. Given that this is a rather niche use case (most people just load up regular camera raw images) I'd rather not do it.
kmilos
commented
3 years ago Btw, in the thread I mentioned, there was a Python only attempt at DNG creation that seems promising. Needs some patching of pylibtiff though...
There is also PyDNG
MarcelDogotari
commented
3 years ago I have used ImageJ to open this 4224 x 3136 little-endian byte order file but for some reason rawpy complains:
with rawpy.imread('template.raw') as raw: File "/home/pas/.local/lib/python3.6/site-packages/rawpy/__init__.py", line 20, in imread d.open_file(pathOrFile) File "rawpy/_rawpy.pyx", line 404, in rawpy._rawpy.RawPy.open_file File "rawpy/_rawpy.pyx", line 914, in rawpy._rawpy.RawPy.handle_error rawpy._rawpy.LibRawFileUnsupportedError: b'Unsupported file format or not RAW file'Is there a way to get additional debug or a way to fix this image loading problem?
// Captured 13:52:29 - Wednesday, October 28, 2020 // // Frame 10 of 10 // Sensor info: // Width = 4224 // Height = 3136 // Top Non-Pixel Rows = 0 // Bottom Non-Pixel Rows = 0 // Image Format = Bayer 10 // Subformat = GRBG // Sensor output = 10 bits per pixel // Color Filter = Monochrome // // .RAW file info: // stored as = 16 bits unsigned short // Endianess = little // bit| 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0 | // Byte 0 | ------------------Pixel[7:0]----------------- | // Byte 1 | 0 0 0 0 0 0 | Pixel[9:8]| // // Application info: // Application Name = C:/Aptina Imaging/bin/DevWareX.exe // Application Version = 6.0.25.54771 // Application Type = 6.0.25_Release // Application Date = 03/12/2020
Hi peteasa! Did you have any luck with this issue? I also need to open a bytestream that is lacking any metadata/header. Did you find a solution? Thanks!
peteasa
commented
3 years ago Hi,
In my case I just open the data as a numpy array. All my images are the same size so it is easy for me
data = np.fromfile(infile, dtype=np.uint16) data = np.reshape(data, (height, width)) data.tofile(outfile)
HongChow
commented
2 years ago Hi,
In my case I just open the data as a numpy array. All my images are the same size so it is easy for me
data = np.fromfile(infile, dtype=np.uint16) data = np.reshape(data, (height, width)) data.tofile(outfile)
hi , could you still use raw post process using numpy data array?
I have used ImageJ to open this 4224 x 3136 little-endian byte order file but for some reason rawpy complains:
with rawpy.imread('template.raw') as raw: File "/home/pas/.local/lib/python3.6/site-packages/rawpy/__init__.py", line 20, in imread d.open_file(pathOrFile) File "rawpy/_rawpy.pyx", line 404, in rawpy._rawpy.RawPy.open_file File "rawpy/_rawpy.pyx", line 914, in rawpy._rawpy.RawPy.handle_error rawpy._rawpy.LibRawFileUnsupportedError: b'Unsupported file format or not RAW file'Is there a way to get additional debug or a way to fix this image loading problem?// Captured 13:52:29 - Wednesday, October 28, 2020 // // Frame 10 of 10 // Sensor info: // Width = 4224 // Height = 3136 // Top Non-Pixel Rows = 0 // Bottom Non-Pixel Rows = 0 // Image Format = Bayer 10 // Subformat = GRBG // Sensor output = 10 bits per pixel // Color Filter = Monochrome // // .RAW file info: // stored as = 16 bits unsigned short // Endianess = little // bit| 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0 | // Byte 0 | ------------------Pixel[7:0]----------------- | // Byte 1 | 0 0 0 0 0 0 | Pixel[9:8]| // // Application info: // Application Name = C:/Aptina Imaging/bin/DevWareX.exe // Application Version = 6.0.25.54771 // Application Type = 6.0.25_Release // Application Date = 03/12/2020