-
-
Save bcyrill/e59fda6c7ffe23c7c4b08a990804b269 to your computer and use it in GitHub Desktop.
#!/usr/bin/python | |
# The Samsung trailer format is based on the implementation from ExifTool | |
# http://www.sno.phy.queensu.ca/~phil/exiftool/ | |
import mmap | |
import struct | |
import os | |
import sys | |
if (len(sys.argv) < 2) or (len(sys.argv) > 3): | |
print("Usage: fix_eoi.py <filename> [dump]") | |
exit() | |
file = sys.argv[1] | |
if len(sys.argv) < 3: | |
dump = 0 | |
elif sys.argv[2] == "dump": | |
dump = 1 | |
else: | |
print("Invalid argument") | |
print("Usage: fix_eoi.py <filename> [dump]") | |
exit() | |
(file_name, file_ext) = os.path.splitext(file) | |
with open(file, 'rb') as fh: | |
m = mmap.mmap(fh.fileno(), 0, access=mmap.ACCESS_READ) | |
b = bytearray(m) | |
trailer_tail = b[-4:] | |
if trailer_tail != 'SEFT': | |
print("No valid Samsung trailer found") | |
exit() | |
else: | |
print("Found SEFT") | |
length = struct.unpack_from("<I",b[-8:-4])[0] | |
trailer = b[-(8+length):] | |
endPos = len(b) | |
dirPos = endPos-(8+length) | |
if trailer[0:4] != 'SEFH': | |
print("No valid Samsung trailer found") | |
exit() | |
else: | |
print("Found SEFH") | |
version = struct.unpack_from("<I",trailer[4:8])[0] | |
if version != 101: | |
print("Unknown Samsung trailer version") | |
exit() | |
count = struct.unpack_from("<I",trailer[8:12])[0] | |
firstBlock = 0 | |
is_pano = 0 | |
for index in range(0, count): | |
entry = 12 + 12 * index; | |
type = struct.unpack_from("<H",trailer[entry+2:entry+4])[0] | |
noff = struct.unpack_from("<I",trailer[entry+4:entry+8])[0] | |
size = struct.unpack_from("<I",trailer[entry+8:entry+12])[0] | |
if firstBlock < noff: | |
firstBlock = noff | |
entryPos = dirPos - noff | |
entryLen = size | |
data = b[entryPos:entryPos+entryLen] | |
# Validate as the type has to match the SEFH/SEFT entry type | |
entry_type = struct.unpack_from("<H",data[2:4])[0] | |
if type != entry_type: | |
print("Missmatch detected. Aborting!") | |
exit() | |
entry_offset = struct.unpack_from("<I",data[4:8])[0] | |
entry_name = data[8:8+entry_offset].decode("utf-8") | |
if entry_name == "Panorama_Shot_Info": | |
is_pano = 1 | |
entry_data = data[8+entry_offset:] | |
if dump: | |
print("Dumping: %s" % entry_name) | |
with open(file_name + '_' + entry_name, 'wb') as f: | |
f.write(entry_data) | |
if not is_pano: | |
print "No Panorama info found" | |
exit() | |
dataPos = dirPos - firstBlock | |
dirLen = endPos - dataPos | |
eoi = struct.unpack_from(">H",b[dataPos-2:dataPos])[0] | |
if eoi == 0xffd9: | |
print("Found EOI") | |
exit() | |
else: | |
print("Inserting EOI") | |
with open(file_name + '_pano' + file_ext, 'wb') as f: | |
f.write(b[0:dataPos]) | |
f.write(bytearray(b'\xff\xd9')) | |
f.write(b[dataPos:]) | |
Great work! Thanks!
This is wonderful thank you, it's helping me import my panoramas into Shotwell on Linux. I'm wondering, if there is this extra MP4 track inside the JPEG (I didn't even know that was possible) and you can "dump" it to a separate file, could there be an option to exclude it from the corrected JPEG? I don't need that track and it's about a third of the whole file size.
@nrrkeene
The following command (if you have jpegtran) will strip the extra junk (after the file has been corrected):
jpegtran -copy all -optimize fixed.jpg > fixed_stripped.jpg
Added a fork that works w. S8+, plus some decorations for batch friendliness.
I made a rust version which can strip the files of the videos and reduce the size in about half. This version merely makes the file a valid JPEG file but the video remains in there. I also made some binaries if somebody is interested: https://github.com/frapa/sampan
@O-Sullivan
Thanks, it works!