def segment_map(filehandle):
"""Get the map of all segments in a jpeg file.
1st parameter = file handle of jpeg file, open for binary read
Returns a list of tuples corresponding to the segments in the file,
in the order they occur in the file. Each tuple contains a segment
ID and an absolute offset in the file; e.g., ('APP1', 2).
Note: we stop at EOI (End Of Image) or SOS (Start Of Scan). This
means that we ignore any segments that might occur after SOS -- a
theoretical possibility, although we've never seen a Jpeg structured
in that manner and stopping at SOS provides better performance
because we don't have to process any image data.
"""
segments = [] # initialize the list of segments
filehandle.seek(2, 0) # first segment starts right after the SOI
while True:
seg_mark = filehandle.read(2)
if len(seg_mark) < 2:
break # file parsing error: we've reached EOF unexpectedly
seg_id = seginfo(seg_mark)['name']
if seg_id == 'APP1':
# determine whether APP1 format is Exif, XMP, or XMP Extended
filepos = filehandle.tell() # note current file position
_ = filehandle.read(2) # skip over thec data size value
id_str = filehandle.read(35) # APP1 identification string
if id_str[:6] == b'Exif\x00\x00':
segments.append(('APP1-Exif', filepos-2))
elif id_str[:29] == b'http://ns.adobe.com/xap/1.0/\x00':
segments.append(('APP1-XMP', filepos-2))
elif id_str[:35] == b'http://ns.adobe.com/xmp/extension/\x00':
segments.append(('APP1-XMPext', filepos-2))
else:
segments.append(('APP1-unknown', filepos-2))
filehandle.seek(filepos, 0) # return to current file position
else:
# non-APP1 segment, add it to the list
segments.append((seg_id, filehandle.tell()-2))
if seg_id == 'EOI' or seg_id == 'SOS':
break # stop processing the image
if seg_mark in [b'\xff\x01', b'\xff\xd0', b'\xff\xd1', b'\xff\xd2',
b'\xff\xd3', b'\xff\xd4', b'\xff\xd5', b'\xff\xd6',
b'\xff\xd7', b'\xff\xd8', b'\xff\xd9']:
# These segment markers have no payload, so we're already
# positioned for the next segment after reading the segment marker
datasize = 0
else:
dsbytes = filehandle.read(2)
if len(dsbytes) < 2:
break # file parsing error: we've reached EOF unexpectedly
datasize = struct.unpack('>H', dsbytes)[0]
# skip forward to next segment, ready to repeat the loop
filehandle.seek(datasize-2, 1)
return segments
def segment_read(filehandle=None):
"""Convert a Jpeg segment to a dictionary.
filehandle = Jpeg file open for binary read, positioned to first byte
of the segment
Returns a dictionary with these keys:
offset = offset of the segment within the Jpeg file
segmark = the 2-byte segment marker
segtype = name of this segment type (e.g., 'APP1')
has_data = whether segment type has a data payload
has_meta = whether segment type's payload contains metadata
payload = segment's data payload
"""
# initialize dictionary object
segdict = {}
segdict['offset'] = filehandle.tell()
segdict['segmark'] = filehandle.read(2)
segdict['payload'] = None # default value
segdict['next_segment'] = None # default value
# get info about this segment type and copy to dictionary
segtype_info = seginfo(segdict['segmark'])
segdict['segtype'] = segtype_info['name']
segdict['has_data'] = segtype_info['has_data']
segdict['has_meta'] = segtype_info['has_meta']
# Stop processing the file when SOS or EOI segment reached. We do this
# because we're only interested in reading metadata, and want to maximize
# performance for scanning large numbers of images quickly. The SOS
# segment is different from the others in that its " data size" is
# merely the size of the SOS header. The compressed data imediately
# follows, and this is by far the largest segment in a typical Jpeg file.
# Furthermore, the compressed data must actually be scanned and decoded
# to find the EOI marker that follows, and we want to avoid the need to
# read all of that data.
# Note: if we terminate at SOS then we should never actually see an EOI,
# but we're checking for EOI as well here to allow for the case where a
# Jpeg file has no image data and only metadata -- we've not seen this
# in an actual Jpeg, but it's theoretically possible and by checking for
# both SOS and EOI here we will gracefully handle any such file.
if segdict['segtype'] in ['SOS', 'EOI']:
return segdict
# if this segment type has no data payload, then the next segment
# starts right after the 2-byte segment marker
if not segdict['has_data']:
segdict['next_segment'] = segdict['offset'] + 2
return segdict
# read data size; note that this size includes the 2-byte size
# itself but doesn't include the 2-byte segment marker
datasize_bytes = filehandle.read(2)
datasize = struct.unpack('>H', datasize_bytes)[0]
# if segment contains metadata, save a copy in the segment's dictionary
if segdict['has_meta']:
segdict['payload'] = filehandle.read(datasize - 2)
else:
# no metadata to save, so just skip past the data
filehandle.seek(datasize-2, 1)
segdict['next_segment'] = filehandle.tell()
return segdict