import struct import PIL.Image from . import hashing def detectpng(content, maxpixels=1024 * 1024 * 32): if len(content) < 33: return None # defer decision if not content.startswith(b"\x89PNG\r\n\x1a\n\0\0\0\x0dIHDR"): return False width, height = struct.unpack(">II", content[16:24]) if width * height > maxpixels: return False return True class ImageHash(hashing.StoredHash): """A hash on the contents of an image. This disregards mode, depth and meta information. Note that due to limitations in PIL and the image format (interlacing) the full contents are stored and decoded in hexdigest.""" maxsize = 1024 * 1024 * 32 # max memory usage is about 5 * maxpixels in bytes maxpixels = 1024 * 1024 * 32 def __init__(self, hashobj): """ @param hashobj: a hashlib-like object """ self.hashobj = hashobj hashing.StoredHash.__init__(self, self.computehash, self.detect, self.maxsize) self.hashobj = hashobj def detect(self, bytesio): return detectpng(bytesio.read(33), self.maxpixels) def computehash(self, bytesio): hashobj = self.hashobj.copy() img = PIL.Image.open(self.content) width, height = img.size pack = lambda elem: struct.pack("BBBB", *elem) # special casing easy modes reduces memory usage if img.mode == "L": pack = lambda elem: struct.pack("BBBB", elem, elem, elem, 255) elif img.mode == "RGB": pack = lambda elem: struct.pack("BBBB", *(elem + (255,))) elif img.mode != "RGBA": img = img.convert("RGBA") for elem in img.getdata(): hashobj.update(pack(elem)) return "%s%8.8x%8.8x" % (hashobj.hexdigest(), width, height)