Seam carving based anonymization
As image source attribution techniques have become significantly sophisticated and are now becoming commonplace, there is a growing need for capabilities to anonymize images and videos. Focusing on the photo response non-uniformity noise pattern based sensor fingerprinting technique, this work evaluates the effectiveness of well-established seam carving method to defend against sensor fingerprint matching. We consider ways in which seam-carving based anonymization can be countered and propose enhancements over conventional seam carving method. Our results show that applying geometrical distortion in addition to seam carving will make counter attack very ineffective both in terms of computational complexity and accuracy.
1. A method of anonymization of images against source attribution comprising:
irreversibly transforming an original digital image into an anonymized image by:
determining a plurality of seams for the image;
adding or removing at least one pixel from one of the determined seams;
wherein pixel correspondences between the original image's sub-blocks and the anonymized image's sub-blocks are not preserved,
removing one or more fake seam by:
a) identifying an image region that lies between two determined seams of the plurality of seams;
b) selecting a pixel in the identified image region;
c) calculating gradient energy in a first direction and a second direction from the pixel in the identified region until a minimum energy is found as a full horizontal or vertical fake seam; and
d) removing the full horizontal or vertical fake seam.
2. The method of claim 1 , further comprising adding or removing at least one pixel from each of the determined seams.
3. The method of claim 2 , further comprising: creating one or more fake seams by:
treating the removed full horizontal or vertical fake seam as a seam of the determined plurality of seams;
repeating removal of the one or more fake seams until no image region between two determined seams of the plurality of seams larger than a predetermined size are left.
4. The method of claim 3 , wherein the predetermined size is equal to or less than about 50×50.
5. The method of claim 2 , wherein altering the image comprises removing the determined plurality of seams and shifting the remaining pixels.
6. The method of claim 2 , further comprising the distorting the image prior to determining the plurality of seams.
7. The method of claim 1 , wherein the one or more fake seams cross hot and dead pixels.