System, method, and computer program for encoding and decoding a unique signature in a video file as a set of watermarks

A non-transitory, computer-readable medium comprising a computer program, that, when executed by a computer system, enables the computer system to perform the following method for encoding a digital signature for a user in a video and for decoding the digital signature from the video, wherein the digital signature is substantially invisible to a viewer of the video, easy to detect during the decoding process, and resilient to transcoding, video editing, and high-compression encoding, the method comprising:

• associating the user with a unique number, wherein the unique number is the user's digital signature;
• converting the digital signature into a binary number;
• dividing the binary number into sequences, wherein each sequence includes a subset of data bits from the binary number and is associated with a sequence ID that enables the data bits in the sequences to later be combined in the correct order to recreate the binary number;
• creating a watermark for each of the sequences, wherein the watermark is a P×P grid of cells with certain cells corresponding to data bits and certain cells corresponding to the sequence ID, where P is a positive integer;
• finding suitable areas to embed the watermarks in the video by performing the following:
• (i) for N sequential frames in the video, generating an image that illustrates areas that have substantially the same pixel value in a color channel across the N sequential frames, where N is a positive integer of 2 or more;
• (ii) filtering out the areas that are too small for a watermark;
• (iii) determining that the remaining areas are suitable areas;
• (iv) for each of the suitable areas, calculate a luminosity change measure for the area across the N sequential frames, a color channel change measure for the area across the N sequential frames, and a mean pixel value for the color channel across the N sequential frames;
• (v) store the location of all the suitable areas found in the N sequential frames, along with the corresponding luminosity change measure, color change measure, and mean pixel value;
• (vi) repeating steps (i)-(v) for a plurality of sets of N frames;
• ranking the suitable areas in order of least visibility by sorting the areas first in ascending order of luminosity change, and, for areas with the same luminosity change measure, further sorting them by ascending order of color change;
• embedding one watermark per suitable area in order of the ranking, starting from the first sequence and proceeding through the last sequence, wherein each watermark is embedded in a subset of the N sequential frames of the suitable area, wherein embedding a watermark comprises the following:
• for either all cells representing binary 0 or all the cells representing binary 1 in the watermark, modifying the pixels in the suitable area corresponding to such cells by adding or subtracting a constant value Z to said pixel values, wherein the value of Z is such that the modification of the pixels is small enough to be invisible to the human eye, but large enough to be preserved during video transcoding;
• decoding the watermarks from the video by performing the following:
• extracting the color channel from the video;
• for each pair of consecutive frames, generating an image that illustrates areas that have a pixel difference across the two frames substantially equal to Z (“blobs”);
• filtering out blobs that are smaller than the cells of a watermark;
• identifying blobs that substantially match the size and shape of a watermark cell (“candidate blobs”);
• extracting any watermark patterns by determining whether any groups of candidate blobs match predefined constraints for a watermark;
• repeating the above decoding steps for each pair of consecutive frames in a video;
• mapping extracted watermark patterns to sequence IDs and extracting the data bits from each watermark pattern;
• and
• reconstructing the digital signature by combining data bits from each sequence.

A computer system for encoding a digital signature for a user in a video and for decoding the digital signature from the video, wherein the digital signature is substantially invisible to a viewer of the video, easy to detect during the decoding process, and resilient to transcoding, video editing, and high-compression encoding, the system comprising:

• one or more processors;
• one or more memory units coupled to the one or more processors, wherein the one or more memory units store instructions that, when executed by the one or more processors, cause the system to perform the operations of:
• encoding a unique signature into a video file by performing the following:
• associating the user with a unique number, wherein the unique number is the user's digital signature;
• converting the digital signature into a binary number;
• dividing the binary number into sequences, wherein each sequence includes a subset of data bits from the binary number and is associated with a sequence ID that enables the data bits in the sequences to later be combined in the correct order to recreate the binary number;
• creating a watermark for each of the sequences, wherein the watermark is a P×P grid of cells with certain cells corresponding to data bits and certain cells corresponding to the sequence ID, where P is a positive integer;
• finding suitable areas to embed the watermarks in the video by performing the following:
• (i) for N sequential frames in the video, generating an image that illustrates areas that have substantially the same pixel value in a color channel across the N sequential frames, where N is a positive integer of 2 or more;
• (ii) filtering out the areas that are too small for a watermark;
• (iii) determining that the remaining areas are suitable areas;
• (iv) for each of the suitable areas, calculate a luminosity change measure for the area across the N sequential frames, a color channel change measure for the area across the N sequential frames, and a mean pixel value for the color channel across the N sequential frames;
• (v) store the location of all the suitable areas found in the N sequential frames, along with the corresponding luminosity change measure, color change measure, and mean pixel value;
• (vi) repeating steps (i)-(v) for a plurality of sets of N frames;
• ranking the suitable areas in order of least visibility by sorting the areas first in ascending order of luminosity change, and, for areas with the same luminosity change measure, further sorting them by ascending order of color change;
• embedding one watermark per suitable area in order of the ranking, starting from the first sequence and proceeding through the last sequence, wherein each watermark is embedded in a subset of the N sequential frames of the suitable area, wherein embedding a watermark comprises the following:
• for either all cells representing binary 0 or all the cells representing binary 1 in the watermark, modifying the pixels in the suitable area corresponding to such cells by adding or subtracting a constant value Z to said pixel values, wherein the value of Z is such that the modification of the pixels is small enough to be invisible to the human eye, but large enough to be preserved during video transcoding;
• decoding the watermarks from the video by performing the following:
• extracting the color channel from the video;
• for each pair of consecutive frames, generating an image that illustrates areas that have a pixel difference across the two frames substantially equal to Z (“blobs”);
• filtering out blobs that are smaller than the cells of a watermark;
• identifying blobs that substantially match the size and shape of a watermark cell (“candidate blobs”);
• extracting any watermark patterns by determining whether any groups of candidate blobs match predefined constraints for a watermark;
• repeating the above decoding steps for each pair of consecutive frames in a video;
• mapping extracted watermark patterns to sequence IDs and extracting the data bits from each watermark pattern;
• and
• reconstructing the digital signature by combining data bits from each sequence.