Method for embedding and extracting digital data in images and video

• US 6,332,030 B1
• Filed: 01/14/1999
• Issued: 12/18/2001
• Est. Priority Date: 01/15/1998
• Status: Expired due to Fees

• Current Assignee: Regents of the University of California (University of California)
• Original Assignee: Regents of the University of California (University of California)
• Inventors: Mukherjee, Debargha, Manjunath, B. S., Chae, Jong Jin, Mitra, Sanjit K.
• Primary Examiner(s): Boudreau, Leo
• Assistant Examiner(s): Tabatabai, Abolfazl
• Application Number:
  US09/231,547
• Time in Patent Office: 1,069 Days
• Field of Search: 382/100, 382/248, 382/253, 382/232, 713/176
• US Class Current: 382/100
• CPC Class Codes: G06T 1/0028 Adaptive watermarking, e.g. Human Visual System [HVS]-based watermarking
• CPC Class Codes: G06T 1/0057 Compression invariant watermarking
• CPC Class Codes: G06T 2201/0052 Embedding of the watermark in the frequency domain
• CPC Class Codes: G06T 2201/0081 whereby both original and watermarked images are required at decoder, e.g. destination-based, non-blind, non-oblivious
• CPC Class Codes: G06T 2201/0202 whereby the quality of watermarked images is measured; Measuring quality or performance of watermarking methods; Balancing between quality and robustness

Abstract

A method for digital watermarking and, in particular, for digital data hiding of significant amounts of data in images and video. The method employs a discrete wavelet transform for embedding gray scale images which can be as great as 25% of the host image data. A simple control parameter is used that can be tailored to either hiding or watermarking purposes, and is robust to operations such as JPEG compression. The method also uses noise-resilient channel codes based on multidimensional lattices which can provide for embedding signature data such as gray-scale or color images. Furthermore, embedded image data can be recovered in the absence of the original host image by inserting the data into the host image in the DCT domain by encoding the signature DCT coefficients using a lattice coding scheme before embedding, checking each block of host DCT coefficients for its texture content, and appropriately inserting the signatured codes depending on a local texture measure. The method further provides for source coding the signature data by vector quantization, where the indices are embedded in the host by perturbing it using orthogonal transform domain vector perturbations. The transform coefficients of the parent data are grouped into vectors, and the vectors are perturbed using noise-resilient channel codes derived from multidimensional lattices. The perturbations are constrained by a maximum allowable mean-squared error that can be introduced in the host. Also, speech can be hidden in video by wavelet transforming the host video frame by frame, and perturbing vectors of coefficients using lattice channel codes to represent hidden vector quantized speech. The embedded video is subjected to H.263 compression before retrieving the hidden speech.

Claims

1. A method for embedding a signature image in a host image, comprising:

(a) performing a single level discrete wavelet transform decomposition of said signature image and said host image;

(b) quantizing into β

levels each coefficient of said signature image by grouping a set of n coefficients in the host image to form an n-dimensional vector, and perturbing said vector according to a β

-ary channel code comprising a subset of an n-dimensional lattice scaled by a factor α

;

(c) embedding each subband of said signature image into a corresponding subband of said host image to produce a composite image;

(d) subtracting the coefficients of said host image from the coefficients of the composite image to obtain noisy perturbations;

(e) grouping the resulting coefficients into groups of n to obtain a vector {right arrow over (e)};

(f) scaling said vector {right arrow over (e)} by 1/α

to produce a resulting vector 1/α

·

{right arrow over (e)};

(g) nearest-neighbor encoding 1/α

·

{right arrow over (e)} to find an index i of the channel code nearest to it in Euclidean distance;

(h) obtaining a quantized discrete wavelet transform coefficients from said index i.

2. A method for embedding an audio signature in a host video image, comprising:

(a) encoding said audio signature to generate a sequence of symbols from a Q-ary alphabet {s₁, s₂, . . . , s_Q};

(b) injecting one symbol in each coefficient vector V_j, by perturbing it in at least one of Q possible ways in k-dimensional space to obtain the perturbed vector V_j; and

(c) using perturbed coefficients to inverse transform said host video image and produce a composite signal.

3. A method as recited in claim 2, further comprising extracting from each perturbed vector the symbol within whose decision boundaries the vector of the composite signal lies.