A DWT-based Digital Video Watermarking Scheme with Error Correcting Code

Pat P. W. Chan and Michael R. LYU

Computer Science and Engineering Department

The Chinese University of Hong Kong

Fifth International Conference on Information and Communications Security ICICS 2003

Huhehaote City, Inner-Mongolia, China, 10-13 October, 2003

2Computer Science and Engineering Department, CUHK

Outline

Introduction New Video Watermarking Algorithm Experimental Results Future Direction Conclusion

3Computer Science and Engineering Department, CUHK

Introduction

Watermarking is a concept of embedding a special pattern, watermark, into a document.

Watermarking is a key process for the protection of copyright ownership of electronic data.

In this presentation, we will focus on the video watermarking scheme.

Watermarking Algorithm

WatermarkedImage Ì

Stego-Image I

Watermark W

Secret / public key K

4Computer Science and Engineering Department, CUHK

Introduction

Video watermarking is challenging Video watermarking introduces some issues not present in

image watermarking. Due to large amounts of data and inherent redundancy

between frames, video signals are highly susceptible to pirate attacks, including frame averaging, frame dropping, frame swapping, statistical analysis, etc.

However, the currently proposed algorithms do not solve these problems effectively.

Existing techniques are not aware of the usefulness of the audio channel in a video.

5Computer Science and Engineering Department, CUHK

Introduction

A new scheme for robust blind digital video watermarking with error correcting code will be introduced.

The features of the video watermarking algorithm are: Our scheme embeds different parts of a single watermark into

different scenes of a video under the wavelet domain; To increase robustness of the scheme, the watermark is refined

by the error correcting code, while the correcting code is embedded as watermark in the audio channel;

Our video watermarking algorithm is robust against the attacks of frame dropping, averaging and statistical analysis;

It allows blind retrieval of embedded watermark which does not need the original video;

The watermark is perceptually invisible.

6Computer Science and Engineering Department, CUHK

Watermarking Scheme Overview

7Computer Science and Engineering Department, CUHK

Mark with m1 Mark with m3 Mark with m7Mark with m0

Scene change occur

Video frames are transformed to wavelet domain. Perform scene change detection. Each scene is embedded with the same watermark, so it can

prevent attackers from removing the watermark by frame dropping. Different watermarks used for successive different scene can

prevent attackers from colluding with frames from completely different scenes.

Video Preprocess:DWT & Scene Change Detection

8Computer Science and Engineering Department, CUHK

Watermark Preprocess

Scale the watermark to a particular size with the following equations 2n m≦ , n>0 p + q = n , p and q > 0 Size of image = 64˙2p X 64˙2q

Divide the image into 2n small images with size 64 X 64

m -- # of scene change of the video

m=10, n=3, p=1, q=2

9Computer Science and Engineering Department, CUHK

Watermark Preprocess

Preprocessed watermark m0-m7

Original watermark Encrypted watermark m’0

10Computer Science and Engineering Department, CUHK

Video Watermark Embedding

Exchange C[i] with max(C[i], C[i+1], C[i+2], C[i+3], C[i+4]) while W[j] = 1

Exchange C[i] with min(C[i], C[i+1], C[i+2], C[i+3], C[i+4])while W[j] = 0

LL, HH coefficients are not watermarked

11Computer Science and Engineering Department, CUHK

Error correcting code is extracted from the watermark image

Embedded in audio channel as an audio watermark This watermark can provided the error correction and detection capability for

the video watermark

Audio Watermark

Average

12Computer Science and Engineering Department, CUHK

Audio Watermark Embedding

Spread-Spectrum Watermarking Modulated Complex Lapped Transform (MCLT)

13Computer Science and Engineering Department, CUHK

Watermarked Frame & Wave

Watermarked video frame and waveOriginal video frame and wave

14Computer Science and Engineering Department, CUHK

Watermark Detection

Video is split into video stream and audio stream. Watermarks are extracted separately by audio

watermark extraction and video watermark extraction. Then the extracted watermark undergoes refining

process

15Computer Science and Engineering Department, CUHK

Video Watermark Detection

if WC[i] > median(WC[i], WC[i+1],WC[i+2], WC[i+3], WC[i+4])W[j] = 1

elseW[j] = 0

Watermarked video frame

Original video frame

Extracted Watermark

Recovered Watermark

16Computer Science and Engineering Department, CUHK

Watermark Refining

Error correcting codes are extracted from the audio stream The video watermark extracted is refined by this informatio

n with the following equation

k = kth block of the average image i = x-coordinate of video watermark j = y-coordinate of video watermark T = Threshold that the pixel need to correct P: Q = the ratio of importance of extracted watermark to the averag

e

17Computer Science and Engineering Department, CUHK

Experimental Setup

VirtualDub -- a video capture/processing utility (http://www.virtualdub.org/)

A video clip with 1526 frames of size 352 X 288 Experiments:

Experiment with Frame Dropping Experiment with Frame Averaging and Statistical Analysis Experiment with Lossy Compression Experiment with Cropping Watermarked Frame

Measurement:

18Computer Science and Engineering Department, CUHK

Experiment with Frame Dropping As a video contains a large

amount of redundancies between frames, it may suffer attacks by frame dropping.

From the experiment, we found that our scheme achieves better performance than the DWT-based scheme without scene-based watermarks.

19Computer Science and Engineering Department, CUHK

Experiment with Frame Averaging and Statistical Analysis Frame averaging and statistical analysis is another

common attack to the video watermark. When attackers collect a number of watermarked

frames, they can estimate the watermark by statistical averaging and remove it from the watermarked video

Scenario of statistical averaging attack

20Computer Science and Engineering Department, CUHK

Experiment with Frame Averaging and Statistical Analysis The watermarked video is

statistically analyzed by colluding a number of video frames and the watermarks are extracted and NC values are obtained.

It is found that the proposed scheme can resist to statistical averaging quite well.

21Computer Science and Engineering Department, CUHK

Experiment with Lossy Compression The performance of the scheme is

significantly improved by combining with audio watermark, especially when the quality factor of MPEG is low.

When the quality factor of MPEG is low, the error of the extracted watermark is increased and the watermark is damaged significantly.

As the error correcting code is provided from the audio watermark, it can survive the attack by lossy compression which is applied to the video channel.

22Computer Science and Engineering Department, CUHK

Experiment with Cropping Watermarked Frame DWT inherits many

advantages in resisting the attacks on the watermarked frames. It achieves perceptual invisibility and attacks by image processing techniques.

Cropping is one of the most common attack to video watermark.

23Computer Science and Engineering Department, CUHK

Conclusion & Future Direction

Video watermarking is needed since copyright protection is essential. Video watermarking is different from image watermarking. A DWT-based Digital Video Watermarking Scheme with Error

Correcting Code is proposed. Use visual-audio watermark to increase the robustness of the scheme. Increase robustness of the scheme by improving the way to embed

the watermark. Use better error correction coding. Make use of the information from the video, such as time information

to increase the robustness of the watermark.

Q & A