A User Attention Based Visible Watermarking Scheme

1
A User Attention Based Visible Watermarking Scheme

• 2003/12/15
• Chen-Hsiu Huang and Ja-Ling Wu
• Communication and Multimedia Laboratory,
• National Taiwan University,
• E-mail chenhsiu,wjl_at_cmlab.csie.ntu.edu.tw

2
Introduction

• Digital watermarking the process of embedding
  data into a multimedia objects so that the
  embedded watermark can be detected or extracted
  later.
• According to human perceptivity, it can be
  divided into two different types, visible and
  invisible.
• Visible watermark is a second transparent pattern
  or image overlaid with the primary (host) image.

Visible Watermark
3
Criterions about Visible Watermark

• It should be obvious to be seen.
• It should be spread in a large or important area
  of the image.
• It should be visible but not significantly
  obscure the details.
• It must be difficult to remove its more costly
  and labor intensive to remove it.
• Embedding process should be automatic, with
  little human intervention and labor.

4
Challenges for Visible Watermarks

• How can we spread the embedding watermark in a
  large or important area without significantly
  obscure the details?
• If we can achieve these, how can this embedding
  process being labor saving, that is, without too
  much user intervention?

5

• Most of the proposed schemes still treat video
  data as series of pictures and apply image
  watermarking schemes to each frame.
• This may not suitable for video because the most
  significant properties, temporal correlation and
  motion information are not taken into account.
• With the computational attention models, we have
  more objective references for choosing the
  locations to insert watermarks.

6
The User Attention Model

• Attention refers to the ability of a human to
  focus and concentrate upon some visual or
  auditory object, by careful observing or
  listening.

• For features in different levels, various
  saliency maps are generated to capture users'
  focus.
• Once the final saliency map is generated, the
  focus point can be detected and models observers
  attention successfully.

7

• While users pay more attention to the focus
  regions, they usually pay less attention to other
  regions,
• These non-focus regions are good candidates to
  embed watermarks.

8
Selection of Watermark Location

• For those regions far away from the focus point
  are good candidates to embed watermarks.
• In order to decrease the perceptual distortion,
  the regions of lower intensity and higher texture
  are the most suitable ones.

• A weighting function is defined to measure each
  macroblocks unsuitability
• The one with highest unsuitability is selected
  for embedding.

9
Watermark Embedding

• In order not to obscure the original, we select a
  luminance scaling coefficient f based on
  backgrounds intensity and texture to scale
  watermarks strength.

10
Example Frames of with Watermark
11
Shot Based Watermarking

• Since the embedding process is performed on frame
  level, the embedded watermark may look like
  another unstable moving object in the video,
  which is quite annoying.
• For resolving this shortage, we change our policy
  to calculate the watermarking location only once
  in a video location only once in a video shot.
• The watermark appears in the same location
  during each shot and only moves when video
  shot-change occurs.

12
Demo
13
Results

• In our scheme, owners rights claim is preserved
  without spoiling the videos presentation.
• The embedding process is performed once in each
  shot and will not annoy the original video due to
  frequently movements.
• The watermark differs in each shot both in its
  location and strength. Thus cannot be predicted
  in advance due to its content-dependent nature.
• The watermarking can be done fully automatic,
  i.e. labor saving without users intervention.

14

• Borrowing techniques developed in the content
  analysis field, such as modeling of users
  attention, is beneficial to digital watermarking.
• We believe that by exploiting more content
  features with higher semantic meanings,
  approaching a better digital watermarking
  scenario could be possible.