Lecture 17: How to Spend those Bits we Saved…AnnouncementsGoals for TodayTypes of Coding (Orsak et al. Sec. 6.1+)Big Picture of CodingWhy Code???General View of CodingTypes of Coding: Lecture PlanDigital WatermarkingResources for ExamplesVisible WatermarkingInvisible WatermarkingInformation HidingThose Imperceptible Bits…Basic IdeaDigital vs. Traditional WatermarksIssues in WatermarkingDemosThings to Think About…Lecture 17: How to Spend those Bits we Saved…The Digital World of MultimediaProf. Mari OstendorfEE299 Lecture 1714 February 2007Announcements No HW due this week Work on your lab! Holiday next Monday 2/18 Exam Friday 2/22 Review in class 2/20 Last year’s exam is on the course web pageEE299 Lecture 1714 February 2007Goals for Today Big picture of signal coding Spending vs. saving bits Different reasons for different types of coding Building on ideas from lossy compression Watermarking, information hidingEE299 Lecture 1714 February 2007Types of Coding (Orsak et al. Sec. 6.1+) Formatting Describes what type of file it is & how to translate bits into asignal Compression (Source Coding) Reduces the number of bits needed to represent the signal (lossless & lossy) Watermarking Adds unobservable information to the signal so someone can trace where it came from Error detection and correction (Channel Coding) Adds extra bits to make it possible to detect and sometimes correct errors Encryption Modifies the signal or bit sequence to conceal the true valueEE299 Lecture 1714 February 2007Big Picture of Codingoriginal signalSignal to be stored or transmittedEncodingDecoding: reverse the processcompresswatermarkformaterror codeencryptEE299 Lecture 1714 February 2007Why Code??? Formatting Practical implementation Compression Faster communications, less storage Watermarking Establish ownership, information hiding Error Coding Reduce impact of errors Encryption Privacy, securityEE299 Lecture 1714 February 2007General View of Coding Change the representation of the signal Transforms, scrambling Quantization (lossy compression) Expanding (error coding) Need mapping functions to get “original”: Undoing the expansion How bits map to numbers Which bits correspond to different aspects of the signalEE299 Lecture 1714 February 2007Types of Coding: Lecture Plan Formatting Describes what type of file it is & how to translate bits into asignal Compression (Source Coding) Reduces the number of bits needed to represent the signal (lossless & lossy) Watermarking Adds unobservable information to the signal so someone can trace where it came from Error detection and correction (Channel Coding) Adds extra bits to make it possible to detect and sometimes correct errors Encryption Modifies the signal or bit sequence to conceal the true valueLast weekNext timeTodayNot coveredAfter examEE299 Lecture 1714 February 2007Digital Watermarking What is digital watermarking? Imperceptible or barely noticable transformation of a digital signal that embeds a message in it Mainly used for audio, image and video Multiple purposes: Indicate creator/owner and/or purchaser Information hiding Tamper detection Multiple types Visible: Lighten or darken colors Invisible: Use the bits below the noise floor (leverage human perception)EE299 Lecture 1714 February 2007Resources for Examples[1] http://www.dlib.org/dlib/december97/ibm/1 2lotspiech.html[2] www.sagesecure.com[3] en.wikipedia.org[4] www.catchlock.com[5] www.caci.comEE299 Lecture 1714 February 2007Visible WatermarkingDifferent uses:• To remind viewer who owns image• To let viewer know where to find image• For safeguarding previews(from [1], Vatican library image)(from [3])(from [1])EE299 Lecture 1714 February 2007Invisible Watermarking(from [4])Different uses:• Digital rights• Trace bootlegged movie previews• Identify illegal copies • Find source of illegal copies• Access control• Information hiding• Descriptive information (date, source) • Title for broadcast auditing• Secret messageEE299 Lecture 1714 February 2007Information Hiding(from [5])EE299 Lecture 1714 February 2007Those Imperceptible Bits… Recall from audio coding lecture: Below a certain threshold, humans can’t hear the sound (analogous ideas for images) From the bit perspective: Idea in compression: don’t send those bits Idea in watermarking: put a message in those bits1 0 1 1 0 1 1 1 0 1 0 0 1 1 0 1Least significant bits Æbelow the noise floorBits associated with high amplitudeEE299 Lecture 1714 February 2007Basic Idea(from [2])EE299 Lecture 1714 February 2007Digital vs. Traditional WatermarksTraditional Paper Elaborate & costly Visible (though may be hard to see) Lost in copy process Aimed mainly at safeguarding user (by ensuring authenticity)Digital Fast, cheap & easy Often not visible, need computer processing to read Usually kept when signal is copied Aimed mainly at safeguarding ownerEE299 Lecture 1714 February 2007Issues in Watermarking Factors to trade off (interrelated) Number of bits What information is encoded Is encryption needed for privacy Visibility (intentional or unintentional) Robustness to signal modification One big watermark is more fragile than repeats of small watermarks Different media types are subject to different degrees of perceptual masking (watermarking text is very hard!)EE299 Lecture 1714 February 2007Demos Embed text in an image Embed text in a sound file Embed a soundfile in an image Embed an image in an image Approach: use least significant bits of carrier signal to hold hidden signalEE299 Lecture 1714 February 2007Things to Think About… Is identification of the purchaser an invasion of privacy? How vulnerable are watermarks to attacks? Can watermarks be used as evidence in a legal
View Full Document