Objectives of Image CodingSlide Number 2Issues in Image CodingSlide Number 4Methods of Reconstruction Level AssignmentsMethods of Reconstruction Level Assignments cont.Methods of Reconstruction Level Assignments cont.Scalar Case (cont.)Scalar Case (cont.)Scalar Case (cont.)Scalar Case (cont.) Solution to Optimization ProblemSlide Number 12Slide Number 13Scalar Case (cont.)Scalar Case (cont.)Slide Number 16Vector CaseSlide Number 18Codeword Design: Bit AllocationOverviewOverviewMethods of Bit Assignment (cont.)Methods of Bit Assignment (cont.)Methods of Bit Assignment (cont.)Arithmetic CodingMechanics fo Arithmetic CodingExample of Tag Generation:Deciphering the Tag:Generating a binary code:Example Generating a binary code:Efficiency of Arithmetic CodeDictionary TechniquesStatic Dictionary:Adaptive Dictionary:ExampleExample (cont.)Example (cont.)Encoding steps:Adaptive DictionaryWhat to Code (Classification of Image Coding Systems)What to Code (Classification of Image Coding Systems) (cont.)Waveform CoderImprovements of PCM Improvements of PCM (cont.)Improvements of PCM (cont.)Improvements of PCM (cont.)Improvements of PCM (cont.)Slide Number 48Slide Number 49Slide Number 50Delta Modulation (DM)Delta Modulation (DM) (cont.)Delta Modulation (DM) (cont.)Slide Number 54Slide Number 55Differential Pulse Code Modulation (DPCM)Differential Pulse Code Modulation (DPCM) (cont.)Slide Number 58Transform Image CodingTransform Image CodingExamples of TransformsExamples of Transforms (cont.)Slide Number 63Slide Number 64Slide Number 65Slide Number 66Slide Number 67Slide Number 68Discrete Cosine TransformDiscrete Cosine Transform (cont.)Discrete Cosine Transform (cont.)Discrete Cosine Transform (cont.)Discrete Cosine Transform (cont.)Discrete Cosine Transform (cont.)Discrete Cosine Transform (cont.)Discrete Cosine Transform (cont.)DCTDCT (cont.)DCT (cont.)DCT (cont.)Discarding Transform CoefficientsDiscarding Transform CoefficientsScalar Quantization of a Vector SourceDCT-Coded ImagesDCT-Coded Images (cont.)Quantization of DCT CoefficientsBlocking Effect ReductionAdaptive Coding and Vector QuantizationIterative Procedures for Reduction of Blocking Effects in Transform Image Coding by Ruth Rosenholtz and Avideh ZakhorHybrid CodingHybrid Coding (cont.)Hybrid Coding (cont.)Two-Channel Image CoderTwo-Channel Image Coding (cont.)Pyramid Coding and Subband CodingPyramid Coding and Subband Coding (cont.)Pyramid Coding and Subband Coding (cont.)Slide Number 98Gaussian Pyramid RepresentationLaplacian Pyramid RepresentationLaplacian Pyramid Image CodingSubband CodingSubband Coding (cont.)Filter DesignFilter Design (cont.)Filter Design (cont.)Bit Allocation in Subband Coding:2D Subband CodingFrequency DomainFrequency Domain (cont.)WaveletsFamous Wavelet FiltersFractal CompressionImage CompressionImage Compression (cont.)Image Compression (cont.)Image Compression (cont.)Transformations:Transformations (cont.):Insert Fig. 13.11Vector QuantizationVector Quantization (cont.)Properties of Vector QuantizationVQ Removes Linear DependencyVQ Removes Linear Dependency (cont.)VQ Removes Nonlinear DependencyVQ Exploits the Increase in DimensionalityCodebook Design AlgorithmsCodebook Design via K-meansCodebook Design via K-means (cont.)Complexity of K-meansTree Codebook and Binary SearchComplexity of Tree CodebookNearest Neighbor Design AlgorithmVariations of VQVariations of VQ (cont.)Slide Number 137Second Generation Image CodingSecond Generation Image Coding (cont.)Comparison of Image Coding MethodsInterframe Image CodingMotion EstimationRegion Matching TechniqueRegion Matching Technique (cont.)Cross Correlation MethodLogarithmic SearchMSE as Region Matching CriterionRecursive Motion EstimationRecursive Motion Estimation (cont.)Frequency Domain Methods for Motion EstimationMethod of Differential for Motion EstimationMethod of Differential for Motion Estimation (cont.)Color Image CodingSlide Number 154Slide Number 155JPEG StandardJPEG Standard (cont.)JPEGEntropy CodingJPEG MODESJPEG MODES (cont.)MPEG StandardMPEGMPEG 1Picture Types:Picture Types (cont.)Bidirectional Motion CompensationBlock Diagram of MPEG EncodesMPEG 2:Profiles + LevelsSlide Number 171ScalabilityH. 261 Video CodingH. 261Objectives of Image Coding• Representation of an image with acceptable quality, using as small a number of bits as possibleApplications:• Reduction of channel bandwidth for image transmission• Reduction of required storageImage CoderChannel decoderTransmitterImage SourceChannelencoderChannelImage decoderReceiverReconstructed ImageFigure 10.1 Typical environment for image coding.Objectives of Image Coding cont.Issues in Image Coding1. What to code?a. Image densityb. Image transform coefficientsc. Image model parameters2. How to assign reconstruction levelsa. Uniform spacing between reconstruction levelsb. Non-uniform spacing between reconstruction levels3. Bit assignmenta. Equal-length bit assignment to each reconstruction levelb. Unequal-length bit assignment to each reconstruction levelImage sourceTransformationQuantizationCodewordassignmentString of bitsFigure 10.2 Three major components in image coding.Issues in Image Coding cont.Methods of Reconstruction Level AssignmentsAssumptions:• Image intensity is to be coded• Equal-length bit assignmentScalar Case1. Equal spacing of reconstruction levels (Uniform Quantization)(Ex): Image intensity f: 0 ~ 255f !!^ff !!^fUniformQuantizerMethods of Reconstruction Level Assignments cont.Number of reconstruction levels: 4 (2 bits for equal bit assignment)0 32 64 96 128 160 192 224 255Methods of Reconstruction Level Assignments cont.Figure 10.3Example of uniform quantizer. The number of reconstruction levels is 4, is assumed to be between 0 and 1, and is the result of quantizing . The reconstruction levels and decision boundaries are denoted by ri and di , respectively.Scalar Case (cont.)2. Spacing based on some error criterionri : reconstruction levels (32, 96, 160, 224)di : decision boundaries (0, 64, 128, 192, 256) Optimally choose ri and di .To do this, assume J the number of reconstruction levels: probability density function for fMinimizeri , di , : = ==> Lloyd-Max QuantizerThese are not simple linear equations.Scalar Case (cont.)2. Spacing based on some error criterionri : reconstruction levels (32, 96, 160, 224)di : decision boundaries (0, 64, 128, 192, 256) Optimally choose ri and di .To do this, assume J the number of