A Fast LBG Codebook Training Algorithm for Vector Quantization Presented by 蔡進義.

Slides:



Advertisements
Similar presentations
K-MEANS Michael Jones ENEE698Q Fall Overview  Introduction  Problem Formulation  How K-Means Works  Pros and Cons of Using K-Means  How to.
Advertisements

K-means method for Signal Compression: Vector Quantization
Fast Algorithm for Nearest Neighbor Search Based on a Lower Bound Tree Yong-Sheng Chen Yi-Ping Hung Chiou-Shann Fuh 8 th International Conference on Computer.
Coherency Sensitive Hashing (CSH) Simon Korman and Shai Avidan Dept. of Electrical Engineering Tel Aviv University ICCV2011 | 13th International Conference.
CSE 589 Applied Algorithms Spring 1999 Image Compression Vector Quantization Nearest Neighbor Search.
Reconstruction of Lost Blocks Using Codeword Estimation Kuo-Lung Hung, Chin-Chen Chang, Tung-Shou Chen IEEE Transactions on Consumer Electronics, Nov.
Vector Quantization. 2 outline Introduction Two measurement : quality of image and bit rate Advantages of Vector Quantization over Scalar Quantization.
Losslessy Compression of Multimedia Data Hao Jiang Computer Science Department Sept. 25, 2007.
Fast VQ Encoding by an Efficient Kick-Out Condition Kuang-Shyr Wu and Ja-Chen Lin IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY, FEB.
A Multiresolution Symbolic Representation of Time Series
1 Embedded colour image coding for content-based retrieval Source: Journal of Visual Communication and Image Representation, Vol. 15, Issue 4, December.
Image Compression Using Neural Networks Vishal Agrawal (Y6541) Nandan Dubey (Y6279)
Face Recognition Using Neural Networks Presented By: Hadis Mohseni Leila Taghavi Atefeh Mirsafian.
Fast vector quantization image coding by mean value predictive algorithm Authors: Yung-Gi Wu, Kuo-Lun Fan Source: Journal of Electronic Imaging 13(2),
CS 395 T Real-Time Graphics Architectures, Algorithms, and Programming Systems Spring’03 Vector Quantization for Texture Compression Qiu Wu Dept. of ECE.
A Memory-efficient Huffman Decoding Algorithm
Adaptive Multi-path Prediction for Error Resilient H.264 Coding Xiaosong Zhou, C.-C. Jay Kuo University of Southern California Multimedia Signal Processing.
Fast search methods Pasi Fränti Clustering methods: Part 5 Speech and Image Processing Unit School of Computing University of Eastern Finland
1 An Efficient VQ-based Data Hiding Scheme Using Voronoi Clustering Authors:Ming-Ni Wu, Puu-An Juang, and Yu-Chiang Li.
VQ for ASR 張智星 多媒體資訊檢索實驗室 清華大學 資訊工程系.
Date: Advisor: Jian-Jung Ding Reporter: Hsin-Hui Chen.
Chapter 9 DTW and VQ Algorithm  9.1 Basic idea of DTW  9.2 DTW algorithm  9.3 Basic idea of VQ  9.4 LBG algorithm  9.5 Improvement of VQ.
Planar-Oriented Ripple Based Greedy Search Algorithm for Vector Quantization Presenter: Tzu-Meng Huang Adviser:Dr. Yeou-Jiunn Chen Date:2011/11/ /11/161.
Advances in digital image compression techniques Guojun Lu, Computer Communications, Vol. 16, No. 4, Apr, 1993, pp
1 Information Hiding Based on Search Order Coding for VQ Indices Source: Pattern Recognition Letters, Vol.25, 2004, pp.1253 – 1261 Authors: Chin-Chen Chang,
Palette Partition Based Data Hiding for Color Images Yu-Chiang Li, Piyu Tsai, Chih-Hung Lin, Hsiu-Lien Yeh, and Chien-Ting Huang Speaker : Yu-Chiang Li.
A Multiresolution Symbolic Representation of Time Series Vasileios Megalooikonomou Qiang Wang Guo Li Christos Faloutsos Presented by Rui Li.
Vector Quantization Vector quantization is used in many applications such as image and voice compression, voice recognition (in general statistical pattern.
Vector Quantization CAP5015 Fall 2005.
Faculty of Information Engineering, Shenzhen University Liao Huilian SZU TI-DSPs LAB Aug 27, 2007 Optimizer based on particle swarm optimization and LBG.
2016/2/171 Image Vector Quantization Indices Recovery Using Lagrange Interpolation Source: IEEE International Conf. on Multimedia and Expo. Toronto, Canada,
Rate Distortion Theory. Introduction The description of an arbitrary real number requires an infinite number of bits, so a finite representation of a.
Image Compression Using Address-Vector Quantization NASSER M. NASRABADI, and YUSHU FENG Presented by 蔡進義 P IEEE TRANSACTIONS ON COMMUNICATIONS,
Efficient Huffman Decoding Aggarwal, M. and Narayan, A., International Conference on Image Processing, vol. 1, pp. 936 – 939, 2000 Presenter :Yu-Cheng.
S.R.Subramanya1 Outline of Vector Quantization of Images.
An Image Database Retrieval Scheme Based Upon Multivariate Analysis and Data Mining Presented by C.C. Chang Dept. of Computer Science and Information.
Image Compression using Vector Quantization
Chapter 3 向量量化編碼法.
A New Image Compression Scheme Based on Locally Adaptive Coding
Source: ACM International Conference Proceding
A Color Image Hiding Scheme Based on SMVQ and Modulo Operator
Source :Journal of visual Communication and Image Representation
Chair Professor Chin-Chen Chang Feng Chia University
High-capacity image hiding scheme based on vector quantization
Advisor: Chin-Chen Chang1, 2 Student: Wen-Chuan Wu2
A Data Hiding Scheme Based Upon Block Truncation Coding
Clustering 77B Recommender Systems
第七章 資訊隱藏 張真誠 國立中正大學資訊工程研究所.
Foundation of Video Coding Part II: Scalar and Vector Quantization
Advisor: Chin-Chen Chang1, 2 Student: Yi-Pei Hsieh2
An Innovative Steganographic Scheme Based on Vector Quantization
An Innovative Steganographic Scheme Based on Vector Quantization
第 四 章 VQ 加速運算與編碼表壓縮 4-.
Density-Based Image Vector Quantization Using a Genetic Algorithm
Dynamic embedding strategy of VQ-based information hiding approach
Chair Professor Chin-Chen Chang Feng Chia University
Hung, K. -L. and Chang, C. -C. , IEE Image and Signal Processing, vol
A Self-Reference Watermarking Scheme Based on Wet Paper Coding
A Color Image Hiding Scheme Based on SMVQ and Modulo Operator
Hiding Information in VQ Index Tables with Reversibility
Information Hiding and Its Applications
Zhe-Ming Lu, Chun-He Liu, Dian-Guo Xu, Sheng-He Sun,
Authors: Chin-Chen Chang, Yi-Hui Chen, and Chia-Chen Lin
Efficient Huffman Decoding
A Virtual Image Cryptosystem Based upon Vector Quantization
Hung, K. -L. and Chang, C. -C. , IEE Image and Signal Processing, vol
A Self-Reference Watermarking Scheme Based on Wet Paper Coding
Predictive Grayscale Image Coding Scheme Using VQ and BTC
資訊偽裝術 張真誠 講座教授 多媒體暨網路安全實驗室
A New Image Compression Scheme Based on Locally Adaptive Coding
Presentation transcript:

A Fast LBG Codebook Training Algorithm for Vector Quantization Presented by 蔡進義

2 Motivation A fast codebook-training algorithm based on LBG algorithm. A fast codebook-training algorithm based on LBG algorithm. To reduce the computational cost in the codebook training processes. To reduce the computational cost in the codebook training processes.

3 Outline Introduction Introduction Previous Works Previous Works Proposed Method Proposed Method Some Experiments Some Experiments Discussions and Conclusions Discussions and Conclusions

4 Image Compression techniques Block truncation coding Block truncation coding Transform coding Transform coding Hybrid coding Hybrid coding Vector quantization Vector quantization Simple structure and low bit rate Simple structure and low bit rate

5 VQ scheme The VQ scheme can be divided into three parts: The VQ scheme can be divided into three parts: Codebook generation Codebook generation Encoding procedure Encoding procedure Decoding procedure Decoding procedure encoding decoding CodebookCodebook

6 Codebook Generation The most important task for VQ scheme is to design a good codebook. The most important task for VQ scheme is to design a good codebook. LBG (Linde-Buzo-Gray) algorithm / Lloyd clustering algorithm LBG (Linde-Buzo-Gray) algorithm / Lloyd clustering algorithm The LBG algorithm is an iterative procedure. The LBG algorithm is an iterative procedure. cb 0 cb 1 cb n …

7 Euclidean Distance The dimensionality of vector = k (= w*h) The dimensionality of vector = k (= w*h) An input vector x = (x 1, x 2, …, x k ) An input vector x = (x 1, x 2, …, x k ) A codeword y i = (y i1, y i2, …, y ik ) A codeword y i = (y i1, y i2, …, y ik ) The Euclidean distance between x and y i The Euclidean distance between x and y i

8 Codebook Generation

9 Codebook generation Codebook generation Training Images Training set N-1 N VQ Codebook Training

10 Codebook generation Codebook generation Training set Codebook initiation Initial codebook N-1 N VQ Codebook Training

11 Image Index table Vector Quantization Encoder w h Image compression technique VQ Encoding Procedure

12 Image Index table Vector Quantization Decoder w h Image compression technique VQ Decoding Procedure

13 Codebook search To reduce the computational cost for the segmentation procedure in the LBG algorithm, many fast algorithms for codebook search have been developed. To reduce the computational cost for the segmentation procedure in the LBG algorithm, many fast algorithms for codebook search have been developed. Partial Distortion Search (PDS) Partial Distortion Search (PDS) Mean-distance-ordered Partial Codebook Search (MPS) Mean-distance-ordered Partial Codebook Search (MPS) Integral Projection Mean-sorted Partial Search (IPMPS) Integral Projection Mean-sorted Partial Search (IPMPS)

14 Another fast codebook design The tree-structured VQ (TSVQ) The tree-structured VQ (TSVQ) The k-d tree VQ The k-d tree VQ

15 Outline Introduction Introduction Previous Works Previous Works Proposed Method Proposed Method Some Experiments Some Experiments Discussions and Conclusions Discussions and Conclusions

16 Goal To reduce the computation cost in finding the closest codeword in the codebook. To reduce the computation cost in finding the closest codeword in the codebook. PDS PDS MPS MPS IPMPS IPMPS

17 Partial distortion search (PDS) Closest codeword search Closest codeword search If the minimal distance of each input vector could not be found early, the PDS method can just reduce little computation time. If the minimal distance of each input vector could not be found early, the PDS method can just reduce little computation time. (a0, a1, a2, …, a15) input vector (b0, b1, b3, …, b15) codeword

18 Mean-distance-ordered Partial Codebook Search Algorithm (MPS) The Squared Euclidean Distance (SED) The Squared Euclidean Distance (SED) The Squared Mean Distance (SMD) The Squared Mean Distance (SMD) The minimal SED codeword is usually in the neighborhood of the minimal SMD codeword. The minimal SED codeword is usually in the neighborhood of the minimal SMD codeword.

19 Mean-distance-ordered Partial Codebook Search Algorithm (MPS) SMD SED reject

20 Integral Projection Mean-sorted Partial Search Algorithm (IPMPS) Based on multiple distortion measures with different levels of computational complexity. Based on multiple distortion measures with different levels of computational complexity. Three kinds of integral projections: Three kinds of integral projections:

21 Integral Projection Mean-sorted Partial Search Algorithm (IPMPS) Three distortion measures: Three distortion measures: For each codeword Y i Test conditions

22 Outline Introduction Introduction Previous Works Previous Works Proposed Method Proposed Method Some Experiments Some Experiments Discussions and Conclusions Discussions and Conclusions

23 Generalized Integral Projection Model (GIP) To reduce the computational cost To reduce the computational cost MPS and IPMPS MPS and IPMPS IPMPS employs the concept of integral projection to reject further codeword in search. IPMPS employs the concept of integral projection to reject further codeword in search.

24 Generalized Integral Projection Model (GIP) 1. Initially, choose one possible projection map of the pair (p, q).  p segments with q pixels in each segment 2. For each input vector, compute the projection P X (k) of these p segments. 3. The distortion measure corresponding to this projection map is defined as:

25 Generalized Integral Projection Model (GIP) 4. For each codeword, the following inequality can be easily proven true 5. The test condition for this projection map can be constructed. pair(p, q) test conditionpossible projection map

26 Segment maps

27 Fast LBG Algorithm 1. Initially, select a set of test conditions by repeatedly applying the GIP model with different projection maps of the desired pair (p, q). 2. Sort the current codebook by the mean values of the codewords. 3. For each vector, find the corresponding closest codeword.

28 Fast LBG Algorithm 4. Record the index of the closest codeword for each training vector. 5. Update each codeword 6. Overall averaged distortion

29 Outline Introduction Introduction Previous Works Previous Works Proposed Method Proposed Method Some Experiments Some Experiments Discussions and Conclusions Discussions and Conclusions

30 Experiment Methods 512*512 image LBG PDS MPS

31 Experiment Results the property of the training set FLBG-1a FLBG-1b

32 Outline Introduction Introduction Previous Works Previous Works Proposed Method Proposed Method Some Experiments Some Experiments Discussions and Conclusions Discussions and Conclusions

33 Conclusions A generalized integral projection model is developed to produce the test conditions for the speedup of the search process for the VQ codebook design. A generalized integral projection model is developed to produce the test conditions for the speedup of the search process for the VQ codebook design. To use these test conditions to eliminate the need of calculating the squared Euclidean distance. To use these test conditions to eliminate the need of calculating the squared Euclidean distance. The property of image The property of image By choosing proper sets of test conditions for different training sets, a great deal of computation cost can be reduced. By choosing proper sets of test conditions for different training sets, a great deal of computation cost can be reduced.