CURVELETS USING RIDGELETS

Slides:



Advertisements
Similar presentations
Shapelets Correlated with Surface Normals Produce Surfaces Peter Kovesi School of Computer Science & Software Engineering The University of Western Australia.
Advertisements

Chapter 11 Signal Processing with Wavelets. Objectives Define and illustrate the difference between a stationary and non-stationary signal. Describe the.
11/11/02 IDR Workshop Dealing With Location Uncertainty in Images Hasan F. Ates Princeton University 11/11/02.
3-D Computer Vision CSc83020 / Ioannis Stamos  Revisit filtering (Gaussian and Median)  Introduction to edge detection 3-D Computater Vision CSc
Introduction to the Curvelet Transform
Applications in Signal and Image Processing
© by Yu Hen Hu 1 ECE533 Digital Image Processing Image Enhancement in Frequency Domain.
Image Processing A brief introduction (by Edgar Alejandro Guerrero Arroyo)
CS 691 Computational Photography
Chapter 8 Content-Based Image Retrieval. Query By Keyword: Some textual attributes (keywords) should be maintained for each image. The image can be indexed.
Sampling, Aliasing, & Mipmaps
Oriented Wavelet 國立交通大學電子工程學系 陳奕安 Outline Background Background Beyond Wavelet Beyond Wavelet Simulation Result Simulation Result Conclusion.
Lecture05 Transform Coding.
Communication & Multimedia C. -H. Hong 2015/6/12 Contourlet Student: Chao-Hsiung Hong Advisor: Prof. Hsueh-Ming Hang.
Fourier Transform Analytic geometry gives a coordinate system for describing geometric objects. Fourier transform gives a coordinate system for functions.
Wavelet Transform 國立交通大學電子工程學系 陳奕安 Outline Comparison of Transformations Multiresolution Analysis Discrete Wavelet Transform Fast Wavelet Transform.
Basic Concepts and Definitions Vector and Function Space. A finite or an infinite dimensional linear vector/function space described with set of non-unique.
Hui Kong, Member, IEEE, Jean- Yves Audibert, and Jean Ponce, Fellow, IEEE.
2D Fourier Theory for Image Analysis Mani Thomas CISC 489/689.
Introduction to Wavelets
J OURNAL C LUB : Yang and Ni, Xidian University, China “Multimodality medical image fusion based on multiscale geometric analysis of contourlet transform.”
Back Projection Reconstruction for CT, MRI and Nuclear Medicine
Representation and Compression of Multi-Dimensional Piecewise Functions Dror Baron Signal Processing and Systems (SP&S) Seminar June 2009 Joint work with:
Multiscale transforms : wavelets, ridgelets, curvelets, etc.
ENG4BF3 Medical Image Processing
Image Representation Gaussian pyramids Laplacian Pyramids
The Wavelet Tutorial: Part3 The Discrete Wavelet Transform
CAP5415: Computer Vision Lecture 4: Image Pyramids, Image Statistics, Denoising Fall 2006.
Discrete Images (Chapter 7) Fourier Transform on discrete and bounded domains. Given an image: 1.Zero boundary condition 2.Periodic boundary condition.
Wavelet transform Wavelet transform is a relatively new concept (about 10 more years old) First of all, why do we need a transform, or what is a transform.
DCT.
Computer Vision Spring ,-685 Instructor: S. Narasimhan Wean 5403 T-R 3:00pm – 4:20pm.
Wavelets and Multiresolution Processing (Wavelet Transforms)
CS654: Digital Image Analysis Lecture 36: Feature Extraction and Analysis.
Week 11 – Spectral TV and Convex analysis Guy Gilboa Course
By Dr. Rajeev Srivastava CSE, IIT(BHU)
Projects Project 1a due this Friday Project 1b will go out on Friday to be done in pairs start looking for a partner now.
Fourier transform.
Miguel Tavares Coimbra
Jean Baptiste Joseph Fourier
MAIN PROJECT IMAGE FUSION USING MATLAB
Image Resampling & Interpolation
Linear Filters and Edges Chapters 7 and 8
- photometric aspects of image formation gray level images
Multiresolution Analysis (Chapter 7)
… Sampling … … Filtering … … Reconstruction …
Multi-resolution image processing & Wavelet
Linear Filters and Edges Chapters 7 and 8
Degradation/Restoration Model
Image Sampling Moire patterns
Wavelets : Introduction and Examples
Filtering Geophysical Data: Be careful!
Multi-resolution analysis
Fourier Transform.
Fourier Transform Analytic geometry gives a coordinate system for describing geometric objects. Fourier transform gives a coordinate system for functions.
Fourier Transform Analytic geometry gives a coordinate system for describing geometric objects. Fourier transform gives a coordinate system for functions.
Image Sampling Moire patterns
Jeremy Bolton, PhD Assistant Teaching Professor
Computer Vision Lecture 9: Edge Detection II
Image Transforms for Robust Coding
Wavelet transform Wavelet transform is a relatively new concept (about 10 more years old) First of all, why do we need a transform, or what is a transform.
Filtering Part 2: Image Sampling
Image Sampling Moire patterns
Lecture 5: Resampling, Compositing, and Filtering Li Zhang Spring 2008
Fourier Transform Analytic geometry gives a coordinate system for describing geometric objects. Fourier transform gives a coordinate system for functions.
Image Resampling & Interpolation
Resampling.
Wavelet transform application – edge detection
Lecture 7 Spatial filtering.
Image restoration, noise models, detection, deconvolution
Presentation transcript:

CURVELETS USING RIDGELETS BY: RON GRINFELD י"ט/כסלו/תשע"ט

POINTS OF DISCUSSION INTRODUCTION POINT DISCONTINUITIES FAILURE OF WAVELETS ON EDGES MOTIVATION THE CURVELET TRANSFORM ANALYSIS SUMMARY EXAMPLES י"ט/כסלו/תשע"ט

EDGE DEFINITION PHYSICAL – ONE OBJECT OCCLUDES ANOTHER OBJECT INTRODUCTION EDGE DEFINITION PHYSICAL – ONE OBJECT OCCLUDES ANOTHER OBJECT GEOMETRIC – DISCONTINUITIES ALONG CURVES IMAGE PROCESSING – LUMINANCE UNDERGOING STEP DISCONTINUITIES AT BOUNDRIES י"ט/כסלו/תשע"ט

POINT DISCONTINUITIES INTRODUCTION POINT DISCONTINUITIES IDEAL REPRESENTATION: HOW TO DO THIS ? -ASK AN ORACLE -DETECT THE EDGES IS IT POSSIBLE ? -NOISY AND BLURRED DATA -RESOURCES TO AN ORACLE י"ט/כסלו/תשע"ט

WAVELETS AND POINT DISCONTINUITIES INTRODUCTION WAVELETS AND POINT DISCONTINUITIES -WAVELETS USE DIADIC SCALING -EACH SCALING SQUARE SIZE: -TO GET A SCALE OF RATE 1/n ONE NEEDS TO PERFORM n STAGES OF THE WAVELET PYRAMID -AT EACH STEP ONLY A FEW WAVELETS (C) “FEEL” THE POINT DISCONTINUITY י"ט/כסלו/תשע"ט

POINT DISCONTINUITIES ALONG STARIGHT LINES INTRODUCTION CONCLUSION: WAVELETS NEED TO KEEP A FACTOR OF ONLY MORE DATA THEN THE IDEAL REPRESENTATION WHEN HANDLING POINT DISCONTINUITIES POINT DISCONTINUITIES ALONG STARIGHT LINES י"ט/כסלו/תשע"ט

INTRODUCTION Example in 2-D The function: י"ט/כסלו/תשע"ט

FALIURE OF WAVELETS ON EDGES INTRODUCTION FALIURE OF WAVELETS ON EDGES is smooth away from a discontinuity along a curve Note that this defines a line (2-D) discontinuity (edge), and not a point discontinuity At stage j of the wavelet pyramid: squares, size: “feel” the discontinuity Along י"ט/כסלו/תשע"ט

wavelet coeffs needed, each size N’th largest coeff’s size INTRODUCITON wavelet coeffs needed, each size N’th largest coeff’s size Rate of approximation י"ט/כסלו/תשע"ט

MOTIVATION TO GET THE BEST RATE OF APPROXIMATION RATE OF APPROXIMATION: BY TAKING THE BEST m TERMS OF THE TRANSFORM WE WANT THE SMALLEST ERROR RATE Fourier: Wavelets: Curvelets: ? י"ט/כסלו/תשע"ט

THE CURVELET TRANSFORM CvT CvT INCLUDES 4 STAGES: -SUB-BAND DECOMPOSITION -SMOOTH PARTITIONING -RENORMALIZATION -RIDGELET ANALYSIS י"ט/כסלו/תשע"ט

THE CURVELET TRANSFORM CvT SUB-BAND DECOMPOSITION -THE IMAGE IS DEVIDED INTO s RESOLUTION LAYERS BY A BANK OF SUB-BAND FILTERS: - (ALSO CALLED 0) IS A LOW PASS FILTER AND DEALS WITH FREQUENCIES NEAR ||1 - DEFINED AS: 2s(x) = 24s (22sx) DEALS WITH FREQUENCIES NEAR ||[22s, 22s+2] -THUS EACH SUB-BAND CONTAINS WIDE DETAILS -THE SUB-BAND DECOMPOSITION IS APPLYING A CONVOLUTION OPERATOR: י"ט/כסלו/תשע"ט

THE CURVELET TRANSFORM CvT SUB-BAND DECOMPOSITION USING THE WAVELET TRANSFORM TO APPROXIMATE SUB-BAND DECOMPOSITION -Using wavelet transform, f is decomposed into S0, D1, D2, D3, … -P0 f is partially constructed from S0 and D1, and may include also D2 and D3 -s f is constructed from D2s and D2s+1 י"ט/כסלו/תשע"ט

SUB-BAND DECOMPOSITION THE CURVELET TRANSFORM CvT SUB-BAND DECOMPOSITION י"ט/כסלו/תשע"ט

THE CURVELET TRANSFORM CvT P0 f IS “SMOOTH” (NO 2-D EDGES) AND THUS CAN BE REPRESENTED USING WAVELETS ( ) BUT WHAT ABOUT THE DISCONTINUITIES ALONG THE CURVES REPRESENTED IN THE LAYERS s f ? NEXT STEP: SMOOTH PARTITIONING DIVIDING THE LAYERS INTO SQUARES IN A SPEACIAL WAY י"ט/כסלו/תשע"ט

SMOOTH PARTITIONING THE TRICK: THE SUB-BAND FILTERING CAUSED THE EDGES IN LAYER S TO BE WIDE WE WILL SEE A WAY TO DIVIDE THE LAYER INTO SIZE SQUARES, IN A SMART WAY THAT AVOIDS DAMAGING THE EDGES BY THE PARTITION THIS WILL RESOLVE IN THE FOLLOWING ASPECT RATIO OF THE EDGES: width  length2 AND WILL PRODUCE LONG,THIN AND DIRECTION ORIENTED EDGES, TO BE HANDLED BY RIDGELETS י"ט/כסלו/תשע"ט

SMOOTH PARTITIONING DEFINE THE GRID OF DYADIC SQUARES Assume w be a smooth windowing function with ‘main’ support of size 2-s2-s. For each square, wQ is a displacement of w localized near Q Multiplying s f with wQ (QQs) produces a smooth dissection of the function into ‘squares’ י"ט/כסלו/תשע"ט

SMOOTH PARTITIONING The windowing function w is a nonnegative smooth function ENERGY PARTITION: The energy of certain pixel (x1,x2) is divided between all sampling windows of the grid י"ט/כסלו/תשע"ט

SMOOTH PARTITIONING ENERGY PARTITION RECONSTRUCTION PARSERVAL RELATION י"ט/כסלו/תשע"ט

SMOOTH PARTITIONING י"ט/כסלו/תשע"ט

RENORMALIZATION Renormalization is centering each dyadic square to the unit square [0,1][0,1] For each Q, the operator TQ is defined as: Each square is renormalized: י"ט/כסלו/תשע"ט

RIDGELETS – THE FINAL STAGE REMEMBER THE RATIO: width  length2 ? WE HAVE ACHIVED IT, AND NOW WE NEED A SET OF WAVELET BASED FUNCTIONS OF WHICH CONTAIN BOTH ANGULAR AND RADIAL LOCATIONS, AND CAN ENJOY THE BENEFIT OF THE RATIO width  length2 THESE FUNCTIONS ARE CALLED RIDGELETS י"ט/כסלו/תשע"ט

RIDGELETS The ridgelet element has a formula in the frequency domain: where is index to the ridge scale is the location, and are the angular scale and location of the periodic wavelets  on the radon domain [-,  ) where j,k are Meyer wavelets for  י"ט/כסלו/תשע"ט

RIDGELET TILLING י"ט/כסלו/תשע"ט

The energy of the input square sized: And scaled: is defined as: Coefficient’s amplitude N-th largest curvlet coeff. size Letting denote the N-th coeff’s amplitude We get: י"ט/כסלו/תשע"ט

REMEMBER OWER MOTIVATION? TO GET THE BEST RATE OF APPROXIMATION RATE OF APPROXIMATION: BY TAKING THE BEST m TERMS OF THE TRANSFORM WE WANT THE SMALLEST ERROR RATE Fourier: Wavelets: Curvelets: ? י"ט/כסלו/תשע"ט

REMEMBER OWER MOTIVATION? TO GET THE BEST RATE OF APPROXIMATION RATE OF APPROXIMATION: BY TAKING THE BEST m TERMS OF THE TRANSFORM WE WANT THE SMALLEST ERROR RATE Fourier: Wavelets: Curvelets: י"ט/כסלו/תשע"ט

LET US SUMMARIZE THE MAIN CONCEPTS OF THE CURVELET SYSTEM ANISOTROPIC SCALING – CREATING A RATIO OF width  length2 AND BY THAT, CONCENTRAITING THE EDGES, MAKING THEM THIN, STRAIGHT AND DIRECTED THE RIDGELET SYSTEM – A FORMULA CONSTRUCTED OF WAVELETS FITTED TO SCALE AND LOCATION IN BOTH  AND RADON DOMAINS, RECEIVING ANGLE SCALE AND LOCATION AS INPUT י"ט/כסלו/תשע"ט

SUMMARIZE THE TEXTURE (LUMINANCE) OF NATURAL AND SYNTHETIC IMAGES CAN BE DESCRIBED AS A COLLECTION OF CURVES AND POINTS THE CURVELET SYSTEM DIVIDES IMAGES TO POINTS AND LINES (APPROXIMATING CURVES) AND HANDLES THE POINTS BY WAVELETS AND THE LINES BY RIDGELETS י"ט/כסלו/תשע"ט

ILUSTRATING THE PRINCIPLE: image = points + lines (curves) sparse image representation can be achieved by: curvelets = wavelets(points) + ridgelets(lines) Original image Ridgelets part of the image Wavelet part of the noisy image י"ט/כסלו/תשע"ט

LESS THEN 5% OF THE COEFFICIENTS AND YET SO MUCH INFORMATION ABOUT THE EDGES DEFINING THE OBJECTS IN THE IMAGES י"ט/כסלו/תשע"ט

THE END י"ט/כסלו/תשע"ט