Bivariate Splines for Image Denoising*° *Grant Fiddyment University of Georgia, 2008 °Ming-Jun Lai Dept. of Mathematics University of Georgia.

Slides:



Advertisements
Similar presentations
Multimedia Data Compression
Advertisements

Video We will consider how electronic pictures (monochrome) are made up. Consider some important characteristics of electronic images. See how image are.
Inpainting Assigment – Tips and Hints Outline how to design a good test plan selection of dimensions to test along selection of values for each dimension.
Boundary Detection - Edges Boundaries of objects –Usually different materials/orientations, intensity changes.
November 12, 2013Computer Vision Lecture 12: Texture 1Signature Another popular method of representing shape is called the signature. In order to compute.
Topic 6 - Image Filtering - I DIGITAL IMAGE PROCESSING Course 3624 Department of Physics and Astronomy Professor Bob Warwick.
EDGE DETECTION ARCHANA IYER AADHAR AUTHENTICATION.
Math Modeling Final Project APPLICATIONS of FRACTALS Advisor: Professor Alber Fang Qi Pu Wan Xue Rui FRACTAL LANDSCAPES FRACTAL IMAGE COMPRESSION.
Computer Vision Group Digital Image Filters 26 th November 2012.
Katsuyo Thornton*, R. Edwin García✝, Larry Aagesen*
Edge detection Goal: Identify sudden changes (discontinuities) in an image Intuitively, most semantic and shape information from the image can be encoded.
School of Computing Science Simon Fraser University
ANOVA notes NR 245 Austin Troy
1Ellen L. Walker Segmentation Separating “content” from background Separating image into parts corresponding to “real” objects Complete segmentation Each.
January 19, 2006Computer Vision © 2006 Davi GeigerLecture 1.1 Image Measurements and Detection Davi Geiger
Multimedia Data Introduction to Image Processing Dr Mike Spann Electronic, Electrical and Computer.
1 Preprocessing for JPEG Compression Elad Davidson & Lilach Schwartz Project Supervisor: Ari Shenhar SPRING 2000 TECHNION - ISRAEL INSTITUTE of TECHNOLOGY.
Edges and Scale Today’s reading Cipolla & Gee on edge detection (available online)Cipolla & Gee on edge detection Szeliski – From Sandlot ScienceSandlot.
Spatial Analysis Longley et al., Ch 14,15. Transformations Buffering (Point, Line, Area) Point-in-polygon Polygon Overlay Spatial Interpolation –Theissen.
Digital Image Processing Chapter 5: Image Restoration.
Losslessy Compression of Multimedia Data Hao Jiang Computer Science Department Sept. 25, 2007.
Image Enhancement.
VISUALIZATION OF HYPERSPECTRAL IMAGES ROBERTO BONCE & MINDY SCHOCKLING iMagine REU Montclair State University.
Edge Detection Today’s readings Cipolla and Gee –supplemental: Forsyth, chapter 9Forsyth Watt, From Sandlot ScienceSandlot Science.
E.G.M. PetrakisBinary Image Processing1 Binary Image Analysis Segmentation produces homogenous regions –each region has uniform gray-level –each region.
Automatic Estimation and Removal of Noise from a Single Image
כמה מהתעשייה? מבנה הקורס השתנה Computer vision.
Shape Blending Joshua Filliater December 15, 2000.
By Meidika Wardana Kristi, NRP  Digital cameras used to take picture of an object requires three sensors to store the red, blue and green color.
IMAGE SAMPLING AND IMAGE QUANTIZATION 1. Introduction
Texture. Texture is an innate property of all surfaces (clouds, trees, bricks, hair etc…). It refers to visual patterns of homogeneity and does not result.
SIGNAL DETECTION IN FIXED PATTERN CHROMATIC NOISE 1 A. J. Ahumada, Jr., 2 W. K. Krebs 1 NASA Ames Research Center; 2 Naval Postgraduate School, Monterey,
Multimedia Data Introduction to Image Processing Dr Sandra I. Woolley Electronic, Electrical.
1 Lecture 1 1 Image Processing Eng. Ahmed H. Abo absa
Linear Regression Least Squares Method: an introduction.
Image Processing Edge detection Filtering: Noise suppresion.
1 COMS 161 Introduction to Computing Title: Digital Images Date: November 12, 2004 Lecture Number: 32.
GEOSTATISICAL ANALYSIS Course: Special Topics in Remote Sensing & GIS Mirza Muhammad Waqar Contact: EXT:2257.
Digital image processing Chapter 3. Image sampling and quantization IMAGE SAMPLING AND IMAGE QUANTIZATION 1. Introduction 2. Sampling in the two-dimensional.
December 9, 2014Computer Vision Lecture 23: Motion Analysis 1 Now we will talk about… Motion Analysis.
Digital Image Processing Lecture 10: Image Restoration March 28, 2005 Prof. Charlene Tsai.
23 November Md. Tanvir Al Amin (Presenter) Anupam Bhattacharjee Department of Computer Science and Engineering,
VI. Regression Analysis A. Simple Linear Regression 1. Scatter Plots Regression analysis is best taught via an example. Pencil lead is a ceramic material.
Digital Image Processing Lecture 10: Image Restoration
Course 2 Image Filtering. Image filtering is often required prior any other vision processes to remove image noise, overcome image corruption and change.
Correlation & Regression Analysis
Image processing Honza Černocký, ÚPGM. 1D signal 2.
A Tutorial on using SIFT Presented by Jimmy Huff (Slightly modified by Josiah Yoder for Winter )
October 1, 2013Computer Vision Lecture 9: From Edges to Contours 1 Canny Edge Detector However, usually there will still be noise in the array E[i, j],
1 Mathematic Morphology used to extract image components that are useful in the representation and description of region shape, such as boundaries extraction.
Digital Image Processing Lecture 17: Segmentation: Canny Edge Detector & Hough Transform Prof. Charlene Tsai.
Biostatistics Regression and Correlation Methods Class #10 April 4, 2000.
Edges Edges = jumps in brightness/color Brightness jumps marked in white.
12/12/2003EZW Image Coding Duarte and Haupt 1 Examining The Embedded Zerotree Wavelet (EZW) Image Coding Method Marco Duarte and Jarvis Haupt ECE 533 December.
1. 2 What is Digital Image Processing? The term image refers to a two-dimensional light intensity function f(x,y), where x and y denote spatial(plane)
Biointelligence Laboratory, Seoul National University
Katsuyo Thornton1, R. Edwin García2, Larry Aagesen3
WAVELET VIDEO PROCESSING TECHNOLOGY
The Lease Squares Line Finite 1.3.
Digital Image Processing
Image quantization By Student Manar naji. quantization The values obtained by sampling a continuous function usually comprise of an infinite set or.
Pei Qi ECE at UW-Madison
Edge detection Goal: Identify sudden changes (discontinuities) in an image Intuitively, most semantic and shape information from the image can be encoded.
From a presentation by Jimmy Huff Modified by Josiah Yoder
APPLICATIONS of FRACTALS
Product moment correlation
Lecture 2: Edge detection
Wavelet transform application – edge detection
Presentation transcript:

Bivariate Splines for Image Denoising*° *Grant Fiddyment University of Georgia, 2008 °Ming-Jun Lai Dept. of Mathematics University of Georgia

Signals A signal is an observable quantity that varies over time or space. Thus it is parameterizable. The process of sampling involves transforming a continuous signal to a discrete one. There are two methods for this: analog and digital. We focus on digital data sampling.

Image processing Digital data sampling incorporates signal processing based on quantifiable observances ミ measurements. A constant signal is broken into a set of collected data. Images can be sampled this way by measuring color on a 256-color gray spectrum and approximating these values at pixel locations (coordinate pairs in R 2 ). Like all sampling, however, image sampling can be disturbed by noise.

Removing noise Currently, low levels of noise ( σ =10 or 15) can be removed well enough from pictures that the resulting image is acceptable to the eye. Higher levels ( σ = 25 or 30) cannot be so easily removed. Here we deal with denoising images with σ = 30).

Original imageσ = 10 σ = 20σ = 30 Noised Images (1)

Original imageσ = 10 σ = 20σ = 30 Noised Images (2)

Original imageσ = 10 σ = 20σ = 30 Noised Images (3)

PSNR The peak signal-to-noise ratio (PSNR) is the typical method of measuring image quality improvement. MAX I is the maximum possible pixel value of the image (here 255). MSE is mean squared error.

Method There are many methods available for image denoising. One that is popular is wavelet denoising; another is based on PDE methods. We test a new spline-based method to improve upon already wavelet denoised images. This approach has been shown to locally improve PSNR (peak signal to nosie ratio) levels, and visual examination shows noticeably reduced noise levels.

Edge finding The spline method samples a collection of data points (neighboring pixels) within the image. Of course, however, considering large regions involves sampling large amounts of data. As such, we must divide the domain. We do this locally by separating independent regions and then determining their approximate boundaries.

Edge finding (cont.) This involves selecting a starting location and searching radially away from it. Boundary points are collected and the data set is thinned for faster, smoother computation.

Boundary point refinement Our method of refinement relies on testing the angle formed by three successive points on the boundary. This angle’s minimum absolute distance from the horizontal is thresholded. If it is too small, the point is excluded.

Triangulation With these points, we can create a triangulation of the domain. Across the triangulation we fit penalized least squares splines to approximate color values.

Original image LenaTriangulations Wavelet denoised image (PSNR = 2223) Further spline denoised image (PSNR = 2224)

Original imageTriangulations Wavelet denoised image (PSNR = 2223) Further spline denoised image (PSNR = 2224)

Original imageTriangulations Wavelet denoised image (PSNR = 2223) Further spline denoised image (PSNR = 2224)

Results While visual evidence confirms our spline method is successful in locally eliminating noise, it does not yet exhibit significant global PSNR improvement. More intricate image details and small objects demonstrably lose sharpness and contrast shading as a result of spline denoising, which impairs global PSNR improvement. However, with reasonably good triangulations over larger areas, the eye verifies the spline method’s effectiveness in deleting noise.

Original image (Lena)Noised image (σ = 30) Wavelet denoised image (PSNR = ) Further spline denoised image (PSNR = )

Original image (Peppers)Noised image (σ = 30) Wavelet denoised image (PSNR = ) Further spline denoised image (PSNR = )

Original image (Saturn)Noised image (σ = 30) Wavelet denoised image (PSNR = ) Further spline denoised image (PSNR = )

PSNR Readings - Peppers Tdomain Regions [1] [2] [3]

Peppers (Cont.) [4] [5] [6] Global:

PSNR Readings - Lena Tdomain Regions Global

PSNR Readings - Saturn Tdomain Regions (Rings) (Planet)

Original imageNoised image Wavelet denoised imageFurther spline denoised image

Original imageNoised image Wavelet denoised imageFurther spline denoised image

Local spline denoising (T=4)Local spline denoising (T=6) Local spline denoising (T=5)Local spline denoising (T=7)

Local spline denoising (T=4)Triangulation (T=4) Local spline denoising (T=7)Triangulation (T=7)

Improvements The variety of results from spline denoising with fixed degree and smoothness indicate the signifiance of the triangulation of the splines’ domain. Thus it seems the most improvement could be exhibited by creating better triangulations. These give the splines better initial guesses at what they should be predicting. Also, we would like to automate this step as much as possible.

Improvements (cont.) Our next focus will be refining and trimming triangulations so that triangles maximize area – giving each one the appropriate number of sample points to analyze the distributions. We will write and employ MATLAB code to change the triangulations in this way, then denoise regions and images once again using splines.

Improvements (cont.) Once we’ve done this, we can focus on adjusting the degree and smoothness of the spline functions in order to observe their full power in analyzing complex data.

Bibliography Awanou, G. and M.J. Lai, The Multivariate spline method for scattered data fitting and numerical solutions of partial differential equations. Wavelets and Splines (2005), Lai, M.J. and K. Nam, Image denoising using box spline tight-frames. Lian, Q. et al., Wavelets on invariant sets for image denoising. (2007)