HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.1 Outline Motivations Analytical Model of Skew Effect and its Compensation in Banding and MTF.

Slides:



Advertisements
Similar presentations
In InDesign, you can create a new file by pressing Command/Control-N.
Advertisements

CS Spring 2009 CS 414 – Multimedia Systems Design Lecture 4 – Digital Image Representation Klara Nahrstedt Spring 2009.
November 12, 2013Computer Vision Lecture 12: Texture 1Signature Another popular method of representing shape is called the signature. In order to compute.
Chap 4 Image Enhancement in the Frequency Domain.
1 Texture Texture is a description of the spatial arrangement of color or intensities in an image or a selected region of an image. Structural approach:
嵌入式視覺 Feature Extraction
Computer Vision Lecture 16: Texture
EI San Jose, CA Slide No. 1 Measurement of Ringing Artifacts in JPEG Images* Xiaojun Feng Jan P. Allebach Purdue University - West Lafayette, IN.
6/9/2015Digital Image Processing1. 2 Example Histogram.
Resolving the Problem Resolution: Concepts & Definitions.
Digital Image Processing Chapter 4: Image Enhancement in the Frequency Domain.
CS 128/ES Lecture 5a1 Raster Formats (II). CS 128/ES Lecture 5a2 Spatial modeling in raster format  Basic entity is the cell  Region represented.
Introduction to Image Quality Assessment
ISE 261 PROBABILISTIC SYSTEMS. Chapter One Descriptive Statistics.
Face Processing System Presented by: Harvest Jang Group meeting Fall 2002.
DREAM PLAN IDEA IMPLEMENTATION Introduction to Image Processing Dr. Kourosh Kiani
Despeckle Filtering in Medical Ultrasound Imaging
HP - PURDUE CONFIDENTIAL Slide No. Fundamentals 1.
Bitmapped Images. Bitmap Images Today’s Objectives Identify characteristics of bitmap images Resolution, bit depth, color mode, pixels Determine the most.
Dreamweaver Domain 3 KellerAdobe CS5 ACA Certification Prep Photoshop Domain 5: Publishing Digital Images Using Adobe Photoshop CS5 Adobe Creative Suite.
The Digital Image.
Introduction to electrical and computer engineering Jan P. Allebach School of Electrical and Computer Engineering
Chapter 2. Image Analysis. Image Analysis Domains Frequency Domain Spatial Domain.
An automated image prescreening tool for a printer qualification process by † Du-Yong Ng and ‡ Jan P. Allebach † Lexmark International Inc. ‡ School of.
Computer vision.
Lecture 3. Fundamentals of Computer Graphics. Computer Graphics, a very broad term Fields Related to Computer Graphics Bitmap/Vector graphics, 2D/3D graphics,
Screen Ruling, Print Resolution AM, FM and Hybrid Halftoning Sasan Gooran Linköping University LiU-Norrköping.
Digital Watermarking With Phase Dispersion Algorithm Team 1 Final Presentation SIMG 786 Advanced Digital Image Processing Mahdi Nezamabadi, Chengmeng Liu,
IDL GUI for Digital Halftoning Final Project for SIMG-726 Computing For Imaging Science Changmeng Liu
Bitmap Vs. Vector Graphics. To create effective artwork, you need to understand some basic concepts about vector graphics versus bitmap images, resolution,
An efficient method of license plate location Pattern Recognition Letters 26 (2005) Journal of Electronic Imaging 11(4), (October 2002)
1 Chapter 5 Image Transforms. 2 Image Processing for Pattern Recognition Feature Extraction Acquisition Preprocessing Classification Post Processing Scaling.
CS 6825: Binary Image Processing – binary blob metrics
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,
Color and Resolution Introduction to Digital Imaging.
Lecture 03 Area Based Image Processing Lecture 03 Area Based Image Processing Mata kuliah: T Computer Vision Tahun: 2010.
Digital Camera and Computer Vision Laboratory Department of Computer Science and Information Engineering National Taiwan University, Taipei, Taiwan, R.O.C.
CS654: Digital Image Analysis Lecture 25: Hough Transform Slide credits: Guillermo Sapiro, Mubarak Shah, Derek Hoiem.
Digital Image Processing (Digitaalinen kuvankäsittely) Exercise 2
Discrete Fourier Transform in 2D – Chapter 14. Discrete Fourier Transform – 1D Forward Inverse M is the length (number of discrete samples)
Numerical Evaluation of Banding Patterns in the Context of Human Visual System Noise Sensitivity. Submitted as a final study towards the degree of Master.
Gholamreza Anbarjafari, PhD Video Lecturers on Digital Image Processing Digital Image Processing Spatial Aliasing and Image Enhancement.
Dr. Scott Umbaugh, SIUE Discrete Transforms.
Autonomous Robots Vision © Manfred Huber 2014.
Vibrationdata 1 Unit 6a The Fourier Transform. Vibrationdata 2 Courtesy of Professor Alan M. Nathan, University of Illinois at Urbana-Champaign.
CS COMPUTER GRAPHICS LABORATORY. LIST OF EXPERIMENTS 1.Implementation of Bresenhams Algorithm – Line, Circle, Ellipse. 2.Implementation of Line,
Digital Image Processing
Digital Image Processing Image Enhancement in Spatial Domain
Image Quality Measures Omar Javed, Sohaib Khan Dr. Mubarak Shah.
Detecting Image Features: Corner. Corners Given an image, denote the image gradient. C is symmetric with two positive eigenvalues. The eigenvalues give.
Whole Slide Image Stitching for Osteosarcoma detection Ovidiu Daescu Colaborators: Bogdan Armaselu and Harish Babu Arunachalam University of Texas at Dallas.
# x pixels Geometry # Detector elements Detector Element Sizes Array Size Detector Element Sizes # Detector elements Pictorial diagram showing detector.
1 of 32 Computer Graphics Color. 2 of 32 Basics Of Color elements of color:
图像处理技术讲座(3) Digital Image Processing (3) Basic Image Operations
ISE 261 PROBABILISTIC SYSTEMS
Texture.
Fitting: Voting and the Hough Transform
Mean Shift Segmentation
Scott Tan Boonping Lau Chun Hui Weng
Tone Dependent Color Error Diffusion
1.2 Design of Periodic, Clustered-Dot Screens
Computer Vision Lecture 16: Texture II
Volume 6, Issue 5, Pages e5 (May 2018)
Efficient Receptive Field Tiling in Primate V1
Volume 19, Issue 2, Pages (August 1997)
Computer and Robot Vision I
Stability of Cortical Responses and the Statistics of Natural Scenes
7th Annual STEMtech conference
Efficient Receptive Field Tiling in Primate V1
Presentation transcript:

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.1 Outline Motivations Analytical Model of Skew Effect and its Compensation in Banding and MTF Characterization Moiré Artifact Prediction and Reduction in a Variable Data Printing Environment Conclusions References

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.2 Moiré Artifacts in Printing Moiré due to halftoning process Test pattern used to characterize halftoning processing of press Example image to be printed showing moiré artifacts

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.3 Quality of Embedded Images Example: Moiré Artifact Business Week, April 30, 2007 p.56

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.4 Document Composition Affects Artifact Perceptibility Artifact assessment depend on document composition:  Image scaling and rotation  Image cropping  Image position relative to other objects  Background color  Object overlay on image

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.5 Causes and Difficulties to Detect Moiré Artifacts in VDP Halftone screen pattern interacts with digital image  Clustered dot profile Limited spatial resolution of the digital press  Typical digital press : 180 line-per-inch In digital publishing environment with variable data printing  Inspecting each printed page is not cost efficient Moiré artifacts are image content dependent Moiré artifacts vary with the printing device

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.6 Phases and Components of Automatic Workflow [3]

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.7 Spectrum of Halftoned Digital Image in Terms of Spectrum of Original Continuous-tone Image Spectrum of the halftoned digital image can be expressed in terms of the original image and the halftone screen  H(u,v) -- spectrum of halftone image  f[l,k] -- original image  p[m,n;a] -- halftone dot profile  M – size of the halftone cell

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.8 Illustration of Halftone Spectrum for a Sine Wave Image Continuous-tone input image Halftone image Screening Compare Threshold matrix Spectrum of the continuous-tone input imageSpectrum of the halftone image Frequency doubling effect Frequency of the original sinusoidal wave

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.9 Nonlinear Transformation Due to Halftone |P[0,0;a]||P[0,1;a]| |P[0,2;a]| |P[1;a]| f[l] a l l P[1;f[l]] 1 T T 0 A B C A’ B’ C’

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.10 Frequency Doubling Effect Due to Nonlinear Transformation The frequency doubling effect is due to the non-linear transform caused by the screening process Clustered halftone dot profile that is used in laser printing is likely to cause this frequency doubling effect

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.11 Moiré Artifact as Result of Frequency Doubling Effect Continuous-tone input image Halftone image Screening Compare Threshold matrix Spectrum of the continuous-tone input imageSpectrum of the halftone image Moiré artifacts as low frequency component Frequency of the original sinusoidal wave

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.12 Moiré Prediction Image Database Press Profile Detection Algorithm Human Visual System Model Moiré Map Image Analysis Test Pattern Digital Press Real-time analysis of images in documentOffline press characterization process

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.13 Digital Press Characterization Use Bullseye test pattern  Sweep of signal at all angles  Spatial frequency at each location is proportional to its distance to the center Bullseye test pattern is printed using target digital press Moiré inducing frequency (MIF) generates low frequency moiré that forms secondary bullseye pattern on the print After scanning the printout, we detect the secondary bullseye pattern to locate MIF Halftone bullseye test pattern with moiré artifacts

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.14 Moiré Inducing Frequency (MIF) Detection on Test Page This test pattern shows multiple moiré artifacts patterns Each moiré artifact exhibits a pattern of concentric circles The xy coordinates of the center of each pattern of concentric circles correspond to a frequency that may cause moiré artifacts in the printed image moiré artifacts

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.15 Bullseye pattern halftoned with 150 cycles/inch, 0 degree screen; printed at 600 dpi and scanned at 600 dpi. The red dots indicate detected MIF’s Symmetry of the Secondary Bullseye Artifacts The secondary bullseye artifacts are symmetric to the center of the test page Each secondary bullseye artifact forms concentric circles Some pairs of secondary bullseye artifacts that are symmetrical to the center show different gray levels

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.16 1-D illustration Image: 5 cycles per inch Screen: 10 cycles per inch Average: Average: Same frequency

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.17 Anisotropy Measurements on Scanned Bullseye Pattern [4] Each image pixel’s anisotropy measurement is calculated based on a disk area Image pixels within the disk is divided into annuli The width of each annulus is delta, ∆ Image pixels are sorted into annulus (bins) based on their distance to the center of the region Mean and variance are calculated for each bin Calculate Anisotropy for each bin

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.18 Modified Anisotropy Measurement Secondary Bullseye Artifacts Modified anisotropy measurement takes account on the entire region’s energy to give better distinction between concentric circles (secondary bullseye) and random noise region

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.19 Bullseye pattern halftoned with 150 lines/inch, 0 degree screen; printed at 600 dpi and scanned at 600 dpi. The red dots indicate detected MIF’s Printer MIF Detection Result Maximal frequency: 90 cycles/inchMaximal frequency: 55 cycles/inch

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.20 MIF Detection on Test Page Radial Frequency (cycles per inch) Angle (degrees) 37 ±90 50 ±90 75 ±90 57 ±64 67 ±64 75 ±64 50 ±45 72 ±45 75 ±45 57 ±26 67 ±26 75 ±

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.21 MIF detection in the continuous- tone input image Based on press profile, measure the energy of MIF in power spectrum of the digital image Find peaks in the spectrum of the continuous-tone image that corresponding to MIF frequency In frequency domain, calculate a confidence measure in the neighborhood of the peaks Calculate the size of each detected region to eliminate false alarms due to strong edge components

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.22 MIF Detection on Digital Images Sampling frequency of the digital image on print-out: Image Metadata in PPML or XML  Dimension: image width/height size  Position: Determined by the attribute “Position” in MARK and OBJECT elements  Transform Matrix: provides various image properties such as scale, skew, and translation  Clipping size: determined by the attribute “Rectangle” in CLIP_RECT element

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.23 Indices Representing MIF in Frequency Domain Check for MIF on the 2D-DSFT of the digital image:

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.24 Confidence Measurement in Frequency Domain In frequency domain, calculate a confidence measure in the neighborhood of the peaks Power spectrum

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.25 Confidence Measure Strong peak in power spectrum at the MIF location means perceptible moiré is likely to occur in printing Confidence measure helps to reduce misclassification Power Spectrum

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.26 Results: Sinusoidal Grating Digitally generated sinusoidal grating Starting from 10 cycles/inch with 20 cycles/inch increment per row Starting from 0 degree with 10 degrees increment per column Detection is done for 90 cycles/inch with 10 degrees

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.27 Misclassification Due to Strong Edges

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.28 Measure Length and Width of Each Detected Region Project each region to the horizontal and vertical axis of the image plan Count the number of pixels on each horizontal and vertical position Regions with maximal length or width less than 2N (N: the 2D DSFT window size) will be removed from mask. Projection to obtain width region identified in moiré mask

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.29 Misclassification Regions Removed

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.30 Adaptive Scaling to Reduce Moiré For each image identified with moiré we scaled the image to reduce moiré artifacts in print-out Each region on the moiré mask is analyzed to obtain a scale factor Global scale factor is the maximal of all the regional scale factors Entire image is scaled by the global factor

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.31 Results: Shirt Printed using HP LaserJet 5500 with 600 dpi and 150 lpi halftone Visible moiré artifacts on the shirt region Successful detection of using the printer profile

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.32 Results: Hotel Original digital image Moiré mask Scan of the original image print-out Scan of the scaled image print-out

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.33 Results: Kodak Window Original digital image Moiré mask Scan of the original image print-out Scan of the scaled image print-out

HP-PURDUE-CONFIDENTIAL Final Exam May 16th 2008 Slide No.34 Summary Analyze the relationship between the spectrum of halftone image and that of the original image Use bullseye pattern to characterize printer Identified moiré inducing frequency Predict moiré artifacts based on the image content, image pixel size, and actual printed size Adaptive image scaling to resize the image so that the new image will not induce moiré artifacts