halftoning Journal of Electronic Imaging, vol. 11, no. 4, Oct Je-Ho Lee and Jan P. Allebach

Similar documents
System in Weibull Distribution

Several generation methods of multinomial distributed random number Tian Lei 1, a,linxihe 1,b,Zhigang Zhang 1,c

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

BAYESIAN CURVE FITTING USING PIECEWISE POLYNOMIALS. Dariusz Biskup

Excess Error, Approximation Error, and Estimation Error

What is LP? LP is an optimization technique that allocates limited resources among competing activities in the best possible manner.

1 Definition of Rademacher Complexity

LECTURE :FACTOR ANALYSIS

XII.3 The EM (Expectation-Maximization) Algorithm

COS 511: Theoretical Machine Learning

Designing Fuzzy Time Series Model Using Generalized Wang s Method and Its application to Forecasting Interest Rate of Bank Indonesia Certificate

Perceptual Organization (IV)

A New Scrambling Evaluation Scheme based on Spatial Distribution Entropy and Centroid Difference of Bit-plane

Xiangwen Li. March 8th and March 13th, 2001

Applied Mathematics Letters

Least Squares Fitting of Data

Three Algorithms for Flexible Flow-shop Scheduling

4 Column generation (CG) 4.1 Basics of column generation. 4.2 Applying CG to the Cutting-Stock Problem. Basic Idea of column generation

Yong Joon Ryang. 1. Introduction Consider the multicommodity transportation problem with convex quadratic cost function. 1 2 (x x0 ) T Q(x x 0 )

Least Squares Fitting of Data

Lecture 12: Discrete Laplacian

Hidden Markov Models

Gadjah Mada University, Indonesia. Yogyakarta State University, Indonesia Karangmalang Yogyakarta 55281

Computational and Statistical Learning theory Assignment 4

An Accurate Measure for Multilayer Perceptron Tolerance to Weight Deviations

The Fourier Transform

Collaborative Filtering Recommendation Algorithm

PROBABILITY AND STATISTICS Vol. III - Analysis of Variance and Analysis of Covariance - V. Nollau ANALYSIS OF VARIANCE AND ANALYSIS OF COVARIANCE

Some modelling aspects for the Matlab implementation of MMA

Hiding data in images by simple LSB substitution

CONTRAST ENHANCEMENT FOR MIMIMUM MEAN BRIGHTNESS ERROR FROM HISTOGRAM PARTITIONING INTRODUCTION

Lecture 3: Probability Distributions

Lecture 3: Shannon s Theorem

Centroid Uncertainty Bounds for Interval Type-2 Fuzzy Sets: Forward and Inverse Problems

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X

Tutorial 2. COMP4134 Biometrics Authentication. February 9, Jun Xu, Teaching Asistant

Calculation of time complexity (3%)

Week 5: Neural Networks

y new = M x old Feature Selection: Linear Transformations Constraint Optimization (insertion)

BOUNDEDNESS OF THE RIESZ TRANSFORM WITH MATRIX A 2 WEIGHTS

Convexity preserving interpolation by splines of arbitrary degree

A be a probability space. A random vector

EE513 Audio Signals and Systems. Statistical Pattern Classification Kevin D. Donohue Electrical and Computer Engineering University of Kentucky

On the number of regions in an m-dimensional space cut by n hyperplanes

EPR Paradox and the Physical Meaning of an Experiment in Quantum Mechanics. Vesselin C. Noninski

Tests of Exclusion Restrictions on Regression Coefficients: Formulation and Interpretation

Probability Theory (revisited)

College of Computer & Information Science Fall 2009 Northeastern University 20 October 2009

Physics 5153 Classical Mechanics. D Alembert s Principle and The Lagrangian-1

APPENDIX A Some Linear Algebra

Maintenance Scheduling and Production Control of Multiple-Machine Manufacturing Systems

Finite Element Modelling of truss/cable structures

EGR 544 Communication Theory

A DIMENSIONALITY REDUCTION ALGORITHM OF HYPER SPECTRAL IMAGE BASED ON FRACT ANALYSIS

Solving Fuzzy Linear Programming Problem With Fuzzy Relational Equation Constraint

Lagrangian Field Theory

Pop-Click Noise Detection Using Inter-Frame Correlation for Improved Portable Auditory Sensing

Numerical Solution of Ordinary Differential Equations

χ x B E (c) Figure 2.1.1: (a) a material particle in a body, (b) a place in space, (c) a configuration of the body

Using T.O.M to Estimate Parameter of distributions that have not Single Exponential Family

Modified parallel multisplitting iterative methods for non-hermitian positive definite systems

arxiv:cs.cv/ Jun 2000

Block-Parallel Chaotic Algorithms for Image Reconstruction

THE PRACTICE OF AUTOMATIC SATELLITE IMAGE REGISTRATION

On the Multicriteria Integer Network Flow Problem

Optimal Control Scheme for Nonlinear Systems with Saturating Actuator Using ε-iterative Adaptive Dynamic Programming

Our focus will be on linear systems. A system is linear if it obeys the principle of superposition and homogenity, i.e.

Solutions for Homework #9

Canonical transformations

High resolution entropy stable scheme for shallow water equations

Electrostatic Potential from Transmembrane Currents

Study of the possibility of eliminating the Gibbs paradox within the framework of classical thermodynamics *

Simulation and Random Number Generation

Lecture Notes on Linear Regression

ITERATIVE ESTIMATION PROCEDURE FOR GEOSTATISTICAL REGRESSION AND GEOSTATISTICAL KRIGING

1 Review From Last Time

Multipoint Analysis for Sibling Pairs. Biostatistics 666 Lecture 18

Preference and Demand Examples

SELECTED SOLUTIONS, SECTION (Weak duality) Prove that the primal and dual values p and d defined by equations (4.3.2) and (4.3.3) satisfy p d.

Quasi Gradient Projection Algorithm for Sparse Reconstruction in Compressed Sensing

The Quadratic Trigonometric Bézier Curve with Single Shape Parameter

LINEAR REGRESSION ANALYSIS. MODULE VIII Lecture Indicator Variables

Chapter Newton s Method

Integral Transforms and Dual Integral Equations to Solve Heat Equation with Mixed Conditions

COMP th April, 2007 Clement Pang

Reliability estimation in Pareto-I distribution based on progressively type II censored sample with binomial removals

10.34 Fall 2015 Metropolis Monte Carlo Algorithm

A Fast Computer Aided Design Method for Filters

j) = 1 (note sigma notation) ii. Continuous random variable (e.g. Normal distribution) 1. density function: f ( x) 0 and f ( x) dx = 1

SINCE the 1990s, chaotic cryptography has attracted more

Markov Chain Monte-Carlo (MCMC)

CHARACTERISTICS OF COMPLEX SEPARATION SCHEMES AND AN ERROR OF SEPARATION PRODUCTS OUTPUT DETERMINATION

1. Statement of the problem

Difference Equations

Appendix B. The Finite Difference Scheme

DETERMINATION OF TEMPERATURE DISTRIBUTION FOR ANNULAR FINS WITH TEMPERATURE DEPENDENT THERMAL CONDUCTIVITY BY HPM

FUZZY MODEL FOR FORECASTING INTEREST RATE OF BANK INDONESIA CERTIFICATE

[ ] T Journal of Software. Vol.13, No /2002/13(05)

Physics 3A: Linear Momentum. Physics 3A: Linear Momentum. Physics 3A: Linear Momentum. Physics 3A: Linear Momentum

RELIABILITY ASSESSMENT

Transcription:

olorant-based drect bnary search» halftonng Journal of Electronc Iagng, vol., no. 4, Oct. 22 Je-Ho Lee and Jan P. Allebach School of Electrcal Engneerng & oputer Scence Kyungpook Natonal Unversty

Abstract Presentng a new colorant-based color halftonng technque Proposng color desgn crtera for unfor color textures ontrollng the qualty of each colorant texture Settng the dots optally wthout alterng the total arrangeent orrespondng to constraned optzaton proble Solvng ths proble va the swap-only drect bnary search Eployng a lunance huan vsual syste odel Enforcng excluson of vsually non-hoogeneous pattern

. Introducton Three groups of dgtal color halftonng Accordng to the coputatonal coplexty Screenng ausng artfacts such as ore patterns Error dffuson Scalar error dffuson and vector error dffuson ausng artfacts by error accuulaton Iteratve processes Least squares and DBS Takng a DBS-based color halftonng n ths paper

2. olorant based DBS Algorth rtera for Unfor olor Texture If we consder the dots of each prary color (, M, or Y) ndvdually, we would lke these dots to be arranged as unfor as possble We also would lke the overall coposte texture that conssts of all these prary colorant dots to be as unfor as possble We wsh to nze dot-on-dot prntng as uch as possble

Dscussng to b-level three-colorant MY prnters n ths paper. But can be appled to b-level four-colorant MYK prnters hoosng jontly halftone only the and M planes, do the Y plane ndependently, and fnally cobne ther halftones To avod dot-on-dot prntng as uch as possble, are gven by f ', M', and B ' = M M' = B = M ()

(a) (b) Fg. Two halftone patterns llustratng (a) dot-on-dot and (b) dotoff-dot prntng for a 7% gray level prnted at dp

(a) M: ndependently processed (b) Y: ndependently processed (c) MY: ndependently processed Fg. 2 oparson between ndependent and jont processng of colorant halftones for colorant aount of =M=Y=2%: (a)-(c) halftones obtaned by separately halftonng each of the color planes (contnue)

(d) M: jontly processed (e) Y: jontly processed (f) MY: jontly processed oparson between ndependent and jont processng of colorant halftones for colorant aount of =M=Y=2%: (d)-(e) halftones obtaned by jontly halftonng the color planes

(a) (b) Fg. 3 oparson between (a) restrctng Y to go nto spaces between colorant dots of jontly desgned and M halftones, and (b) desgn of the Y halftones copletely ndependently of the and M halftones for =M=Y=2%

(a) yan (b) Magenta (c) yan Magenta Fg. 4 Vsualzaton of relatons gven n Eq. () to nze dot-on-dot prntng

(a) (c) (b) (d) Fg. 5 Steps of color halftonng algorth for the case where =8%, M=5%, and Y=%: (a) set overall dot arrangeent, (b) color the dots the randoly, (c) refne the halftones for and M, and (d) fll holes wth B

Extenson to General Iages f f f M f[ ], ' = fm, [ fm, ' ] = f[ ], M [ ' M ' ] = f f, and f Where = [ = and contnuous spatal coordnates to represent dscrete Where f '[ ] and f M ' are, respectvely, the aount of cyan and agenta coponents whch do not contrbute to blue = or and g f f f f otherwse Applyng the onochroe DBS halftong algorth to M t t, n] and (x) ( x, y) f : R otherwse f 2 M f M R. =,M,Yand f. (2) (3) (4) f M []

s M W, f gm = and f B, f gm = and f =, f gm = and f ' M, otherwse f ' f ' < Let s [ ]: R 2 M {W,, M,B} be the functon that represents the colorng of g M [] r s a rando varable whch s unforly dstrbuted n the nterval (,) Seekng an optal color pattern for s M [] by swappng and M pxels to reduce an error etrc teratvely Extng teraton after the algorth has converged f f M M M > r (5)

g g M, f sm = or B =, otherwse, f sm = M or B =, otherwse g [] and g M [] are obtaned fro s M [] (6)

Fg. 6 Block dagra of the flow the halftonng algorth proposed

Error Metrcs g g ' M ' =, =, f f s s M M = = M g [ ] and g M ' denote the halftone age for f ' and ' (7) f M ' g ( x) = g[ ] p( x X), = ', M ' The halftone age rendered by the prnter (8) p(x) s the spot functon of the prnter and X s the perodcty atrx whose coluns coprse the bass for the lattce of prnter addressable dots

e ( x) = h( x) [ g ( x) f ( x)], = ', M ' (9) Defnton of the perceved error age h(x) s the pont spread functon of the HVS (x) s contnuous-tone orgnal age e f ( x) = e p( x X), = ', M' Where e = g f, ', M ' = () ()

ε 2 = e ( x) dx, = ', M ' Measure of error s based on the total squared value of Eq. () (2) nze ε (3a) subject to g ' g M '[ ] = g M (3b) Where ε = αε ' βεm ' s dual error easure whch penalzes the use of both and M halftone textures that lead to a vsually unpleasng appearance.

Error Mnzaton c ( x) = p( y) p( x p p y) dy Defnton of autocorrelaton functon Dscrete for of Eq. (2) ε e [ = = n ] e c [ n], ', M ' p p (4) orrelaton functon c = c ( X) p p p p g = a δ[ ] aδ[ ], = ', M ' (5) onsderng a tral swap between pxels and Descrbng the change n g ' and g '

g = a δ[ ] aδ [ ], = ', M ' onsderng a tral swap between pxels and Descrbng the change n g ' and g ' (5) Where a and a, = ',M' are gven by a j =, f g[ j ] s changed fro to, f g [ ] s changed fro to j j =, (6) The par ( g '[ ], g' [ ]), ( gm '[ ], gm '[ ]) ( g' [ ], gm '[ ]), g [ ], g [ ]) ust all have dfferent bnary states, we have followng relatonshps aong a and a, = ',M' : ( ' M ' a ' = a M ' = a ' = a M ' (7)

Where s the cross correlaton between and Dual etrc ] [ ) 2( ]) [ ] [ ( 2 ] [ ] [ 2 [] ) 2( ' ' ' ' ' ' c c c a c c a c p p M pe p e M pe pe p p = β α β α β α β α ε ' ', ], [ 2 ] [ 2 ] [ 2 [] ] ) ( ) [( 2 2 M c a a c a c a c a a p p pe pe p p = = ε ' ', ], [ ] [ ] [ M n c n e c p p n pe = = ] [ c e p ] [ c p p e [] (8) (9) (2)

Experental Results oparng color halftones Proposed colorant based DBS algorth Plane-ndependent DBS The HP 97 x nkjet prnter drver hoosng two halftones age Rap and dnng table

(a) (b) (c) Fg. 7 Halftones for the Rap age generated by (a) plane-ndependent DBS, (b) the HP 97 x nkjet prnter drver, and (c) colorantbased DBS. Note that ages are prnted at 5 dp. Ths s uch lower resoluton than would norally be used wth the HP 97 x prnter. Therefore (b) s not representatve of the qualty of the real output of the HP 97 x prnter.

(a) (b) (c) Fg. 8 Halftones for the dnng table age generated by (a) planendependent DBS, (b) the HP 97 x nkjet prnter drver, and (c) colorantbased DBS. Note that ages are prnted at 5 dp.

Suary and oncluson Ths algorth can be effcently pleented by recursvely evaluatng the effect of tral changes olorant-based DBS s stll teratve and slower than non-teratve algorth Future work Extenson of proposed algorth to general case