DEJAVNIKI KAKOVOSTI V TISKU

Slides:



Advertisements
Similar presentations
Ma.
Advertisements

Filtration based on Color distance
Contrast-Aware Halftoning Hua Li and David Mould April 22,
H. R. Sheikh, A. C. Bovik, “Image Information and Visual Quality,” IEEE Trans. Image Process., vol. 15, no. 2, pp , Feb Lab for Image and.
Achromatic and Colored Light CS 288 9/17/1998 Vic.
Guillaume Lavoué Mohamed Chaker Larabi Libor Vasa Université de Lyon
“POLITEHNICA” UNIVERSITY OF TIMIOARA FACULTY OF ELECTRONICS AND TELECOMMUNICATIONS DEPARTMENT OF COMMUNICATIONS DIPLOMA THESIS VIDEO QUALITY ASSESSMENT.
Graphics&Design Correction and Adjustment - Creative: colour & tone, dust & marks, sharp Tech: Gamma, Histograms, & Convolution.
Colour Theory and Application B.Sc. (Hons) Multimedia ComputingMedia Technologies.
Perceptual Evaluation of Colour Gamut Mapping Algorithms Fabienne Dugay The Norwegian Color Research Laboratory Faculty of Computer Science and Media Technology.
Introduction to Image Quality Assessment
Losslessy Compression of Multimedia Data Hao Jiang Computer Science Department Sept. 25, 2007.
Mean Squared Error : Love It or Leave It ?. Why do we love the MSE ? It is simple. It has a clear physical meaning. The MSE is an excellent metric in.
1 Blind Image Quality Assessment Based on Machine Learning 陈 欣
Perceived video quality measurement Muhammad Saqib Ilyas CS 584 Spring 2005.
Colour Image Compression by Grey to Colour Conversion Mark S. Drew 1, Graham D. Finlayson 2, and Abhilash Jindal 3 1 Simon Fraser University 2 University.
Human Visual System 4c8 Handout 2. Image and Video Compression We have already seen the need for compression We can use what we have learned in information.
An automated image prescreening tool for a printer qualification process by † Du-Yong Ng and ‡ Jan P. Allebach † Lexmark International Inc. ‡ School of.
1 Color Processing Introduction Color models Color image processing.
1 © 2010 Cengage Learning Engineering. All Rights Reserved. 1 Introduction to Digital Image Processing with MATLAB ® Asia Edition McAndrew ‧ Wang ‧ Tseng.
بسمه تعالی IQA Image Quality Assessment. Introduction Goal : develop quantitative measures that can automatically predict perceived image quality. 1-can.
What is Image Quality Assessment?
Computer Vision – Fundamentals of Human Vision Hanyang University Jong-Il Park.
MDDSP Literature Survey Presentation Eric Heinen
03/05/03© 2003 University of Wisconsin Last Time Tone Reproduction If you don’t use perceptual info, some people call it contrast reduction.
23 November Md. Tanvir Al Amin (Presenter) Anupam Bhattacharjee Department of Computer Science and Engineering,
Introduction to Computer Graphics
Surround-Adaptive Local Contrast Enhancement for Preserved Detail Perception in HDR Images Geun-Young Lee 1, Sung-Hak Lee 1, Hyuk-Ju Kwon 1, Tae-Wuk Bae.
Chapter 8 Lossy Compression Algorithms. Fundamentals of Multimedia, Chapter Introduction Lossless compression algorithms do not deliver compression.
COMPARATIVE STUDY OF HEVC and H.264 INTRA FRAME CODING AND JPEG2000 BY Under the Guidance of Harshdeep Brahmasury Jain Dr. K. R. RAO ID MS Electrical.
1 Marco Carli VPQM /01/2007 ON BETWEEN-COEFFICIENT CONTRAST MASKING OF DCT BASIS FUNCTIONS Nikolay Ponomarenko (*), Flavia Silvestri(**), Karen.
Performance Measurement of Image Processing Algorithms By Dr. Rajeev Srivastava ITBHU, Varanasi.
Image credit: Wikipedia (Fovea) Human Eye Some interesting facts – Rod cells: requires only low light b/w vision blur, all over retina EXCEPT fovea – Cone.
Masaki Hayashi 2015, Autumn Visualization with 3D CG Digital 2D Image Basic.
SREČNO NOVO LETO 2009 Lara Koren, 2.c.
Področje: 3.05 Reprodukcija človeka
Chapter 8 Lossy Compression Algorithms
Display Issues Ed Angel
Quality Evaluation and Comparison of SVC Encoders
PERFORMANCE ANALYSIS OF VISUALLY LOSSLESS IMAGE COMPRESSION
WAVELET VIDEO PROCESSING TECHNOLOGY
2.1 Direct Binary Search (DBS)
Lossy Compression of Stochastic Halftones with JBIG2
Digital 2D Image Basic Masaki Hayashi
SPACE OF OPPORTUNITIES
Nikolay Ponomarenkoa, Vladimir Lukina, Oleg I. Ieremeieva,
Introduction to Computer Graphics with WebGL
Super-resolution Image Reconstruction
ECE 638: Principles of Digital Color Imaging Systems
R V P 2 Predavanje 04 Animacija RVP2 Animacija.
Fotografiranje hrane Predmet: Tipografija in reprodukcija.
Pei Qi ECE at UW-Madison
Digital Multimedia Coding
Microsoftove rešitve za šolstvo
Glavne lastnosti multimetrov so:
MS Excel, Open Office Calc, Gnumeric …
Presenter by : Mourad RAHALI
Šalej Mirko Iskraemeco, d.d. Kranj
A Review in Quality Measures for Halftoned Images
Ela Reven, Katarina urbančič
Ugani število Napišimo program, ki si “izmisli” naključno število, potem pa nas v zanki sprašuje, katero je izmišljeno število. Če število uganemo, nas.
E-mobilnost in njena integracija v elektroenergetski sistem
TERMODINAMIKA / THERMODYNAMICS
Strukture podataka i algoritmi 5. VRIJEME IZVRŠAVANJA ALGORITMA
Digital television systems (DTS)
Image Perception and Color Space
Increasing Sensitivity of Ca2+ Spark Detection in Noisy Images by Application of a Matched-Filter Object Detection Algorithm  Cherrie H.T. Kong, Christian.
Lecture 2: Image filtering
Image quality measures
Review and Importance CS 111.
Presentation transcript:

DEJAVNIKI KAKOVOSTI V TISKU METODE OBJEKTIVNEGA VREDNOTENJA Tadeja Muck

+ + DOLOČANJE KAKOVOSTI PRINTING PROCESS + INK + PRINTING SUBSTRATE Print quality had been defined through many attributes: lightness blur printer type paper flatness sharpness details paper whiteness contrast effective resolution structure properties banding dinamic range structure changes gloss text quality paper roughness tone reproduction gloss uniformity line quality process colour gamut Many authors tried to define attributes that affect print quality in order to develop print quality metrics which correlates well with human perception. Experiments and many researches highlighted some of the attributes like: colour, lightness, gloss, sharpness, tone reproduction, structure properties, micro uniformity, line and text quality etc. Different authors suggested various properties to be investigated, but all of these are ceartainly important and influence print quality to a greater or lesser extent. As can be seen there are plenty of these attributes. Some of them are direcly related to colour like: tone reproduction, hue, chroma, saturation, colour shift, colour rendition, process colour gamut etc. Printer type, effective resolution and tone levels depends of printer used, while parameters like paper roughness, flatness, whiteness, structure properties and changes depends of a substrate. Gloss can be connected with gloss uniformity, while some attributes are related to human perception only– naturalness and perceived gray value. Evaluating so many attributes in practical application is not possible or even necessary. In order to reduce the number of parameters to be assessed, some attributes need to be grouped and observed as more universal properties. ordered noise chroma saturation micro uniformity colour adjacency patchlines colour shift hue naturalness noise/graininess colour rendition perceived gray value artifacts mottle

ATRIBUTI KAKOVOSTI … print quality metrics which correlates well with human perception … Grouping quality attributes enables efficient quality assesment: Color aspects related to: hue, saturation … Lightness is important, separate from the color… Contrast differences in lightness and chromaticity, within the image… Sharpness clarity of details and edges definition … Artifacts noise, contouring, and banding … Possibe reduction to a more manageable set of attributes is shown on this image. Here, there are five basic attributes which is shown to affect print quality the most.   • Color contains aspects related to color, such as hue, saturation, and color rendition, except lightness. • Lightness is considered so perceptually important that it is beneficial to separate it from the color. • Contrast can be described as the perceived magnitude of visually meaningful differences, global and local, in lightness and chromaticity, within the image. • Sharpness is related to the clarity of details and definition of edges. • And, since in color printing some artifacts can be perceived in the resulting image, these are also included into attributes of importance. These artifacts, like noise, contouring, and banding, contribute to degrading the quality of an image if detectable. Beside these, which are general parameters not related to substrates and inks used, physical QAs had also been defined. These attributes contains all physical parameters that affect quality, such as paper properties and gloss. Overall image quality can be influenced by one, two, three or more quality attributes and their interactions. + physical quality attributes; paper properties, gloss …

KONTROLA KAKOVOSTI Image processing Densitometry Spectrophotography / Colorimetry Densitometry

UVOD kako objektivno določiti z očesom zaznano razliko; original : reprodukcija? algoritmi rastriranja > vnos napak kvantizacije (quantization error/noise) > zmanjšanje bitne globine slike npr. iz 8 (256 odtenkov sivin) na 1 (binarna slika) kako objektivno (numerično) določiti z očesom zaznano razliko med originalno in reproducirano (rastrirano) sliko? algoritmi rastriranja vnesejo neko napako – t.i. napako (šum) kvantizacije (quantization error/noise) > zmanjšanje bitne globine slike npr. iz 8 (256 odtenkov sivin) na 1 (binarna slika)

METODE VREDNOTENJA KAKOVOSTI subjektivne > zamudne objektivne > numerično izraziti zaznano razliko med večtonskim originalom in rastrirano sliko določiti objektivno optimalen algoritem rastriranja > problem > kakovost rastriranja se lahko spreminja z aplikacijo Splošno dva razreda metod objektivnega vrednotenja kakovosti: matematično / statistične metode (enostavne, hitre, neodvisne od pogojev opazovanja, opazovalca) vključen HVS (človeški vizualni sistem za napoved percepcijske vizualne kakovosti, low-pass filter, omejitve vidnega sistema) subjektivne (časovno zamudne) objektivne - numerično izraziti zaznano razliko med večtonskim originalom in rastrirano sliko > avtomatsko rangirnje določiti objektivno optimalen algoritem rastriranja problem > kakovost rastriranja se lahko spreminja z aplikacijo Splošno dva razreda metod objektivnega vrednotenja kakovosti: matematično / statistične metode (enostavne, hitre, neodvisne od pogojev opazovanja, opazovalca) vključen HVS (človeški vizualni sistem za napoved percepcijske vizualne kakovosti, low-pass filter, omejitve vidnega sistema) problem > kakovost rastriranja se lahko spreminja z aplikacijo ( slika dobro izgleda na zaslonu, če se isti algoritem rastriranja uporabi v tisku – porazno!) (človeški vizualni sistem za napoved percepcijske vizualne kakovosti, upoštevajo, da naše oko deluje kot low-pass filter in hkrati lastnosti ter omejitve vidnega sistema. Zato je potrebno narediti računski model HVS

STATISTIČNE METODE VREDNOTENJA MATEMATIČNO / STATISTIČNE METODE Matematično osnovana metrika MSE – Mean Squered error RMS – Root mean Square SNR - Signal-to-Noise Ratio PSNR - Peak Signal-to-Noise Ratio Za vse velja: per-pixel based method hitre, enostavne ne upoštevajo strukture, kontrasta le vidik moči signala napake in ne kako le-ta vpliva na sliko Za vse velja: per-pixel based method hitre, enostavne ne upoštevajo razmerja med piksli ne upoštevajo strukture, kontrasta le vidik moči signala napake in ne kako le-ta vpliva na sliko

STATISTIČNE METODE VREDNOTENJA STATISTIČNE METODE - MSE, RMS, SNR, PSNR x(i,j) > sivinska slika y(i,j) > rastrirana slika M·N > velikost slike D > maks. peak-to-peak vrednost signala (255 za 8-bit) x(i,j) y(i,j) Višja vrednost SNR, PSNR > boljša je reprodukcija, bolj podobna originalu Emse, rms ravno obratno > višja vrednost slabša je reprodukcija …. A ni nujno, da so to pravi kazatelji …

STATISTIČNE METODE VREDNOTENJA STATISTIČNE METODE - MSE, RMS, SNR, PSNR Leva slika – dodan gaussian white noise Desna – FM rastrirana slika > ERROR DIFFUSION)

ČLOVEKOV VIDNI SISTEM - HSV HVS METODE HVS = Human Visual System CSF = Contrast Sensitivity Function

ČLOVEKOV VIDNI SISTEM - HSV Človeško oko kot nizkoprepustni filter značilnosti človeškega očesa > upadanje njegove občutljivosti pri višjih prostorskih frekvencah visokofrekvenčni vzorec (šum) bo torej težje zaznan oz. manj moteč kot nizkofrekvenčni

ČLOVEKOV VIDNI SISTEM - HSV Modeli HVS občutljivost človeškega vidnega sistema pri različnih frekvencah podaja funkcija kontrastne občutljivosti – Contrast sensitivity function (CSF) CSF > vlogo pri določanju ločljivosti slike, izboljšanju kakovosti in izbiri najustreznejšega algoritma rastriranja CSF CSF štirih modelov človeškega vidnega sistema občutljivost človeškega vidnega sistema pri različnih frekvencah podaja funkcija kontrastne občutljivosti – Contrast sensitivity function (CSF) CSF ima pomembno vlogo pri določanju ločljivosti slike, izboljšanju njene kakovosti in izbiri najustreznejšega algoritma rastriranja f – prostorska frekvenca cikl/stopinjo (funkcija ima vrednost 1 pri f=8 cikl/stopinjo in nima smisla na frekvence nad 60 cikl./stopinjo. Frekvenca (cikli/st)

ČLOVEKOV VIDNI SISTEM - HSV Funkcija kontrastne občutljivosti (CSF) Campbell 1969 Mannos 1974 Nasanen 1984 Daly 1987

ČLOVEKOV VIDNI SISTEM - HSV CSF

ČLOVEKOV VIDNI SISTEM - HSV Wang 03 – članek Prototipni sistem za oceno kakovosti slik osnovan na občutljivosti za napake. CSF je lahko implementirana kot samostojna stopnja ali pa znotraj normalizacije napake.

HVS METODE UNIVERSAL IMAGE QUALITY INDEX – Zhou Wang Modeliranje popačenja slik kot kombinacija 3 faktorjev: izguba korelacije (loss of correlation) popačenje svetlosti (luminance distortion) popačenje kontrasta (contrast distortion) Universal – neodvisen od: slike (vsebine) pogojev opazovanja od opazovalca http://www.ijser.org/researchpaper%5CComparison-of-Image-Quality-Assessment-PSNR-HVS-SSIM-UIQI.pdf

HVS METODE UNIVERSAL IMAGE QUALITY INDEX – Zhou Wang koeficient korelacije med x, y (meri stopnjo lin. korelacije, max. 1) kako blizu je srednja vrednost svetlosti med x, y, max 1 kako podobna sta kontrasta slik; max 1 Članek; A Universal Image Quality Index x= (xi; i =1,2 …N), y=(yi; i=1,2 …N) – signal originalne in testne slike. Trije faktorji – komponente korelacijski koeficient med x in y, ki meri stopnjo linearne korelacije med x in y in njihovo dinamično območje (-1, 1). Najboljša vrednost je 1, ko je yi=axi+b, kjer sta a in b konstanti in a > 0. Tudi če je zveza med x in y linearna, lahko obstajajo relativna popačenja med njima, kar je ovrednoteno v 2. in 3. koeficientu komponenta z območjem med (0 in 1) meri kako blizu je srednja vrednost luminance med x in y. Komponenta meri kako podobna sta si kontrasta slik, najboljša vrednost je 1. Zgornji enačbi sta enaki, le zapisani drugače. Na desni je delitev na tri faktorje in sicer (od leve proti desni): izguba korelacije (loss of correlation) popačenje svetlosti (luminance distortion) popačenje kontrasta (contrast distortion)

HVS METODE UNIVERSAL IMAGE QUALITY INDEX Original Salt-pepper noise Gaussian noise Speckle noise MSE UIQI b) 255 0,6494 c) 255 0,3891 d) 255 0,4408

HVS METODE UNIVERSAL IMAGE QUALITY INDEX MSE UIQI a) 255 0,9894 b) 255 0,9372 c) 255 0,3461 d) 215 0,2876 Mean shifted image Contrast streched image Blurred image JPEG compresed image

HVS METODE SSIM – structural similarity index loči strukturo, svetlost in kontrast > simuliranje HSV rezultat primerjava vhodna : izhodna slika > vrednost (0 – 1) človekovo oko je občutljivo že na majhne spremembe v svetlosti, kontrastu in strukturi, kar danes vključuje že veliko algoritmov SSIM opisuje kakovost slike s primerjavo lokalnih korelacij svetlosti, kontrasta in strukture med originalno in testno sliko metoda, ki v osnovi loči strukturo, svetlost in kontrast, na ta način simulira HSV rezultat primerjava vhodne – izhodne slike izražen z vrednostjo med 0 in 1 človekovo oko je občutljivo na majhne spremembe v svetlosti, kontrastu in strukturi, kar danes vključuje že veliko algoritmov osnovna funkcija človekovega očesa je ločitev strukturnih informacij iz vidnega polja oko je zelo občutljivo na strukturne distorzije SSIM opisuje kakovost slike s primerjavo lokalnih korelacij svetlosti, kontrasta in strukture med originalno in testno sliko.

HVS METODE SSIM – structural similarity index Diagram of the structural similarity (SSIM) measurement system Diagram merilnega sistema strukturne podobnosti.

HVS METODE http://www-ee.uta.edu/dip/Courses/EE5356/ssimzhouwang.pdf

HVS METODE SSIM – structural similarity index Komponente relativno neodvisne; sprememba svetlosti in/ali kontrastu ne vplivala na strukturo slike Slike enak MSE, različen SSIM Iz seminarja marjane vlaović in Wanga

HVS METODE http://www-ee.uta.edu/dip/Courses/EE5356/ssimzhouwang.pdf

HVS METODE

LITERATURA ImageJ vtičniki (P)SNR, RMSE, MAE: http://bigwww.epfl.ch/sage/soft/snr/ SSIM indeks: http://rsbweb.nih.gov/ij/plugins/ssim-index.html MSSIM in MSSIM* indeks: http://rsbweb.nih.gov/ij/plugins/mssim-index.html http://waset.org/publications/9998522/a-new-categorization-of-image-quality-metrics-based-on-a-model-of-human-quality-perception http://www.imaging.org/ist/publications/reporter/articles/REP26_2_EI2011_PEDERSEN_7867_1.pdf …