Download as pdf or txt
Download as pdf or txt
You are on page 1of 54

Digital Image Processing

Syllabus & Grading System, and Introduction

Semester Gasal 2010/2011


Prof.Dr. Aniati Murni Arymurthy (Bldg. A, R 1202)
Faculty of Computer Science
University of Indonesia
Grading System, Text Book, Programming

Grading System:
• Homework and Programming Assignments (~30%)
• Mid Test / UTS (~35%) Final Exam / UAS (~35%)
References:
• Slides & Hand outs
• Gonzalez & Woods, 2002/2008, Digital Image Processing
Programming:
• C++ or Java or MATLAB

2
Teaching and Learning Methods

• Lectures (target of course content) – 28 sessions


• Independent Study (critical thinking) - 1 topic
• Presentation (communication ability – small class)
• Paper/report writing (knowledge composition)
• Team work (project management)
• Programming (ability to implement methods) – 2 assignments
• Mid Test and Final Test (comprehension test)

3
University / Faculty Regulations

„ Minimum attendance in class : 75%


„ No free rider in working group
„ No plagiarism in report / academic work and
writing
„ No cheating in examination
„ No UTS/UAS susulan (without appropriate
reason)

4
Three Areas of Study that Related to
Image and Picture Processing

5
Three Areas of Study that Related to
Image and Picture Processing

„ Computer Graphics
„ Image Processing
„ Pattern Recognition / Computer Vision /
Artificial Intelligence

6
Pavlidis, 1986
1950 Image Processing

Image
1970 Computer Vision
1970 Computer
Graphics
1960 Pattern Recognition
Artificial Intelligence

Description
Michigan State University, 1990

Description Generated Image


Computer Graphics (1)

Object Description Generated Image


(Murni, 1979) 9
Computer Graphics (2)

Wire Frame Drawing Realism Drawing


(Hearn and Baker, 1986) 10
Computer Graphics
„ A process, technique, and method to
generate a picture based on the
description of both its objects and
backgrounds;
„ A process, technique, and method to
create a realism effect on the objects
and backgrounds contained in a picture;
„ Drawing a picture and animating
objects using a computer.
11
Image Processing (1)

(JPL, 1972)

Input/Degraded Image Output/Enhanced Image 12


Image Processing and
Pattern Recognition (2)

CASI image of Bogor area Classified / Thematic Image


(Source: The Map Indonesia) (Source: Wiweka, 2006)
13
Digital Image Processing
„ Image quality enhancement: (a) radiometric aspect
(contrast enhancement, colour transformation, image
restoration); and (b) geometric aspect (rotation,
translation, scale, geometric transformation);
„ Feature/image extraction and selection to obtain
images that would be optimal for analysis purpose;
„ Data reduction and image compression (for the
purpose of efficiency in data storage, data
transmission, and data processing time);
„ Information extraction, object recognition and
description that is contained in an image.

14
Computer Vision (1)

Garage Bushes Grass House Sky Tree1 Tree2

Roof Side Roof Side1 Side2

(Ballard, 1992)
Input Images Descriptions 15
Computer Vision (2)

Bahasa isyarat lainnya: menggunakan bahasa tangan dan ada


juga yang disebut sebagai ‘body language’; mengangguk (jarak
antara garis alis dan mulut mengecil), menggeleng (jarak antara
garis mata kiri dan kanan mengecil).
16
Computer Vision
„ Pattern Recognition: Segmentation and
Classification, Speaker Recognition – siapakah yang
membacakan kalimat tersebut?;
„ Computer Vision: Object Recognition and
Description (Object Structure), Word and Vowel
Recognition – kata-kata apa saja yang membentuk
kalimat tersebut;
„ Artificial Intelligence: What is illustrated by this
image?, Speech Understanding – apakah arti dari
kalimat tersebut?.
17
Assignment-1 (Tugas Kelompok)
„ Bentuk kelompok yang masing-masing terdiri dari 3
mahasiswa.
„ Untuk setiap kelompok, cari 3 papers yang masing-
masing masuk kategori: (i) computer graphics; (ii) digital
image processing; dan (iii) pattern recognition atau
computer vision atau artificial intelligence. Lebih baik
lagi bila dari problem domain yang sama.
„ Setiap mahasiswa dalam setiap kelompok membahas 1
paper yang dipilih kelompok.
„ Buat laporan sesuai dengan format laporan (lihat
halaman berikut).
„ Laporan diserahkan pada tanggal 5 Oktober 2010. 18
Report LayOut

„ Tuliskan identifikasi kelompok dan anggota-


anggotanya.
„ Pengantar: jelaskan apakah ada kaitannya atau tidak
antara 3 papers yang dibahas dalam kelompok.
„ Secara berurutan (paper 1 s/d paper 3), tuliskan:
„ Judul paper dan identifikasi pembahas

„ Jelaskan mengapa paper tersebut termasuk pada

kategori yang dimaksud, serta jelaskan inti dari isi


paper (misal: input, proses, output).
„ Jumlah halaman laporan adalah 3 – 5 halaman.
19
Digital Image Processing: Applications

„ Orthodontic and Dentistry (Kedokteran Gigi);


„ Biomedical (Kedokteran);
„ Remote Sensing (Penginderaan Jarak Jauh);
„ Industry;
„ Gesture Language (Bahasa Isyarat);
„ Character Recognition (Pengenalan Karakter);
„ Digital Signature & Biometric Data (for e-Commerce,
Banking etc.).
20
Research Activities related to
Lab for Pattern Recognition, IP, and CBIRS
„ Fasilkom Multi-Lab Research: Mobile-based Batik Content-
Based Image Retrieval System (2009)
„ Sponsored by UI: Change Detection Based on Multidate
Remote Sensing Data (2009)
„ Sponsored by UI: Breast Cancer Image Detection Based on
Lab and Mammogram Data (2009)
„ Sponsored by UI: ICT based Health Services (2010)
„ Sponsored by DepDikNas: Content-Based Image Retrieval
System (2008-2010)
„ Sponsored by MenRisTek: Cultural Heritage Artefact
Inventory System (2010)
„ Sponsored by UI (being reviewed): ECG for Handling Sleep
Disorder (2010)
Orthodontic Application (1)
(Source: Budhiantini Bagyo, S2 Thesis, UI, 1993)

22
Image Registration: Citra Biomedis
Source: J. Kusnoto and A. Murni, 2007

Pasangan Titik Kontrol


Orthodontic Application (2a)
(Sumber: Joko Kusnoto et al., AICBET-2007)

Cephalo Image Indonesian Deuteromalay


(hanya contoh, Normative facial profile Preferred facial profile
bukan data pemilik foto). Display simulasi facial profile
Cephalo data untuk dengan metrik normative dan preferred.
menentukan nilai metrik. Untuk menampilkan simulasi hasil
treatment yang akan diperoleh.
24
Orthodontic Application (2b)
„ Orthodontic treatment is usually done based
on normative measures (normal average
measures)
„ As an alternative, before treatment the
patient could have an opportunity to choose
her own preference measures based on
simulated preference face profiles
„ The treatment is then done based on the
patient’s preference measures instead of the
normative measures
„ Further works: 3-D profiles
25
Dentistry Application (1a)
(Source: Tirza, S1 Technical Report, 2004)

Bregma
Nasion

X-Y Axis of The Forehead Sample points from (x0,y0) to (x31,y31)


(Source of Image: M. Inoue, 1990)
26
Dentistry Application (1b)
(Source: Tirza, S1 Technical Report, 2004)

„ A curve from Nasion landmark (lekukan dahi ke hidung)


to Bergman landmark (ubun-ubun) is extracted from a
cephalometric image;
„ Coefficients of Fourier transform and wavelet transform
are computed for the curve;
„ A discriminant analysis is used to determine the gender
type of the person recorded in the cephalometric image
based on the Nasion and Bergman curve and Fourier /
wavelet analysis using a number of man and woman
training sample data;
„ The experimental results show that the use of wavelet
transform gives better recognition accuracies compared
to the use of Fourier transform.
27
Biomedical Application (1)
(Source: Kartono, S2 Thesis, 1990)

Thorax X-Ray Landmarks Standard Thorax Tissue


(Source: A. Murni, 1990) (L. Sider, 1986) (Kartono, 1990)
28
Biomedical Application (2a)
Diagnosing a Pap Smear Cell Image
(Source: Farida and Addiati, 2007; Amalia and Phyllisia, 2008;
and E.P. Giri and A. Murni, 2008)

(J. Indarti, FKUI) (T. Farida, Fasilkom UI) (Jantzen et al., TUD)

Single pap smear cell image Seven condition categories


Nucleus
Cytoplasm 20 Features Classification Method
Background.

29
Diagnosing a Pap Smear Cell Image Based on
CBIR (2b)
Image to be Query By Example Searching
diagnosed
and
Matching

Retrieved Similar Images


carcinoma treatment

Diagnosed Image Database


carcinoma (Source of images:
Jantzen et al., TUD)
severe dysplasia

moderate dysplasia

mild dysplasia
ECG Analysis & Sleep Disorder
„ Device design
„ Deep sleep detection
„ Polsomnography and ECG
ECG and Arrythmia
(Sumber: Asep Insani, 2010)

Normal

Bradycardia

Tachycardia

Sudden Cardiac Death

ECG Signal Fourier Transform Spectrum


Remote Sensing Application (1)
(Source: A. Murni, 1997)

From Left to Right


Top to Bottom:
Optical-Sensor Image;
Classified Optical Image;
Joint Prob. Based Fusion;
SAR-Sensor Image;
Classified SAR Image;
High Rank based Fusion.
(Source of original /
unprocessed images:
BAKOSURTANAL RI)

33
Remote Sensing Application (2)
(Source: A. Murni, 1997)

From Left to Right


Top to Bottom:
Optical-Sensor Image;
Classified Optical Image;
Mosaic Image;
SAR-Sensor Image;
Classified SAR Image;
Fused Image.
(Source of original /
unprocessed images:
BAKOSURTANAL RI)

34
Remote Sensing Application (3)
(Source: Rohmah and Murni, 1997 and 2001)

From Left to Right


Top to Bottom:
Optical-Sensor Image;
Classified Optical Image;
Cloud free optical image;
SAR-Sensor Image;
Classified SAR Image;
Cloud-free classified image.

(Source of original /
unprocessed images:
BAKOSURTANAL RI)

35
(a)

Ground Truth
Multispectral
Image
Classification
(Source: S.H. Wijaya, 2008) 36
Automatic Image Registration
(Source: Skripsi S1 Gunawan, 2006)

Reference Image
Sensed Image
Registered Image
(Source of image http://imagers.gsfc.nasa.gov) 37
Research Design (2006-2010)
Content-Based Image Retrieval System

(Status 2008) 38
Query By Example to CBIR System
(Source of raw images: http://earth.google.com/
CBIR system: Eka Aditya, Fasilkom UI, 2006)

Example of
Vegetation Area

the least relevant


The 1st Indonesian Geospatial Technology Exhibition, Jakarta, August 23-27, 2006.
Query By Composition of Objects
(Source of raw images: http://earth.google.com/
CBIR system: Eka Aditya, Fasilkom UI, 2006)

30% Water; 30% Vegetation; and 40% Buildings.

The 1st Indonesian Geospatial Technology Exhibition, Jakarta, August 23-27, 2006.
Query By Example
(Mostly Tree)
(Source: S.H. Wijono, 2008)

(Source of raw images :


http://earth.google.com and
PT. The Map Indonesia Data)

Different scale of image can


also be retrieved.

41
Aplication in Industry (1)
(Source: Castleman, 1972)

Redness

Cherries Apples

Fruit Sorter Jackfruits


Lemons

Diameter

Feature Space Diagram (2 features)


42
Aplication in Industry (2)
(Source: Jain dan Murni, 1990)

Original Image Edge Image Recognized Circle


(MSU, 1990) (A. Murni, 1990)
43
Aplication in Industry (3)
(Source: Jain, 1990)

Types of Object Feature Space Diagram

44
Applications in Gesture Language
(Source: MSU, 1990)

Bahasa isyarat lainnya: menggunakan bahasa tangan dan ada


juga yang disebut sebagai ‘body language’; mengangguk (jarak
antara garis alis dan mulut mengecil), menggeleng (jarak antara
garis mata kiri dan kanan mengecil).
45
Applications in Character Recognition (1)
(Source: Murni, 1990)

Huruf A hasil scanning Huruf A terputus setelah ‘thinning’

46
Applications in Character Recognition (2)
(Source: Edi, 2002)

Huruf hasil scanning Huruf setelah ‘skeletonizing’

47
Applications in Character Recognition (3)
(Source: Skripsi S1 Juanita Rohali, 1991)

BAP AK BER UANG YANG J AH AT


PAD A ZAMAN DAHU L U KALA ADA SE E KOR BER
UANG YANG NAKAL DA N JA HAT
BER UA NG I TU BE RN AMA XAM I N
DIA S UKA ME M AKAN AN AK AN AK YANG TI DAK D
I SU KAI OLEH I BUN YA
SEHINGG A LA M A KELAMAAN ANAK AN AK DI D
UNIA ME NJADI MU SNAH B I NASA KAR EN A
DI MAKAN OLEH BAPAK BE RUAN G TERS EBUT
48
Human Biometrics & Features
(Source: Kompanets et al.; Uludag, 2000; Skripsi S1 Maukar, 1991)

Citra Wajah Citra Sidik Jari

49
Old Document Image Restoration
(Sumber: Hilda Deborah, 2010)

Need a further work in image preprocessing and postprocessing


Realism vs Abstractionism in Paintings
(Sumber: Tieta Antaresti, 2010)

Correctly Recognized as
Abstractionism painting (Left) and
Realism painting (Right)

Incorrectly Recognized as
Abstractionism painting (Left) and
Realism painting (Right)

Average recognition accuracy is


66.23%
CBIRS: Batik Application (1)
(Sumber: E. Aditya, E. Margaretha, A. Murni, 2009)

„ Feature Extraction:
„ Log Gabor Filter (sudah selesai untuk publikasi)
„ Keyblock (sedang dilanjutkan s/d Desember 2009
oleh Wahyudi)
„ Beberapa Contoh Pola Batik: (Sumber: M. Manurung, 2008)

Kawung Kawung Lereng Udan Parang Parang Parang


Beton Picis Liris Barong Chentung Curiga

Parang Parang Rusak Sekar Sido Sido


Kesit Barong Jagat Asih Mulyo
CBIRS: Batik Application (2)
(Sumber: E. Aditya, E. Margaretha, A. Murni, 2009)

„ Contoh Hasil Eksperimen:


Query Retrieved

Kawung Beton Kawung Beton Parang Barong (tidak relevan)

Kawung Beton Kawing Beton Kawung Picis

Total Average Precision 46.5%


Total Average Recall 31.5%
The End

You might also like