Image Processing - Fakultät Informatik — TU Dresden · • Image processing, and particularly computer vision, are important research fields • Their importance grows rapidly

© 2007 Thomas Brox 1

Image Processing

Image Processing

Class 1Introduction


Image Processing Organizational issues

• What type of lecture?– 2 hours classroom lecture, 2 hours tutorials– Tutorial consisting of mostly programming assignments and some theoretical

assignments– Introductory course in the fields of image processing, computer vision, and

pattern recognition– (Soft) requirement for starting a diploma thesis in this field

• Prerequisites– Undergraduate mathematics– Some programming experience in C/C++

(for the programming assignments)


Image Processing Organizational issues

• Grading policy– Oral or written exam (depends on the number of participants)

after the lecturing period– Regular attendance in the tutorials required to qualify for the exam

• Tutorials start next week

• These slides will be made available athttp://www.inf.tu-dresden.de/content/institutes/ki/is/bv_ws0708.de.html

• They are password protected. Access via…– username: open– password: sesame


Image Processing Other courses

• This winter term:– Pattern Matching in Computer Vision (2+2, called Mustererkennung)

• Next summer term (planned):– Computer Vision (2+2) – Image Segmentation (2+2)

or Pattern Recognition and Machine Learning (2+2)


Image Processing What is an image?

• Digital image: regular grid of intensity values

• Continuous function

Author: Daniel Cremers


Image Processing Where do images and image processing play a role?

• Imaging (e.g. photography, ultrasound, x-ray, magnetic resonance)

• Image enhancement (e.g. Adobe Photoshop)

• Image and video compression

• Computer graphics (model image)

• Computer vision (image model)

• Focus of this and the following courses: image enhancement and computer vision


Image Processing Imaging – ultrasound, MRI

Ultrasound image of a human embryo Magnetic resonance image of a human head


Image Processing Image enhancement - denoising


Image Processing Image enhancement - deblurring

Welk et al. 2005: variational deblurring


Image Processing Image compression - JPEG


Image Processing Computer graphics – water simulation

Guendelman et al. 2005: Simulating the flow of water with level sets


Image Processing Computer vision – 3D reconstruction

Kolev et al. 2007: multiview reconstruction


Image Processing Computer vision - segmentation

Wedel et al. 2007: obstacle detection and segmentation


Image Processing Computer vision – human tracking

Brox et al. 2007: 3D human pose tracking


Image Processing Computer vision – human tracking

Brox et al. 2007: 3D human pose tracking


Image Processing Difference between image processing and computer vision

• Usually, the term “image processing” is used for the general subject of processing images with a computer

• Computer vision signifies the specific task to make the computer interpret the content of images

• This is a difficult and so far unsolved task

• Computer vision has similar motivations as artificial intelligence and cognitive neuroscience


Image Processing Why is computer vision difficult?

• Vision is a natural and easy task for humans (and many animals)

• This is not for free: ~50% of the primate’s cortex deals with the processing of visual information (Felleman-van Essen 1991)

• Making a computer see like humans see means to solve a large part of the AI problem (this cannot be easy)

• What do you see in this image?

• A lot of high level knowledgeand content information necessary


Image Processing

• What’s that?

• This is how the computer sees Einstein

• Demonstrates what our brain achieves at the unconscious level

Images are only a structured grid of numbers

Author: Daniel Cremers


Image Processing Importance of the spatial arrangement

• Image content is defined by the spatial arrangement of intensities

• It is not sufficient to treat images as vectors and to analyze these vectors

Zebra image Same image with a different row length


Image Processing Importance of image processing and computer vision

• Computer vision is a very young research field– Main computer vision conference (ICCV) founded in 1987– Main computer vision journal (IJCV) founded in 1988

• Nonetheless, it belongs already to the most influential subfields of computer science; tendency: growing rapidly

• By 2006, the International Journal of Computer Vision had the highest impact factor of all computer science journals: 6.085

• Impact factor in 2002: 2.03

• Computer vision applications not restricted by a small market but by the limited quality provided so far

• More research more products

INT J COMPUT VISION 6.085ACM T INFORM SYST 5.059BIOINFORMATICS 4.894MIS QUART 4.731IEEE T PATTERN ANAL 4.306ACM COMPUT SURV 4.13ACM T GRAPHIC 4.081J AM MED INFORM ASSN 3.979IEEE T EVOLUT COMPUT 3.77IEEE T MED IMAGING 3.757NEUROINFORMATICS 3.541J CHEM INF MODEL 3.423VLDB J 3.289MED IMAGE ANAL 3.256J ACM 2.917J CRYPTOL 2.833IEEE T IMAGE PROCESS 2.715MACH LEARN 2.654IEEE T NEURAL NETWOR 2.62IEEE WIREL COMMUN 2.577COGNITIVE BRAIN RES 2.568IEEE T MOBILE COMPUT 2.55CHEMOMETR INTELL LAB 2.45IEEE INTELL SYST 2.413HUM-COMPUT INTERACT 2.391J MOL GRAPH MODEL 2.371J BIOMED INFORM 2.346J COMPUT PHYS 2.328DATA MIN KNOWL DISC 2.295IEEE ACM T COMPUT BI 2.283ARTIF INTELL 2.271J MACH LEARN RES 2.255NEURAL COMPUT 2.229IEEE NETWORK 2.211ACM T DATABASE SYST 2.143COMPUT BIOL CHEM 2.135IEEE T SOFTWARE ENG 2.132INFORM MANAGE-AMSTER 2.119QUANTUM INF COMPUT 2.105J COMPUT AID MOL DES 2.089IEEE T KNOWL DATA EN 2.063IEEE PERVAS COMPUT 2.062J COMPUT BIOL 2MATCH-COMMUN MATH CO 2NEURAL NETWORKS 2

Impact factors of computer science

journals 2006


Image Processing Some image processing applications

• Quality control, visual inspection

• Security systems– Fingerprint recognition– Iris recognition– Face recognition– Tracking

• Medical systems– Image enhancement– Data analysis– Routine diagnostics

• Entertainment industry– 3D reconstruction– Motion capturing– Augmented reality– Human-machine interaction

• Information systems– Document analysis– Image Google

• Earth surveillance– Weather forecasts– Google Earth

• Driver assistance systems– Lane control– Collision avoidance– Autonomous driving

• Robotics– Autonomous robots (e.g. pathfinder)– Domestic robots

• Artificial intelligence


Image Processing Related sciences

• Computer science– Audio processing– Pattern recognition– Optimization– Machine learning– Data retrieval– Computer graphics– Robotics– Control theory– Software engineering

• Mathematics– Numerics– Statistics– Linear algebra– Functional analysis– Graph theory– Geometric algebra

• Physics– Optics– Computational physics

• Electrical engineering– Signal processing

• Neuroscience– Neurophysiology– Computational neuroscience– Psychophysics

• Medicine

• Philosophy


Image Processing Conferences and Journals

• Conferences– ICCV: International Conference on Computer Vision– ECCV: European Conference on Computer Vision– CVPR: Int. Conference on Computer Vision and Pattern Recognition– NIPS: Neural Information Processing Systems– DAGM: Tagung der Deutschen Arbeitsgemeinschaft für Mustererkennung– ICIP: International Conference on Image Processing– ICPR: International Conference on Pattern Recognition– SSVM: Int. Conf. on Scale Space and Variational Methods

• Journals– International Journal of Computer Vision– IEEE Transactions on Pattern Analysis and Machine Intelligence– IEEE Transactions on Image Processing– Computer Vision and Image Understanding– Image and Vision Computing– Journal of Mathematical Imaging and Vision– Journal of Visual Communication and Image Representation– Realtime Imaging– Pattern Recognition


Image Processing Overview (planned)

• Class 1: Introduction• Class 2: Image acquisition and representation• Class 3: Noise, point operations• Class 4: Fourier transform• Class 5: Linear filters• Class 6: Wavelets• Class 7: Morphological filters• Class 8: Nonlinear diffusion filters• Class 9: Variational methods I• Class 10: Variational methods II• Class 11: Deconvolution• Class 12: Texture analysis and filtering• Class 13: Color spaces• Continued next term with the course Computer Vision


Image Processing Summary

• Image processing, and particularly computer vision, are important research fields

• Their importance grows rapidly with the number of successful applications

• Image processing makes use of techniques from various other sciences

• Especially mathematics provides many helpful tools


Image Processing Literature

• R. C. Gonzalez, R. E. Woods: Digital Image Processing, Addison-Wesley, Reading, 2nd Edition, 2002.

• J. Bigün: Vision with Direction, Springer, Berlin, 2006.

• K. D. Tönnies: Grundlagen der Bildverarbeitung (in German), Pearson Studium, Munich, 2005.

• CV Online: Online compendium on numerous image processing and computer vision topics,http://homepages.inf.ed.ac.uk/rbf/CVonline/

Documents

Image Processing - Fakultät Informatik — TU Dresden · • Image processing, and particularly computer vision, are important research fields • Their importance grows rapidly