630 likes | 866 Views
Introduction to Computer Vision. Lecture 1 Dr. Roger S. Gaborski. Course Goals. Obtain a working knowledge of computer vision Gain an understanding of current research in computer vision Become familiar with programming in the MATLAB environment. Where to Find Me. Office: 70 – 3647
E N D
Introduction to Computer Vision Lecture 1 Dr. Roger S. Gaborski
Course Goals • Obtain a working knowledge of computer vision • Gain an understanding of current research in computer vision • Become familiar with programming in the MATLAB environment RS Gaborski
Where to Find Me • Office: 70 – 3647 • My lab 70-3400 • Office Hours: • Tuesday noon – 2:00pm , in Rm3400 or 3647 • Email: rsg@cs.rit.edu • Homework email: rsg_introcv@gmail.com RS Gaborski
Course Outline • Textbook – Digital Image Processing using MATLAB • SECOND EDITION 2009 Gatesmark Publishing • Online MATLAB tutorial-Register at Mathworks: • http://www.mathworks.com/academia/student_center/tutorials/launchpad.html • Topics • Homework • Quizzes and Exams • Projects (4005-757 only) • Grading • Webpage: www.cs.rit.edu/~rsg (includes course calendar on CV page) • Lecture slides will not always be posted on webpage RS Gaborski
Homework • Questions concerning Homework • Do not wait until the night before its due to start working on the HW • Ask questions in class concerning HW • First, ask the TA during his office hours • If TA cannot answer your questions, see me during my office hours • Do not send me email concerning the HW after noon the night before it is due. I will not be able to respond to your email. RS Gaborski
Grader • Jeffrey Zullo • E-Mail: <jlz9811@rit.edu> RS Gaborski
Grading - With Final • Homework 35%(457) 20%(757) • Quizzes/Exams 45% 45% • Final 20% 20% • Project* --- 15% • No Project for 4003-457 • *Project: 757 Individual only, weekly presentation updates RS Gaborski
Grading - Without Final • Homework 35%(457) 20%(757) • Quizzes/Exams 65% 65% • Project* --- 15% • No Project for 4003-457 • *Project: 757 Individual only, weekly presentation updates • Exam and Quiz grade must be at least 80% at the end of week 10 to be excused from final RS Gaborski
Course Grade • 90%-100% A* • 80%-89% B • 70%-79% C • 60%-69% D • <60% F * Note: For example, 89.4 is a ‘B’, 89.5 is rounded to 90 which is an ‘A’ RS Gaborski
Project • Choose from a list of projects provided on course Project Page • Ten minute verbal proposal presentation (see course calendar) • Verbal updates (see course calendar) • *Project grade includes verbal proposal, verbal update reports and final presentation, code and report RS Gaborski
Images are Everywhere • On the web – flickr, Google Images, YouTube • On your computer – iPhoto, Picasa • Video Surveillance: • Streets • Hotels • Businesses • Parking lots RS Gaborski
Computer Vision – Interpretation of Images • Digital photographs • Medical radiographic images • Functional magnetic resonance imaging (fMRI) • Medical ultrasound • Industrial radiographic images • Digital video images • Satellite images • Astronomy RS Gaborski
Digital Image RS Gaborski
Digital Image RS Gaborski
Digital Image RS Gaborski
Medical Related Images Information obtained from images: Bone structure Soft Tissue Brain Activity
Medical Radiographic Image www.4umi.com/image/x-ray.jpg RS Gaborski
Medical Ultrasound http://keystone.stanford.edu/~huster/photos/i/ultrasound.640.jpg RS Gaborski
Functional MRI A 20-year old female drinker A 20-year old female nondrinker Response to the spatial working memory task. Brain activation is shown in bright colors. RS Gaborski www.alcoholism2.com/
Industrial Applications Non Destructive Testing Inspection / Security
Industrial Radiographic Image www.vidisco.com/ CabinetXrayMic80A_01.htm RS Gaborski
Industrial Radiographic Image Pseudo- color www.vidisco.com/ CabinetXrayMic80A_01.htm RS Gaborski
Satellite Images RS Gaborski www.noaa.gov
Astronomy Images www.sdsc.edu/ sciencegroup/astronomy/ RS Gaborski
Astronomy Images astro.martianbachelor.com/ RS Gaborski
A Few Observations • Object recognition is a very difficult problem • Objects can be rigid, or flexible • Finding a specific object ( is easier than finding all objects that belong to a category RS Gaborski
Find a 911 Porsche RS Gaborski
Find All Cars in an Image RS Gaborski
What About Background IssuesSeparating the car from the background RS Gaborski
Image Database Problem • Assume you have taken pictures with your digital camera the last three years • You now have 4000 pictures stored on your computer’s hard drive • How do you sort them? RS Gaborski
Student Result RS Gaborski
More Categories RS Gaborski
How Else Could You Identify Locations? KISS- simpler approach to recognize location that recognizing objects in the image?
iPhoto 09 "Places" Geotagging • http://www.youtube.com/watch?v=GVW8700LrvE RS Gaborski
How do you find a particular face • How do you find a particular object in an image? • Faces • Cars • Buildings • etc RS Gaborski
Level of Vision • Low level processing: • Pixel level • Gradient (uniform area, edges) • Intermediate level processing: • Group pixels into line • High level processing • Interpretation of a scene beyond grouping RS Gaborski
Image Models • Task: “Look for an object in an image” • Assume the task is to find rectangle and washer objects RS Gaborski
Image models, continued RS Gaborski
Image Models • Task: “Look for an object in an image” • Assume the task is to find rectangle and washer objects • Find edge pixels and group • Find outlines of objects in the image • Create a model of the object • Rectangle: Four straight lines, Opposite lines equal in length, 90 degree angles, lines connected • Washer: Two concentric circles RS Gaborski
Image models, edges RS Gaborski
Image models, continued One object partially overlaps another RS Gaborski
Objects are 3 Dimensional Rotating Disk Frame 1 Frame 2 Frame 3 RS Gaborski
License Plate Model • Rectangular (depending on viewpoint) • Aspect ratio 2:1 • Textures (characters on license plate) RS Gaborski
Face Model http://www.faceresearch.org/ RS Gaborski