Instructor: Pamela Cosman
Phone: 822-0157, e-mail: pcosman@ucsd.edu
Office hours: Tuesday 2-3, Friday 12-1, and by appointment
Office: EBU-1, Room 6407
Teaching Assistants:
Tsung-Yi Lin, tsl008@ucsd.edu
Kritika Muralidharan, krmurali@ucsd.edu
Office hours for Tsung-Yi: Thursday 11am-12pm, in room 4506,
Friday 3pm-4pm, in 5706
Office hours for Kritika: Wed 5-6 in room 5706
Administrative Assistant:
Molly Condon, mcondon@ece.ucsd.edu
858-534-2222, EBU-1 room 2605, 7:30am to 4:00pm.
Course Requirements and Grading:
6 Homeworks/Lab Exercises, 30%
Quiz 1, 15%
Quiz 2, 15%
Final Exam, 40%
Text: Digital Image Processing by Gonzalez and Woods. 3rd edition, Prentice Hall, 2008
3/22 Here is the Solution to the Final Exam. Final exams can be retrieved from Molly Condon in EBU-1 room 2605 starting Tuesday March 27. If you want to ask for a re-grade, do not remove the exam from her office. Just look at it there, and attach a note to your exam saying what is the basis for your re-grade request. If you are NOT planning to ask for a re-grade, then you can take your exam away from her office.
3/16 Here is the Solution to homework set 6. Note also that for the final exam, the 3 sheets of paper you are allowed for notes can have front and back both.
3/15 The final exam will be at 11:30 on Monday. It will be held in two rooms: our usual room as well as WLH 2207. This will allow students to be spaced out with one empty seat between students. Please come a few minutes early so we can get everyone sorted out into the two rooms with correct seating. During the exam, any clarification which is given out in one room will be given in the other as well, so it doesn't matter which room you're in.
3/12 Here is the Solution to homework set 5.
3/11 On this last week, Wed and Fri lectures will be essentially used for review, not for new material. If you have questions, or particular topics you'd like me to review, send me an e-mail. For those of you who tend to skip lectures, maybe you should attend Wed/Fri (or at least get good notes from someone). For the final exam, you can have 3 sheets of paper with notes. The final will be comprehensive, so you might want to just retain the two sheets which you had for quiz 2, and create an additional one which covers the new material since quiz 2.
3/9 Here is Homework set 6 and the files you need for it: coke.tif, tran.tif.
3/7 Kritika asked me to clarify something about the Hurst operator. It is a plot of log(range) on the y-axis versus log(distance) on the x-axis. Here we are including in the range calculation all the pixels at distances less than or equal to the distance whose log is on the x-axis. So, as we move to the right on the plot, the range can only increase (or stay the same). So the slope of the line fitting the points can be zero or positive, but never negative.
3/6 Here is an item on Image Registration. Please print it out... it will be helpful in class on Wed 3/7 and even more helpful on Friday 3/9.
3/2 Here is an item on Gray Level Co-Occurrence Matrices. It was handed out and discussed in class on 3/2.
3/2 Here is the Solution to Quiz 2. Kritika graded quiz 2, so talk to her if you have re-grade issues. mean score is 25.3. Median is 24.5. Range is 9 to 39.
2/28 Here is Homework set 5 and the files you need for it: oranges.tif, cell.tif, badcell.tif, truth.tif,
2/25 Mean scores on the homeworks: HW1 79.3, HW2 78.6, HW3 71.9, HW4 82.2
2/23 I was asked whether you need to know the equations in the textbook on the watershed algorithm. There are a few equations relating to watersheds on page 775, but you don't need to know those. You need a conceptual understanding of how the watershed algorithm works and what it is used for.
2/22 Here are the Answers to the Practice Problems.
2/21 Here are some Practice Problems to help you prepare for the next quiz. Obviously the actual quiz wouldn't have this many problems. The quiz may include anything from the first lecture through the Wednesday 2/22 lecture. The emphasis will be on material after the first quiz, but something from earlier will be included too. You may have two pages of notes with you for the quiz. You can find additional good problems at the end of each chapter in the textbook (some of these practice problems are in fact from the textbook).
2/21 Here is the Solution to homework set 4.
2/13 Here is the Solution to homework set 3.
2/9 Here is Homework set 4 and the files you need for it: bookcover.tif, plotLines.m, plotSHT.m, plotPeaks.m, cie
2/9 Some comments from Tsung-Yi about data types.
2/7 Quiz 1 statistics: Mean = 31, median = 32, S.D. = 6.4, Range = 15-40. Kritika graded the quiz, so please direct re-grade questions to her.
2/6 Here are some notes and figures related to color. Please print this out, and bring it to class on Wed 2/8 and Fri 2/10.
2/6 A company named ecoATM is looking to hire a graduating senior for a computer vision engineering job. The job info is here. ecoATM makes kiosks that buy back used cell phones (you may have seen some of them in the Westfield malls or at Ralphs). The computer vision goal is to identify what type of phone you have and assess its condition. You can contact Jeff Ploetner at jploetner@ecoatm.com if you're interested.
2/6 For homework 3, the true binary edge map
for problem 1 was obtained using imfilter, and it had 11368 edge pixels.
You might get a different number if you use filter2 instead, largely because
of the different treatment of the boundary of the image. The boundary of
the image might show up as edge pixels, or might not, depending on how it
is treated. So don't worry about this if your number is off from the one
in the homework set. But the best way to do this is to have
the boundary NOT get counted as edge pixels. You can do this with filter2
for example by using 'valid' where the filtered output is slightly smaller
than the input array, and is defined only where the filtering operation
does not hang over the boundary of the image.
Another issue someone pointed out is that pyr7 will give an output
of type double, which will show up as all white if you use imshow. Might
need to convert it.
2/4 Here is Homework set 3 and the files you need for it: peppers512.tif, noipep512.tif, pyr7.m, medfilt.m.
2/4 Here is the Solution to Quiz 1.
2/2 Kritika has added office hours:
Wed 5-6 in room 5706
Also, here are some
quiz 1 review notes from Kritika's
Wed 2/1 discussion section.
1/31 Here are my lecture notes for Jan 30. Also, Tsung-Yi mistakenly took some points off if you didn't include all the figures for homework 1. The homework didn't explicitly say you had to include all the figures. So you can go see him during office hours and get those points back. However, please do try to make a homework report which is clear and complete!
1/30 Note: the 1st quiz will cover everything up to and including the lecture of Friday Jan 27. Don't forget: You can use 1 page of notes for the quiz. Use the page well! Here is the Solution to homework set 2.
1/27 Tsung-Yi graded the 1st homework set. If you have grading questions, send him an e-mail or talk to him in office hours.
1/27 List of directional derivatives
1/25 Here is the Solution to homework set 1. The last page has a couple of useful Matlab commands from the TA.
1/25 A couple of people were hoping for
homework help on Friday afternoon, instead of 2 TA hours on Thursday.
So the TA office hours will be:
Thursday 11am-12pm, office 4506
Friday 3pm-4pm, office 5706
1/24 Quiz 1: You can bring one standard 8.5 by 11 inch sheet of paper, with writing on both sides of the page, to quiz 1. Other than your one sheet of notes, the quiz will be closed-book and closed-notes.
1/24 Here is the
2nd homework set and the files you need for it:
pep.tif,
n2.tif,
n4.tif,
bab.tif, and
bab2.tif.
The first problem uses matlab. Problems 2-6 are pencil-and-paper problems,
no computer needed. Problems 2-6 are the style of problems which appear on
quizzes and exams. A typical 50-minute quiz would probably have 4
problems. Problems 2-6 shouldn't take you much more than an hour.
1/21 I realized the eyechart.tif image which
I gave you before is of datatype uint8, which some versions of Matlab don't
like for binary image processing operations. And also it had values 0 and
255, rather than 0 and 1. Here is a different version:
neweyechart.tif which is a logical datatype
and has values 0 and 1,
which should make certain versions of Matlab happier. The previous image
could be converted with a command like
bw = round(double(eye)/255);
which both converts the datatype and makes the values 0 and 1.
Also we could have done
bw = 1- round(double(eye)/255);
which will flip the 0's to be 1's and vice versa. This is because we want
the letters to be processed as objects (so they need to have value 1).
If you don't flip the 0's to be 1's and vice versa,
then an operation such as erosion,
which should make the letters smaller, will instead make them larger,
because it is the background which is being treated as the object,
and which is getting eroded.
1/20 Quiz dates:
Quiz 1 will be on Friday Feb 3.
Quiz 2 will be on Monday Feb 27.
1/17 Kritika wrote up a set of background notes on convolution
1/13 I will be out of the office Tuesday afternoon 1/17, so my Tuesday office hours 2-3 will be covered by Tsung-Yi (in my regular office EBU-1 6407). For people who missed class today to take an extra long weekend, we finished iterative modification schemes (see the extra page of notes provided below) and covered generalized dilation, erosion, opening and closing (topics very well covered in your textbook, section 9.1, 9.2, 9.3, just read those).
1/13 Here is the 1st homework set which is due on Monday January 23. The image files you need for it are eyechart.tif and bincell256.tif.
1/12 Here is some material on Iterative Modification Schemes which will be helpful for homework 1.
1/10 Here is some material on binary image processing. Please print it out and bring it to class Wed & Fri. That will make taking notes a lot easier in class.
1/10 The first discussion section will be Wed at 1 in WLH 2205. Kritika will review 2-d convolution and filtering and various other useful background items. TA office hours have been determined: Thursday 11-12 and 1-2 in EBU-1 room 4506.
1/6 Regrading policy: If you wish to request a regrade of any material (homework or quiz or final) you must submit the request, in writing, to one of the TAs with 1 week of the date that the graded material is handed back to the class. Otherwise the regrade will not be considered.
1/6
The following sections of the textbook (in roughly the following order)
will be covered in this class,
in case you want to get started on the reading:
Introductory material: Chapter 1 (all), Chapter 2: sections 2.1, 2.2, 2.4
Chapter 3, sections 3.4.1, 3.4.2
Binary Image Processing: Chapter 2: section 2.5, Chapter 9: sections
9.1 through 9.5.
Noise reduction and Edge Detection: Chapter 3: section 3.5, Chapter 10,
sections 10.1, 10.2
Color: Chapter 6: sections 6.1 through 6.7.
Segmentation: Chapter 10, sections 10.3 through 10.5
Texture analysis: Chapter 11, section 11.3, primarily 11.3.3
Registration: material not in textbook
Image compression: Chapter 8: section 8.1, portions of 8.2
1/6 Here is
a Matlab tutorial for the image processing toolbox (I'm sure you can find
lots of other such tutorials...):
http://www-cse.ucsd.edu/~sjb/classes/matlab/matlab.intro.html
Pamela Cosman / pcosman@ucsd.edu