EECS 396L/490: Digital Image Processing

Fall 2004, Prof. Frank Merat

General Information

Class Lectures: TR 4:30-5:45pm, Rockefeller 309

Course Instructor: Frank Merat, flm at po.cwru.edu, Glennan 518, x4572

Teaching Assistant: Xiying (Frank) Li, xxl20 at po.cwru.edu

Instructor's Office Hours:

• Monday 11am-12 noon except when I am meeting with visiting prospective students
• Other times by appt.

 

Required Text

• Rafael C. Gonzalez and Richard E. Woods, Digital Image Processing, 2nd Edition. Prentice-Hall, 2001. ISBN:0-20-118075-8.
  Buy it for $117 (new) from www.bn.com: Digital Image Processing
  Buy it for $117 (new) from www.amazon.com: Digital Image Processing

Syllabus

• Syllabus (DOC, 32k) Reflects current grading policy. Mid-Term and Final will be projects. Watch for postings below.

Homework

I will typically pass out solutions in class.

• Homework 1 Due September 10th (PDF, 31 kB)
• Homework 2 Due September 23rd (PDF, 36 kB)
• Homework 3 Due September 30th (PDF, 36 kB)
• Homework 4 Due October 14th (PDF, 28 kB)
• Homework 5 Due October 28th (PDF, 28 kB)
• Homework - Extra Credit Due November 9th (PDF, 28 kB) You may do either problem, both or none. Your report for the second problem should be no more than 1-2 pages including figures.

Computer Assignments

• Project 1 Histogram equalization. Due September 14th (PDF, 66 kB)
• Project 2 Spatial image processing. Due September 21st (PDF, 40 kb)
• Project 3 Basic Fourier transforms, magnitude and phase information, periodic noise removal. Due September 28th (PDF, 123 kb)
• Images for computer projects are available through the"Images" link below.
• Project 4 Periodic noise additiona and removal. Due October 14th (PDF, 36 kb)
• Project 5 Color histogram equalization. Due November 2nd (PDF, 30 kb)
• Project 6 Color image segmentation. Due November 9th (PDF, 30 kb)
• Project 7 Principal component analysis. Due November 23rd (PDF, 30 kb) <<Change in date due to my illness.

Mid-term and Final Projects

In lieu of a mid-term and final exam you are required to do two computer projects each of which will be worth 25% of your final grade. In general, your projects will be chosen from those in the following list. If you have an idea for another project please discuss it with me and I may accept it for one of the listed projects. The mid-term project is due November 29th (change in date due to my illness); the final project is due December 9th. Details on the format of your report will be provided shortly.

• Project 1 - OCR preparation. It is quite common when you are digitizing a copy of an original document to encounter pages of text which are off center and skewed by an unknown angle. Devise an algorithm which will (a) detect the skew angle and align the text to be horizontal, and (b) will fit a given bounding box to the text, erasing all image information outside the text box. Demonstrate an implementation of your algorithm. Several images will be provided to test your program. However, you may immediately begin testing your programs by simply rotating "Image5" and "Mesh_text" by a known angle in MATLAB and determining if you can estimate this angle using your algorithm. Three starting references on text de-skewing are provided in the Course References.
• Project 2 - Face recognition. The goal of this project will be to locate faces in a given image which will contain one or more faces. The goal of this project is not to recognize the person as some sophisticated security systems now do, but simply to locate faces in an image. Common techniques include color segmentation and principal component analysis. Three starting references on face detection are provided in the Course References. I have provided a number of facial images which you may use for this project.
• Project 3 - Counting Chocolate Chips. Image processing is being increasingly used in food inspection. The goal of this project will be to locate and count the chocolate chips visible in images of chocolate chip cookies. Possible techniques include color segmentation, filtering and morphological operations. A starting references on counting chips is provided in the Course References.
  • Zip file of top images of chocolate chip cookies (zip, 9.15MB)
  • Zip file of bottom images of chocolate chip cookies (zip, 9.78MB)
  • Zip file of circus cookies (zip, 2.57MB)
• Project 4 - Photomask Analysis. A critical element of semiconductor fabrication is projecting a pattern on an optical mask onto a wafer. These masks are designed on a computer and then optically written onto very accurate, flat glass plates. As the required resolution increases (already at a small fraction of a micron) many types of production errors appear in the masks. These errors take the form of geometric distortions (warping), noise (both Gaussian and salt&pepper) and defects (which takes the form of large black and white spots) on the mask. The goal of the project is to compare an image of the produced mask (containing unknown warping and noise) with the as-designed mask image and subtract the two images to identify the defects. This project was used at Stanford in their EE368 course in the Spring 2004 semester. A significant number of triplets of images (image_DB indicated the as-designed mask image; image_D indicates the produced mask with noise and warping and errors due to dirt, and image_mask is an image of the errors your algorithm should find) are available for download from the Stanford EE368 project Web site. These are listed in Steps 3 and 5 on Stanford's Web page. You should use the images from Step 5 to evaluate your program (i.e., how successful was your algorithm at finding all the errors) for your report. You may ignore Stanford's Step 4. Implementation Details with the exception that your program should be implemented in MATLAB. Thanks to Prof. Girod of Stanford of letting us use his class project Web page and images!
  • Zip file of image triplets containing relatively easily detected defects (zip, 4.44 MB)
  • Zip file of image triplets containing harder to detect defects (zip, 2.97 MB)
  • Zip file of image triplets containing a mix of defects which you should described in your report (zip, 2.55 MB)

Report format (Mid-term and Final Projects only)

• CALL FOR PAPERS. To provide practice in writing papers these project reports will be modeled after a typical conference paper:
  • Full Paper— The final paper is due one week after the last class. This paper must be no longer than six typed pages, including figures. See the grading policy for guidelines on how the content of the paper will be assessed. A dual column format similar to IEEE conference papers is encouraged. Download a conference paper in .doc format (919 kB) which was presented at IEEE Sensors 2003 in Toronto. You may use it as a template for formatting your report.
  • Submission procedure — Papers should be submitted to me electronically by e-mail. There should be no problem with file size as university e-mail can now handle 50 MB attachments. The document submitted must be in Word doc format (preferred) or PDF.
    
• GRADING POLICY. The ultimate goal of the paper is to demonstrate to me that you have learned something from the course. However, most technical papers are peer reviewed by known experts in the field who will rank the paper according to different criteria including originality and technical contribution. In the spirit of this process I will review your paper and assign two letter ranks* using the following criteria.
  • Content — this will be 70% of the total paper grade
    Abstract
    Key words which can be used to electronically search for your paper
    Introduction. This is usually a description of the problem.
    A description of what you did. This can be broken down into several sections and is where your results should be included.
    A short summary of what you did and how well it worked.
    References. You should properly use references and footnotes. Failure to do so will result in loss of points.
    Any acknowledgements.
  • Style — this will be 30% of the total paper grade
    Clear, concise text
    Appropriate figures, charts, tables
    Spelling and grammar.
    * Letter ranks will be translated into numerical grades using the following equivalences.
    
    

    A+	100%	B+	88%	C+	78%
    A	95%	B	85%	C	75%
    A-	90%	B-	80%	C-	70%

  
  

Student Papers

Student papers from the Fall 2004 semester are now available for downloading.

Resources

• There is an extensive Gonzalez and Woods Digital Image Processing, 2/e Web site which contains all the artwork and images found in the textbook, some homework solutions, and a lot of other useful stuff to accompany the text.

MATLAB

The course will make extensive use of MATLAB and the Image Processing Toolbox.

• Gonzalez and Woods have teamed up with Eddins from the Mathworks to write Digital Image Processing Using MATLAB which is very similar to our 2/e textbook and has additional MATLAB material but no homework.
• Mathworks has a very good set of Image Processing Toolbox demos along with the source code.
• There is a pretty good list of MATLAB tutorials at this Web site.
• Sandberg at UC Boulder has written an Introduction to the MATLAB Image Processing Toolbox.
• Check out Prof. Branicky's EECS 246 Signals and Systems website for more MATLAB tutorials.
• It never hurts to check out Mathworks' Image Processing Toolbox website for MATLAB help, files, etc. They also have many examples of code which you may find useful.
• Mathworks has a forum for the exchange of several thousand MATLAB routines you may find useful. There are at least 150 image processing functions available for downloading.
• Peter Kovesi has a really interesting Web site containing many useful image processing and computer vision functions implemented in MATLAB.
• Mathworks maintains a page of links to MATLAB related image processing sites.

POWER POINT PRESENTATIONS

Several of you have expressed a desire for a more convenient local copy of the author's power point presentations.

• Chapter 1 describes imaging and shows a sample of the images one can work with. (PPT, 5.02 MB)
• Chapter 2 presents some fundamentals of digital images such as gray levels, spatial resolution, and connectivity. (PPT, 2.53 MB)
• Chapter 3 describes gray level transformations, histogram equalization, and spatial filtering including low-pass, gradient, Laplacian. (PPT, 6.45 MB) UPDATED 9/10 with my MATLAB slides.
• Chapter 4 describes Fourier domain image processing techniques. (PPT, 4.27 MB) UPDATED 9/10 with my MATLAB slides.
• Chapter 5 describes image reconstruction techniques including Wiener filters. (PPT, 6.80 MB)
• Chapter 6 describes color representation and color image processing. (PPT, 8.54 MB)
• Matrices and eigenvectors reviews basic matrices and vectors, vectors spaces and basis vectors, and eigenvalues and eigenvectors. The Hotelling transformation is then used to analyze image regions and boundaries (Section 11.4 of GW). (PPT, 1.3 MB)
• Chapter 7 presents basic wavelets and their application to image processing. (PPT, 3.68 MB)

Images for various assignments

Lecture Notes/Other Course Material

PPT notes with my lecture notes.

• Chapter 3 (PDF, 3.76 MB)
• Chapter 4 (PDF, 2.47 MB)
• Chapter 5 (PDF, 3.22 MB)
• Chapter 5 Updated 10/23 (PDF, 3.42 MB)
• Chapter 6 B/W (PDF, 2.40 MB)
• Chapter 6 Gray (PDF, 3.54 MB)
• Chapter 6 Color (PDF, 8.27 MB)
• Chapter 7 (PDF, 2.85 MB)
• Chapter 9 (PDF, 1.78 MB)
• Chapter 10 (PDF, 1.40 MB)
• Chapter 13 Castleman (PDF, 860 kB)
• Linear Algebra & Eigenvectors from various sources (PDF, 2.3 MB)

Other Resources

• EEAP 431Digital Image Processing as I taught it in Spring 1996 using Castleman's Digital Image Processing.
• EEAP 431Digital Image Processing as I taught it in Spring 1997 using Gonzalez and Woods, Digital Image Processing, 1/e.
• The Image Processing Handbook, 4th Edition by John C. Russ is an excellent guide to what you can do with image processing. The book contains hundreds of illustrated examples. Many of the algorithms demonstrated in the book are available on a fairly expensive [$249.95] CD-ROM from Reindeer Graphics. They are implemented as Photoshop® plug-in's .
• Generate your own color cube. Check out this program for printing your own color cube. [Some assembly with scissors and glue is required.]
  

Other Course References