MATH4900 Seminar

MATH4900 Seminar MATH4900 Seminar

Course Description: This seminar course is on variational methods for segmentation and their applications to different problems in imaging, including vascular segmentation, hyper-spectral image classification, and point cloud classification.

Course Format: The students will first present some background materials on segmentation and then be divided into groups to work on different applications of segmentation. Each group will report their progress once every two weeks in the 1-hour bi-weekly meetings. They will have to give two presentations (one at mid-term and one at end-of-term) about their projects and to write a report to summarize their projects at the end.

Presentations are to held at the assigned classrooms. Bi-weekly meetings are in Raymond Chan's office (LSB 201) and the times are to be arranged between the teachers and students beforehand.

Lecture Hours: Monday 4:30pm-6:15pm at LSB C1 and Wednesday 12:30-1:15pm at LSB C4

Office Hours: Every Tuesday 5:30pm-6:15pm, please send me an email (rchan@math.cuhk.edu.hk) or call me first (3943-7970), if possible.

About the Lecturer: Raymond Chan

Reading Materials:

Chapter 10, Digital Image Processing by R. Gonzales and R. Woods, Pearson Education Inc., 2010.
Chapter 5, Image Processing, Analysis, and Machine Vision by M. Sonka, V. Hlavac and R. Boyle, PWS Publishing, 1999.
X.H. Cai, R.H. Chan, and T.Y. Zeng, A Two-stage Image Segmentation Method using a Convex Variant of the Mumford-Shah Model and Thresholding, SIAM J. Imag. Sci., 6 (2013), 368-390.
X.H. Cai, R.H. Chan, S. Morigi, and F. Sgallari, Vessel Segmentation in Medical Imaging Using a Tight-Frame Based Algorithm, SIAM J. Imag. Sci., 6 (2013), 464-486.
R.H. Chan, H.F. Yang, and T.Y. Zeng, A Two-stage Image Segmentation Method for Blurry Images with Poisson or multiplicative Gamma Noise, SIAM J. Imag. Sci., 7 (2014), 98-127.
X.H. Cai, R.H. Chan, M. Nikolova, and T.Y. Zeng, A Three-stage Approach for Segmenting Degraded Color Images: Smoothing, Lifting and Thresholding (SLaT), Journal of Scientific Computing, 72 (2017), 1313-1332.
R.H. Chan, A. Lanza, S. Morigi, and F. Sgallari, Convex Non-convex Image Segmentation, Numer. Math., 138 (2018), 635-680.

Meeting Schedule:

September 3: Organization Meeting and A brief introduction to Segmentation by Raymond Chan
September 10, 24, 26, and Oct 3 (if necessary): Presentation of background materials by students
Week of October 8: First Group Meeting (1 hour for each group)
Week of October 22: Second Group Meeting (1 hour for each group)
October 29: Mid-term Presentation (20 minutes for each group)
Week of November 12: Third Group Meeting (1 hour for each group)
Week of November 26: Fourth Group Meeting (1 hour for each group)
December 3 (1630-1830 @ C3, LSB): Final Presentation (30 minutes for each group)
December 15: Deadline for Final Report

Project Milestones:

September 3-October 3:

Presentation of background materials
Introductory power-point file on segmentation

First Meeting (Week of October 8):

Present the 2-stage segmentation paper:
X.H. Cai, R.H. Chan, and T.Y. Zeng, A Two-stage Image Segmentation Method using a Convex Variant of the Mumford-Shah Model and Thresholding, SIAM J. Imag. Sci., 6 (2013), 368-390.

Second Meeting (Week of October 22):

Present numerical testing results of examples in the 2-stage segmentation paper.
Code is here.

Mid-term Presentation (October 29):

Each team will have 20 minutes to present what have been done in the last two meetings.
A short introduction to hyperspectral image classification.

Third Meeting (Week of November 5):

Present the hyperspectral image classification paper:
R.H. Chan, K.K. Kan, M, and R.J. Plemmons, A two-stage method for spectral-spatial classification of hyperspectral images.

Fourth Meeting (Week of November 19):

Present numerical testing results of examples in the 2-stage HSI classification paper.
Link of code will be sent separately.

Final Presentation (December 3):

Each team will have 30 minutes to present what have been done in the course.

Final Report (December 17):

Within two weeks after the presentation, please hand in a 10-page project summary.
It should contain the aim of your project, the method used, the results obtained by you,
suggestions for improvement, and the bibliography.

Assessment Scheme:

Presentation of background materials: 20%
Two presentations of group project: 40%
Performance in the biweekly meetings: 20%
One 10-page report related to the presentations: 20%

Important Remarks:

CUHK places very high importance on honesty in academic work submitted by students,
and adopts a policy of zero tolerance on cheating and plagiarism. Any related offence
will lead to disciplinary action including termination of studies at the University.
For information on categories of offenses and types of penalties, you should consult this link.
If you are found cheating (in the programming assignment or the homework assignments),
you will automatically get an F grade in this course and your act will be reported
to the Department for necessary disciplinary actions.
To avoid copying of programs, your programs may be spot-checked, i.e. you will
be asked questions regarding the statements in your program.