Dong-Hui Xu
DMP Experience 2004

About Me · My Mentor · Project · Final Report · Pictures

Research Journal
week 1, 2, 3, 4, 5&6, 7, 8, 9, 10,

Week 1:

I arrived at the University of Washington one week after spring quarter ended. Prior to my arrival, my mentor, Professor Linda Shapiro, had set up a CS Research account for me and had assigned me an office on the third floor of the Paul G. Allen Center. The building houses the Computer Science and Engineering department, and is dedicated to Microsoft co-founder Paul G. Allen.

I met my mentor, Dr. Shapiro, on Monday afternoon. We discussed some of the projects that I might work on. For the rest of the week, I spent most of my time reading and trying to decide the focus of my summer project. I also spent some time walking around the campus, familiarizing myself with the new environment and meeting the people I would work with.

Week 2:

Selecting a project is an important but difficult part of doing research. After much reading, I decide to work on techniques for selecting texture scales and extracting texture features from color images and gray level images at the selected scale. I choose this project because I felt that I would enjoy working on it, and it seems to have a wide variety of applications, particularly in automatic segmentation and classifications using texture features. I also have some experience in medical image processing, so I felt that I would be better prepared to get something done during the 10 weeks, while also exploring something new.

Coffee room:
One afternoon Brianna, the other DMP student, and I discovered a coffee room on the second floor of the CSE department. I have always been fond of tea and coffee. I had previously learned to brew variety of tea: Chinese tea, Japanese tea, Indian tea and Thai tea, but I had not gotten a chance to learn how to make mocha or latte. There for the first time, I learned to brew latte: for one or more shots of espresso, top with a cup of steamed milk and a quarter inch of steamed milk foam, add sugar or vanilla syrup to sweeten it and most importantly, finish with a shot of condensed love. Then you can enjoy the best of coffee moments. Since then, almost every day at noon, Marta, a doctoral student from Spain, Brianna and I would visit the coffee room, make coffee, and sometimes chat with other students from the CSE department.

Week 3:

In the third week, I spent most of the week reading and understanding "Blobworld: Image Segmentation Using Expectation-Maximization and Its Application to Image Querying." I discussed the paper with Yi and learned some basic concepts, such as a Gaussian smooth filter, gradient, Polarity, etc. and their applications in image processing. I also spent some time going through the MatLab code and try to understand the implementation of each procedure.

Week 4:

We have meetings every Tuesday in the Irish Room. We exchange ideas, discuss what we have been working on for the last week, what difficulties we have encountered and what our plan is for the current week. Linda noted down the comments in her notebook and gave us advice. I was impressed that Linda managed multiple projects at ease and never reluctant in acknowledging our progress.

Week 5&6:

I started implementing dynamic texture scale analysis by using Java. I spent most the time writing utility functions for vector, matrix, 3D matrix and complex number and other functions. I also tested each function for bugs.

This weekend Linda invited all of her students and their families to a barbeque party at her house. During the party, I got the chance to meet people from different cultures. Some were from the Middle East, Spain, Mexico and East Asia. I also met Linda's son, Michael, who plays football very well and may be receiving a football scholarship.

Week 7:

By this week, I have finished coding. I combined all the functions and tested a protein crystal image. I calculated three texture descriptors for each pixel at a selected texture scale for this image. Linda suggested that I represent the results of the image in RGB format. The results did not turn out very well. The problem was I used the crystal image in JPEG format, which compresses the image by storing only the factors rather than the pixels, to save space. For the rest of the week I modified the code and processed some protein crystal images in PGM format which improved the results dramatically.

Week 8:

On Wednesday, Linda, Marta, Brianna and I decided to go out for lunch since Brianna is leaving for home this Friday. At noon, we walked to a newly opened Japanese Sushi restaurant on University Way. The restaurant was crowed, but we were able to find comfortable seats at a corner table and enjoyed delicious sushi, salmon teriyaki and chicken teriyaki.

After the lunch we stopped at the fountain near the CSE department to take some pictures. The sun shined brightly and there was a hint of sweet roses in the air, but I felt a bit sad knowing that Brianna was leaving soon.

Week 9:

I have generated three texture features at the pixel level for some crystal images and outdoor-scenery images. Linda asked me to classify these images using clustering technique. I have applied neural network and decision trees in my previous work but did not get chance to apply any clustering technique. I felt very excited that I can learn clustering since it is one of the most widely used data-mining techniques, simple but efficient.

With Yi's help, I applied K-means clustering on the images and present each cluster in different color. I love these cluster images because they look like commentary art work.

Week 10:

Before I left Seattle, I visited the Pike Place Market, an open air market, located at the heart of Seattle. The Pike Place Market has been operating for nearly a century and people come from all over the world to see the fish crew fish and having fun with the customers. They also came to buy some of the best seafood and have it shipped home.

It was Saturday and it was time to say Goodbye. It is sad that I have to leave for Chicago, but I know the good memories will always be with me: from the very first second I arrived at Seattle till the moment we said our farewells and good wishes to each other.