Comments for Prof. Xin Li's Weblog

Comment on What I have learned from “poor-quality” PDF files? by Ravindra

Ravindra — Sun, 06 Sep 2009 15:01:31 +0000

well said Dr.li…

Comment on The Relationship of Mathematics to Image Processing by zuhe

zuhe — Fri, 04 Sep 2009 20:35:42 +0000

Einstein’s relationship with Math is an interesting topic.
Actually, Einstein encountered problem seems “unbeatable” when he tried to develop his general relativity theory until he learned Riemann geometry from his mathematician friend Marcel Grossmann. It’s safe to say that Einstein had been a great mathematician when he was able to propose general relativity theory in a reasonable manner. That’s probably why he once said
“How can it be that mathematics, being after all a product of human thought which is independent of experience, is so admirably appropriate to the objects of reality?”.
Unlike image processing, a lot of theoretical physicist’s theories can not be justified by experiments. That is why the greatest theoretical physicists nowadays, like Hawking and Witten can not be recognized by Noble physics prize while their work are highly respected by Math community.

Comment on The Relationship of Mathematics to Image Processing by Cunjian

Cunjian — Wed, 02 Sep 2009 01:29:53 +0000

If there would be some courses offered like mathematics for image processing or mathematics for pattern recognition, that will greatly bridge the gap between application and theory. And I am glad to see this course could cover many math topics,with the purpose of demystifying the heavily used math equations in image processing and explaining its motivations.

Comment on Tips related to MATLAB-based image processing by Sreekanth Nambu

Sreekanth Nambu — Fri, 31 Jul 2009 19:16:39 +0000

Hi Dr Lee

Thanks for the valuable tips on programming and debugging MATLAB.

Regards
Sreekanth Nambu

Comment on Some Thoughts on Compound Video Coding by Mike

Mike — Sun, 01 Mar 2009 15:33:53 +0000

Just passing by.Btw, your website have great content!

_________________________________
Making Money $150 An Hour

Comment on Challenges in Assessing Subjective Quality of Images by masterxinli

masterxinli — Tue, 27 Jan 2009 16:45:26 +0000

It depends on how we define information. In Shannon’s mind, information is the quantification of uncertainty and his famous entropy formula is based the probability model of a source. Image quality can be abstractly viewed a kind of information but its concrete meaning varies from application to application.

In consumer electronics, image quality refers to the sharpness or freedom from artifacts of an image – at least intuitively that is how HVS tells us. But unfortunately, we still have not understood HVS well enough to come up with a computational model for measuring sharpness or related quantities. In other words, there exist low-level probabilistic models accounting for intensity uncertainty in the literature but high-level ones that can account for something like sharpness remain elusive.

In pattern recognition, image quality bears a different meaning because the objective has changed. An innocent-looking fingerprint image (from the perception perspective) might fail the registration process because not enough minutiae can be found. In other words, image quality in pattern recognition application is tied to the matching algorithm. In formation theoretic approach might be more promising in this scenario because unlike HVS, pattern matching algorithms do admit rigorous mathematical analysis. However, a tantalizing question down this path is: how do we accommodate new matching algorithms? For instance, a fingerprint with large deformable distortion might be called low-quality in a system that cannot tolerate that kind of distortion; but suppose the matcher has improved (e.g., using TPS), do we still call it low-quality or not? It seems to me that it is certainly possible to come up with engineering solutions to image quality assessment assuming a fixed matcher. But I doubt its scientific merit because essentially we are forced back to solve the matching problem first.

Comment on Challenges in Assessing Subjective Quality of Images by Jidnya Shah

Jidnya Shah — Mon, 26 Jan 2009 23:27:50 +0000

How about information theory approach? Quality of an image could be a measure of how much information it carries. Low quality could mean less information (about edges, texture, or other desired features).

Comment on Welcome to the wavelet course by Antonio

Antonio — Mon, 08 Dec 2008 19:31:59 +0000

I need a deep course on wavelet concepts and biological applications.

Comment on Career Advices to MS Students by Deepthi Garigipati

Deepthi Garigipati — Tue, 25 Nov 2008 03:38:53 +0000

Dr. Li,
The blog written by you was very helpful in reminding us many things which we often overlook. This semester, as you have said, I came to know the areas in which am weak and I would definitely brush up my skills.

Thanks.

Comment on Career Advices to MS Students by Raghunandan

Raghunandan — Mon, 24 Nov 2008 18:35:08 +0000

Dr.Li

Sometimes, may be most of the times, we overlook many small issues like the ones you mentioned and we need to be reminded at constant intervals about improving social skills, broadening the horizons of our thinking and most importantly knowing about ourselves.

At this juncture, I would like to stress the importace of being motivated for longer times. At this age, no one but ourselves can motivate oursleves – SELF-MOTIVATION. Motivation is a leaking bucket and it needs to be constantly refilled.