A shift in object detection: http://pdollar.wordpress.com/2013/12/

I am looking for the PhD topic in Computer Vision and Image processing. However, this field is huge, so I would like to ask for help.

Hi Adrian,

Thank you very much for your advice!

Actually, I have taken some courses related to image processing ( 2D image) at University. After these cosurses, I decided to follow this field, and my thesis was FPGA based 2D Discrete Wavelet Transform.

Currently, I want to join the real time image processing group. The requirement is knowledge of Computer vision and Image processing. I dont know exactly what I will investigate right now. However, I intend to use the combination of software and hardware (FPGA) in my fututer research.

Best regards,

Tuan Huynh

I think that many interesting tasks can be found in the field of computer vision and augmented reality in mobile devices (smartphones and tablet PC's).

I'd generally agree to Adrian: The ability to implement your idea will heavily depend on your research group. The available equipment, know-how and experience is an important foundation for a future thesis or PhD topic.

That said, you could of course have a look at what other universities are looking for. Most institutes have a website listing open topics for their respective fields, as for example with my institute in the attached link (which has some open topics for robot vision, a related field). As you have probably already have some research experience in the field, universities that publish a lot of well researched papers, for example in the journal posted by Aparna, are probably a good place to start. Note though that these are often topics topics that have not yet been covered for a reason (if they are not VERY new). Many more exist but never get published due to being handed out internally. A more reliable method is to contact the respective institutes directly, as many of them are looking for future research assistants and PhD students.

http://www.lsr.ei.tum.de/teaching/theses-announcements/

Trends that I observed recently:

1. Avoid binary segmentation before classification.

2. Use general low-level features instead of hand-crafted features.

Excuse me Tuan to post another query in yours.. What if there are no groups... Is this field still suitable for individual research?

In answer to the question for this thread, you may find the following book interesting:

J.F. Peters, Topology of Digital Images. Visual Pattern Discovery in Proximity Spaces

now in press at Springer (see the attached cover for this book).

I'm working on my PhD as well, and my topic is related to object recognition and image annotation. What I have found more fascinating in this field is to look for paths that less people are looking for. Let me explain. Nowadays, there is a trend to image analysis based on regular image decomposition (or segmentation), for instance, the largely popular Bag of Features, for object recognition, deals with visual words, that are usually extracted from a grid decomposition of the image. Many approaches are based on regular windows analysis, and, although they have proven to improve current results, I think (and many other people do as well) they will get stuck eventually. We, as humans, do not perceive the world as regular patches. On the other hand, when you deal with irregular image segmentation, most popular approaches use no high-level knowledge. That is, the segmentation of the image is based on some kind of mathematical equivalence among low-level features (pixels in the image). But, how can you actually segment an image if you don't know in advanced what are you looking for? For me, and what I'm trying to do is not just to repeat and enhance a little bit what most people are doing, but to find other ways of simulating how humans perceive the world.

If you like this kind of approach, I recommend the book "Object Categorization: Computer and Human Vision Perspectives", edited by Sven J. Dickinson, 2009.

A shift in object detection: http://pdollar.wordpress.com/2013/12/

The current trend in computer vision and image processing is focused on many tasks. Bridging the Semantic Gap in Image Retrieval, Video analysis in mobile devices, Image and Video Forensics, Deep-learning and feature learning are on the rise within the Computer Vision community, Mid-level patch discovery is a hot research topic, RGB-D input data is also a current trend in image processing.

Current and future trend, a interesting point of view in this article: http://savvash.blogspot.it/2013/07/cvpr-2013-three-trending-computer.html

The main trend that I notice is the introduction of deep learning and sparse coding to large data analysis in image/vision.

I agree with Anette to look for paths that researcher do not go through but of course with care. Finding a thesis topic is a multiple objective optimization problem indeed.

Objectives can be your own interest at first because you have to stay long time with your ideas. Fitting in your supervisor and group research interest and considering availability with equipment. It is the intelligent interaction of all these parameters that should make you decide on the topic. My personal advise is that you should visualize the outcome of your thesis at first. Dont enforce false objectives such as the necessity of doing hardware implementation such as FPGA. My supervisor in Japan once asked me to do the same. After a while I realized that the real outcome should be a journal publication in high impact journal. The industry can do magic about hardware if you find the right algorithm for them, so your chances of doing real genuine work in hardware is limited and so is publication in hardware ideas. A journal publication is easier if you studied well the literature and come up with new idea or algorithm, people will appreciate the idea more than the hardware implementation and your idea will be more appreciated if you make it look like a natural consequence of the literature. My second advise is really to work with low level vision and how primitives can be used for better measurements, few people do this and best luck for you :)

Dear All, i have just started pursuing my PhD Research. And i am also facing the same dilemma of selecting research topic. i am not working in any research group.

can you indly suggest me some topics.