Research Orientation of Image Processing

Hello!

Although I am engaged in image processing research, what I do is not very thorough. I can only give you some rough suggestions.

I feel that there are two levels of development in image processing:AlgorithmThe study requires a lot of mathematical foundations, such as partial differential equations (PDE) and various spatial transformations (wavelet, curved waves, shear waves, etc ). There are a lot of research literature in these fields, but it is really difficult to come up with something new. If you can study it in depth, the written papers have a certain theoretical depth, and it is easy to be hired.

The other layer is to expand horizontally, find new applications, and combine other technologies. The key is to find the research content. The algorithm only applies the existing theories. Most common pattern recognition algorithms are combined, such as face detection, pedestrian detection, and motion detection in videos. Only new ideas,ArticleIt should also be easier to hire.

Instructor Wang:

Hello, thank you for taking the time to read my emails during your busy work and study. I am a cloud

A teacher from a local school in South China recently studied digital image processing.

I know basic concepts such as image enhancement, restoration, compression, encoding, and segmentation, but I think there are a lot of details.

I don't really understand it. I don't know what I need to learn to study image processing. Read the online

Some discussions have said that the theoretical research on image processing is difficult to develop new things.

You can see which aspect of image processing is easier to post. I want to choose my master's degree recently

I don't know where to start. I hope I can get some suggestions from Teacher Wang. I am not very grateful here!

I saw instructor Wang's article and email address in "China NLP graphics news", so I took the liberty

I have sent you an email. Sorry for the exception.

 

Image processing is an old and difficult problem. Image processing has been studied for many years and many problems have not been solved. A simple example, such as de-noise, has not been fully studied till now. However, it is difficult to propose a better solution. There is also a natural image model problem, for example, at least till now there is no good model that can describe a large number of natural images. Some models also apply to some situations. Therefore, image processing still has to be done, but it is not that easy to do. Just like the reply from the above teacher, it requires a deep theoretical foundation, especially in mathematics, signal Processing and Statistics   There are a lot of research contents and researchers. No matter what the study is, it should be possible to put forward your own opinions or make improvements on the basis of others.   Image resolution enhancement. For example, a small image must be magnified several times to ensure the amplification effect. Specific applications, such as facial Resolution enhancement, are applied to environments such as surveillance video face detection. This is what a teacher gave me. I checked the documents about resolution enhancement, but there are still few Chinese characters in this regard. I have some bugs in this aspect to talk about.   Image resolution enhancement. For example, a small image must be magnified several times to ensure the amplification effect. Specific applications, such as facial Resolution enhancement, are applied to environments such as surveillance video face detection. This is what a teacher gave me. I checked the documents about resolution enhancement, and there are quite a few Chinese characters in this regard... It seems that there are a lot of enhancements to multi-resolution images.   Innovation is really not easy. I feel that the combination of image processing theory and other disciplines is the best way to achieve this.   I think this teacher's reply is still very pertinent. Basically, the multimedia field (including video and audio) is both of these two ideas. Personal feeling: If you are a student of mathematics and do not feel obvious about the media, you can follow the first idea. After all, most people studying computer science are inferior to you in terms of theory, applying some commonly used mathematical transformations or optimizations may make the methods seem novel; If you are a computer student and have a general mathematical knowledge, you can follow the second idea. As long as you can find an interesting application, the method used in it is not too good, it's a good article. Of course, if you are strong in both aspects, you can do whatever you like. PS: journals of graphics and image may be the best journals in China in terms of image processing. But if you want to do research, pay more attention to well-known international conferences, such as mm, ICIP ...... The content of the article is much newer.   I think this teacher's reply is still very pertinent. Basically, the multimedia field (including video and audio) is both of these two ideas. Personal feeling: If you are studying mathematics and do not feel obvious about the media, you can follow the first idea. After all, most people who are studying computer science have theoretical skills... Thank you for your reply. I cannot leave without it. Poor mathematics and computers. Very depressing. It is too difficult to engage in scientific research. It is not suitable for me. Haha   Except a very small number of students, most of them won't be too prominent in both aspects when they graduated from undergraduate courses. As long as you can still calculate in the same grade, there will be no major problem. If you are lacking in both aspects, but want to do research, we recommend that you first take advantage of the young math. Although it takes a little longer to start posting articles, reading and writing articles later will be much faster, with fewer detours and no mistaken cutting-back engineers. If you don't really want to do research, go to work. In fact, work is very good, at least to make a lot of money, it is not very hard to do it later.   Ah, it's harder for us to make watermarks.   I am also doing algorithms, which is difficult to do. I need to calm down and look at a lot of things .. I have to work hard .. The Chinese Ghost graphics report is very good. But not the EI core.   The Chinese Ghost graphics report is very good. But not the EI core. I think it is amazing to be able to cast it to the ghost graphics news in China, because I am really a newbie. Although I have been studying every day, I still don't understand anything, I tried to write the first paper and made a bottom-level journal. I was hired after I was asked to change the format. As a result, I paid 1700 yuan for the publication. Hey, I am dead. Forget it, I will not pay it. in the future, we can really invest 1700 of the core.   Image resolution enhancement. For example, a small image must be magnified several times to ensure the amplification effect. Specific applications, such as facial Resolution enhancement, are applied to environments such as surveillance video face detection. This is what a teacher gave me. I checked the documents about resolution enhancement, and there are quite a few Chinese characters in this regard... This is called Super Resolution Reconstruction, which is a good research direction.   This is called Super Resolution Reconstruction, which is a good research direction. Thank you very much for your prompt! I used "Resolution enhancement" to search for articles and found very few. After your prompt, I found that there are still a lot of documents in this area.   This is called Super Resolution Reconstruction, which is a good research direction. This research has been going on for many years. I remember that many people were doing this at an academic exchange CONFERENCE IN THE 04 years.   Well, I just learned the digital image processing course, and I am very touched by what you have said.   Image processing is indeed difficult to develop further, but it also shows the need for development. It is not easy to develop its own theories, so most people in China are using it. I think the bottleneck of image processing is that, unlike one-dimensional signal decomposition and reconstruction tools such as FFT and wavelet, images also need to be decomposed and reconstructed, and image processing requires their own tools. However, image processing still uses the one-dimensional signal analysis method, which is similar to nature itself? I hope you can think about images from the perspective of images. Fortunately, some scientists have already done this, such as beamlet and ridgelet.   Image processing is indeed difficult to develop further, but it also shows the need for development. It is not easy to develop its own theories, so most people in China are using it. I think the bottleneck of image processing is that, unlike one-dimensional signal decomposition and reconstruction tools such as FFT and wavelet, images also need to be decomposed and reconstructed, and image processing requires their own tools. ... Haha, it should be a bull   Make sense!   Image processing is indeed difficult to develop further, but it also shows the need for development. It is not easy to develop its own theories, so most people in China are using it. I think the bottleneck of image processing is that, unlike one-dimensional signal decomposition and reconstruction tools such as FFT and wavelet, images also need to be decomposed and reconstructed, and image processing requires their own tools. ...