The graph theory is quite interesting and useful.Some year ago,although I learned some knowledge of graph theory in mathematical modeling course,I just think it can only be used to solve some math problems,like paths optimization and TSP (Travelling Salesman Problem).Now, I realize it can be used in social networking.
In those what we have learned from this course,the case about social multimedia searching,especially, the photos ranking technique stimulates my interest.
I once made some research effort to realize the ranking of photos.Because there are a lot of photos online and it is difficult for us to rank photos by their qualities. For example,we search "dog" by using Google and the web react very immediately and show a lot of photos of dogs.But what should be noticed is that these photos are not shown by their qualities.Some photos are very professional taken and some are bad qualities,like blur and dark.
In order to solve this problem, we can take some factors that affecting the photo qualities into consideration. These factors include some rules in photography,like the Rule of Third, focus on the object and blur the background, color hue and saturation and simplicity. If one photo satisfy some criterion,the ranking system can give it a higher point and show it more previous in webpage.
For example,the Rule of Third used when we take a photo of a tree in river.Shown as below:
In order to solve this problem, we can take some factors that affecting the photo qualities into consideration. These factors include some rules in photography,like the Rule of Third, focus on the object and blur the background, color hue and saturation and simplicity. If one photo satisfy some criterion,the ranking system can give it a higher point and show it more previous in webpage.
For example,the Rule of Third used when we take a photo of a tree in river.Shown as below:
The photo of white flowers above is bad quality.Because it is failed to focus its object and makes the whole photo blur. But the photo of butterfly is quite good, it has clear object and blur background.
This system,now can achieve a accuracy above 80% in simulation by using Matlab. Also this ranking system has some shortcoming,like long delay and some difficulties in categorizing the photos. However after I learned the graph theory, I come out some ideas that this theory may be helpful for photos categorization.
Therefore, in my opinion, photos ranking is a very potential and useful tool for social network multimedia.
This system,now can achieve a accuracy above 80% in simulation by using Matlab. Also this ranking system has some shortcoming,like long delay and some difficulties in categorizing the photos. However after I learned the graph theory, I come out some ideas that this theory may be helpful for photos categorization.
Therefore, in my opinion, photos ranking is a very potential and useful tool for social network multimedia.
It a good idea about the photo ranking.It maybe very helpful in our daily life and I found that the computing about how to figure out whether the photo is good or not may have different views. I recommend that you can show us more rules about how to put a value to one picture, and photoes of different areas may have different features, according to my opinion, is it right?
回覆刪除Yes,you are right!There are many features can be used to evaluate the photos qualities.Some feature as I have mentioned in this blog,like "Rule of Third, focus on the object and blur the background, color hue and saturation and simplicity". We also have features like "clarity","contrast" and some analysis methods in frequency domain.
刪除Another thing you have mentioned is also correct.Different kinds of photos need different features to evaluate.For example, the photos of people and photos of scenery are quite different.So it is important to do some categorization work before ranking them.
Photo ranking is very interesting. But I think the problem is how to handle large numbers of images quickly. And as the increasingly popularity of digital cameras on different devices, more photos will be produced. The ranking process will consume more computing power. Though there are data centers to do such a ranking, I think the solution of distributed processing may be better way for it can use cheaper machines over internet.
回覆刪除yes,one problem of this technique is time consuming.When I processed 2000 photos,it almost costed me 2 hours then gave the result!I think your idea is quite right by using distributed processing.Thanks a lot,because you give me a very good idea!!
刪除I am quite agree with your opinion that photo quality ranking system is a potential tool for social network. And I think the most difficult aspect lies in that how to separate the object region and background region. How can a computer figure out the theme of a photo or picture automatically? It is also a key to make it easier and more efficient to search a picture or photo.
回覆刪除Besides, there is an application which fascinates me deeply. You can imagine that you see a film online and you are interested in the clothes the leading roles are wearing. You can click on the clothes and find out the information about the clothes and then share them with your friends on the social network. It is also an useful application of computer vision and social network. It can help to do some advertisement at the same time.
Photo ranking now is mostly done by human. Your idea is great if it is really feasible. Face recognition is a technology long time ago, but is it possible to identify the object like a dog, tree, diamond ring and so on?
回覆刪除Thanks for sharing the concept of photo ranking technique.
回覆刪除It's very interesting. As you know, it is good for website to filter out the bad photos for us so that we can find the best and interesting photos easily.
Photo ranking is a pretty exciting technology if it can effectively solve coding problem.In fact,I think it can greatly change the future trend of social networking.People can get access to more information in SN,including photo searching,new ways of advertising with the help of photo ranking.And the more abundance of social network will absolutely help to enhance the technology of photo ranking.
回覆刪除This concept is really cool! In the past, only people especially the professionals can judge if a photo is well taken, but machines can't. By using the technology you just said, it could be easier and faster to do that. And it will save much when we are searching pictures on the Internet. I am very interested in it, and I also think if this technology could be used in social network, there will be much fun!
回覆刪除I really agree with your opinion, the picture quality evaluation system is a potential tool for social network. And I think that the most difficult is that how to separate the target area and background of the area. The computer can find out how to theme pictures or pictures to be automatic? It is also a key to make it easier and more efficient search a painting or photos.
回覆刪除In addition, there is an application, I deeply fascinated. You can imagine, you see a movie online and your interest in the main characters clothes to wear. You can click on the clothes and find out the information about the clothes, and then share them with your friends social network. This is also a useful the application of computer vision and social networks. It can help to do some advertising at the same time.
I think Photo ranking will become a very useful tool in social networking in the future. Photo is a very important component in our social network. Nowadays, Blog is not made of word only. Blog is full of words and photos. With the help of Photo ranking, we can find good photo in our social network.
回覆刪除