Printable View
how can i compare two images in java? its not like 100% pixel match but something like detecting a person. I take a persons photo and compare it with the database photo to check for a certain threshold. how it can be done? any code or any idea/suggestions?
If image is not possible/too hard i can also use signature but i think it will make things worse.
please help me.
i think i will need JAI packages...
Uhm... Yeah..
I took a couple of course at the university to get an idea on how to do this. It is very very complicated to do even for a basic set of no more than 20-30 photos in perfect condition. The courses were, amongst others, Artificial Intelligence, Neural Networks, Advanced Image Transforms and math-classes out of my a*se. They were all great fun and I recommend them.
The point I'm trying to make here is that this is a very complex task and you are unlikely to find a generic API in java of the kind:
boolean Image.kind_of_looks_like(Image image)
lolz...got ur point...i dont need a perfect algorithm...i need something i can start to work with...i can start from the beginning but i dont have time...:S...
n e ways thanx for ur suggestion...i have a working algorithm that has an accuracy of 50% (:D) so i have to work hard i think.
You should study about how to read and write Java image. Then try to implement your own way. If you can read the image, you should able to find a pattern to compare images.
I know how to read and write java image, can u provide me any algorithm that does the comparing effeciently??
If you can read an image, you can find the header details right?
What is the expected results of comparing two images?Quote:
compare two images
For example:
image 1 is the same size as image 2
image 1 is a subsection of image 2 (ie it is contained with it)
image 1 uses same amount of color red as image 2
The location of the dark intensity colors in the two images has an agreement of > xx%
how to compare two images that to say its equal or not?
As we discuss previously, we can do this is several ways. Each image has a header, and contain lots of details. Some of them are specific, and some of them are common to each same type.Quote:
Originally Posted by nithiyanandam
What's the way you want to compare those images?
This is one of the so-called "AI-Complete" problems - i.e. it is incredibly difficult. I suggest you look up some recent academic papers on computer vision and face recognition. You'll probably be dealing with wavelets, DCTs, eigenfaces, inference-based choice engines, etc.
If you do manage to create an efficient working solution, I'm sure a number of international governments would be interested in talking to you.
I agreed with your. But in lots of applications, I seen that reading the header and comparing some basic details compared the images. Actually those are not advance level application. And I don't think once you edit an image quite similar to the other one, headers are given different details though. Like the color depth and so on. So I think for low level applications it's not a bad selection.
This question is about artificial intelligence, not Java. If you need to know how to load/write images, manipulate pixles, display images, etc...this is the place. But it isn't the place for "how do you create a truly intelligent machine" type questions.Quote:
Originally Posted by GhosT
Seriously, the question is too vague to be answered correctly at here or at any site..
Hi all,
I think he needs something like OpenCV (wrapped with Swig) or AForge (but it is for .Net) using Image Similarity and Computer Vision
I'm interested in this argument too
could anyone help us?
Thanks in advance
Revenge