Hi,
I would need to match some images between different categories by complexity i.e. in terms of colour patterns and shapes. Is there any step-by-step procedure to follow?
I probably need to use a program which measure the number of unique colours within each image. Does anybody know whether there is any paper/tutorial which explain it?
Thanks!
This is an interesting topic. It would be nice to try to employ various entropy measures to make categories and subsequently use machine learning techniques to automatically distinguish among those categories. When you are interested, introduction into given techniques is part of my paper on the prediction of Torsades de Pointes arrhythmias.
Combination of entropy measures and ML has a relatively big future.
On the top of it, it can be fully automatized.
Hi Denise,
it was nice to see a question coming from Loughborough. Many, many years ago I worked in a project with people from HUSAT, which belonged to your university, but I think that institute doesn't exist anymore.
I hope you have found a solution for your problem by now. If not, or for others interested, I would suggest to look into Leeuwenberg's structural information theory. He developed a systematic approach to counting the information units in all sorts of patterns, geometrical, musical, movement etc.
The problem may be, as you indicated, that you first need countable things. For example, if you wanted to determine the information in a piece of music that consists of continuously floating pitches rather than of clearly different notes on a scale then you must first find of way of converting the floating pitches into a sequence of discrete ones, e.g. into steps of just notatable differences. In your case, continuous colour changes from lighting or random changes from textures might be a hindrance. There are algorithms for removing such unwanted variations, but unfortunately I cannot suggest a concrete one.
- Badges
- Science topic