Evaluate the edge quality of IplImages

How can开发者_运维技巧 I evaluate two 'edge images' (generates from CVCanny) in Opencv..? As a example I have 2 Iplimages. First one is clear but second blur little bit. When I convert these images to edges using CVCanny function in Opencv , Is it possible to identify best image? If I get edge population can I filter the best one..? If yes how can I get edge Population..? Please help me.

The usual way to measure blurriness is to evaluate the loss of high frequencies, via the variation of the edges spreading. This is a totally different task if you have a reference image or not. There are not a lot of paper on this topic, especially without reference. It doesn't seem to be your case, but I'm not sure sources are exactly the same in your setup.

If you have a reference of some sort, you may want to look at these papers:

• http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.27.7536
• http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=1689062
• http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.107.8712

which are usually more accurate than a simple standard deviation over the whole image, as proposed on the "Get most accurate image using OpenCV" topic. Be careful to compare things that are truly comparable in terms of frequency content!

If you don't have the reference image, have a look at Marziliano's dedicated paper: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.7.9921

This solution is not perfect, as none are, but may suffice to your needs.