Automatically detect blurry image areas

Question

I am working on images that partially blur on some sections. These are the noises that should be taken care of, but here is the problem:

Are there any methods for determining if the image is blurry or partially blurry in some parts of the image? For example, take a look at the sample image below:

In the image you can see that there are 3 sections that are visually blurred: bottom left, near the central area and top right. Is it now possible to detect that any part of the image is blurry in programming or mathematically?

Answer 1

As lain_b noted, with this image you can use the edge detector and look for the absence of edges. I tried this on your image and it seems to work very well. I used the kernel first

[0,1,0, 1,-4,1, 0,1,0] 

This is a simple detector. His result was

Then I used the threshold to get

Then I closed the image and opened it to get

This is obviously not a finished version, the upper right side does not recognize at all. Perhaps you could improve it by blurring before executing threshold values ​​or by choosing the best values ​​for the threshold and the radius of the open and close operations. Many of the solutions you need depend on the limitations that you can solve. I think this method will work for you.

Edit If you are looking for blur detection of arbitrary images, you will have to explore a variety of methods. Everything is much simpler if you can make assumptions about your set of input images. Without any assumptions, I do not know what will be best for you. Here are some related readings.

Image blur rates

Restart paper using Harra wavelet transform

Similar SO question and look at the question that refers to

Blur detection is a very active area of ​​research, no answer. You just need to try all the methods that you can find (they were detected when googling was detected in the image).

Answer 2

This document may be helpful. It makes a blur estimate (mainly due to focus, but I think it also blurs) to recreate a similar blurry object in the image.

I think you can use it to detect blurry areas and how blurry they are. This should be especially relevant for your problem, as it is designed to work with images of the real world.