Detect blur, sharpness of the single image

asked 2016-02-21 04:43:16 -0600

vlad.gorbenko
1 ●1 ●1 ●1

updated 2016-02-21 08:45:28 -0600

6772 ●3 ●48 ●79 https://github.com/stu...

I am trying to create validation of the image into my mobile application for different platforms(iOS, Android, etc.) Googled a lot I found some solutions and snippets but unfortunately they are not opaque for me. The idea is take a picture and validate it. For example attached image is not valid.

Is it possible to make this kind of validation using OpenCV? I want to have functions like isBlurry(image), isSharped(image)

sample blurry image ( 0,99 MB ! )

edit retag flag offensive close merge delete

Comments

maybe you will support the idea of adding calcBlurriness function to imgproc

sturkmen ( 2016-02-21 07:52:09 -0600 )edit

With one image I don't know if an image is blur or not. With two images you can compare values of bluriness but you need to define a region of interest. It's like when you take a picture do you want to see a blur backgound or a blur foreground?

LBerger ( 2016-02-21 09:07:58 -0600 )edit

add a comment

1

answered 2016-02-21 19:27:55 -0600

Tetragramm

7376 ●13 ●37

Take a look at this SO answer. http://stackoverflow.com/a/7768918/56...

These aren't absolute measures, but they are relative. So you can use these to tell if one image is blurrier than another.

edit flag offensive delete link

Comments

Thank you. But the question is what to do when you have only one image without comparing.

vlad.gorbenko ( 2016-02-21 23:26:42 -0600 )edit

If you find an answer after you can post an answer to this comment "What if you take a perfectly on-focus picture of a blurred image?" When you measure something all is relative

LBerger ( 2016-02-22 03:30:47 -0600 )edit

It can actually be done if you have a perfectly sharp edge aligned with either the horizontal or vertical line of pixels. But other than a very few lab condition papers I've never seen it used.

Basically, take a perfect picture, measure the metric immediately surrounding the knife edge, set that to 1.0, and find the knife edge in subsequent images and compare.