Advanced opportunities to remove real motion blur

[Rasenmaeher]

Hey Vladimir Yuzhikov!

First off all: Your tool is really amazing! From a photographers point of view it is already stunning what this tool can do to bad images.

However when it comes to removing motion blur, the real motion blur is many times very complex. Imagine shaking the camera to the left and afterwards up. To remove this blur, a kernel looking like a "L" would be required. Thus the user is not able to represent the kernel only using the sliders.

To solve this problem, I suggest creating an option to load a premade kernel. This kernel can be obtained by the user by using a image editing program (e.g. by extracting a bright spot on a dark background).

I'd be great if you could think about adding this new feature, because it would make your software so much more powerfull.

Greetings from Germany,

Rasenmaeher

[Y-Vladimir]

Thanks!

Actually, I am planning to include blind deconvolution (with complex path
of blur supporting) in the next release.
19.11.2012 16:11 ÐÏÌØÚÏ×ÁÔÅÌØ "Rasenmaeher" notifications@github.com
ÎÁÐÉÓÁÌ:

Hey Vladimir Yuzhikov!

First off all: Your tool is really amazing! From a photographers point of
view it is already stunning what this tool can do to bad images.

However when it comes to removing motion blur, the real motion blur is
many times very complex. Imagine shaking the camera to the left and
afterwards up. To remove this blur, a kernel looking like a "L" would be
required. Thus the user is not able to represent the kernel only using the
sliders.

To solve this problem, I suggest creating an option to load a premade
kernel. This kernel can be obtained by the user by using a image editing
program (e.g. by extracting a bright spot on a dark background).

I'd be great if you could think about adding this new feature, because it
would make your software so much more powerfull.

Greetings from Germany,

Rasenmaeher

Reply to this email directly or view it on GitHubhttps://github.com//issues/13.

[Enchiridion]

Thanks for the great program! Do you know when the next version might be released (with the blind deconvolution)?

[brianberneker]

I've seen tools like Neat Video (which is used for noise removal) employ a temporal parameter that uses up to 5 frames forward and backwards to not only anticipate the direction of movement (presumably via point cloud) but also to use for additional image reconstruction source.

Have you ever thought of making a video version of this application that employs such techniques?