alt.hn

3/28/2025 at 6:07:39 AM

Estimating Camera Motion from a Single Motion-Blurred Image

 https://jerredchen.github.io/image-as-imu/

by smusamashah

3/28/2025 at 11:13:47 AM

Visual effects, VFX, has been doing this type of work for decades. The guy that I know doing this the best, Eugene Vendrovsky ended up at Nvidia as a robotics computer vision division director. Back in the early 2000's he had all kinds of camera motion recovery algorithms we used in the Tracking department for use on feature films like the live action Scooby Doo, Riddick and the Narnia films. Eugene retired recently, after a very productive career.

by bsenftner

3/28/2025 at 1:41:21 PM

Going to read this in depth. I have been on and off trying to find a way to distinguish between in focus, motion blur (camera shake) and out of focus. Motion blur is surprisingly tricky to distinguish using a computer from the others. I can glance at a 2D Fft plot and tell, but the computer - not yet.

by porker

3/28/2025 at 2:47:41 PM

Blind Deconvolution is around for a while and from the estimated PSF you can gather the shake pattern (or tell whether it’s just overall out of focus). It just never worked good enough in practice but I remember some impressive results already a decade ago

by fp64

3/28/2025 at 3:10:59 PM

Thank you, blind deconvolution and PSF sound like what I have been scratching around the edges of in my experiments, without knowing the right terms to search for to discover prior work. I shall dig into the literature!

by porker

3/28/2025 at 10:58:10 AM

Can't upvote this enough. Finally a piece that's not about LLMs or how are they going to ruin the world.

by anovikov

3/28/2025 at 12:17:42 PM

And not about Rust, either!

by fortran77

3/28/2025 at 1:43:29 PM

And no cod psychology about how to live a better life, either!

by damnitbuilds

3/28/2025 at 2:03:46 PM

or worse, make the world a better place

by dylan604

3/28/2025 at 7:46:44 AM

Three days after posting- gets inverted and reused as a directional camera motion blur shader ever after

by InDubioProRubio

3/28/2025 at 8:09:50 AM

Isn't a logical next step to extract the depth field? Possible?

by drsopp

3/28/2025 at 8:17:33 AM

Isn't the depth decoder part of the processing already?

by tetris11

3/28/2025 at 12:08:23 PM

I should have read the abstract.

by drsopp

3/28/2025 at 8:45:16 AM

Right there in the abstract:

"Our approach works by predicting a dense motion flow field and a monocular depth map directly from a single motion-blurred image"

by skywal_l

3/28/2025 at 8:46:14 AM

I'm not familiar with all of this but is there then a tool to remove the blur ?

by wiz21c

3/28/2025 at 8:50:20 AM

That’s for the next grad student to solve.

by xattt

3/28/2025 at 2:54:38 PM

Yes, those have existed for decades, with various degrees of success.

by esafak

3/28/2025 at 9:04:19 AM

Can this be used in LLM's like gemini which don't really have a motion idea but rather they just take things frame by frame and so can't really understand motion b/w them

by Imustaskforhelp