0

If you take a picture of - for example - a car and you get the license plate number in it; how do you blur out the numbers?

(I would also like to know how to do this with a video if possible).

Michael C
  • 175,039
  • 10
  • 209
  • 561
L.B.
  • 121
  • 4
  • You could just draw over it with a color, or clone-stamp something nearby onto it, or swirl it around with a blur tool... – Ryan Nov 26 '14 at 07:40
  • 1
    This seems completely trivially easy for still photographs. We accept a lot of basic questions as on topic here, but... What have you tried, and why didn't it work? – mattdm Nov 26 '14 at 09:17
  • On the other hand, it's suddenly a lot more complicated with video, where an object might move across many frames. That suggests this should be migrated to the video site. – mattdm Nov 26 '14 at 09:18
  • @mattdm I actually hadn't tried anything because I wasn't sure where to start! :) Also, I was not aware of the fact that there was a video site... I'll look into that :) – L.B. Nov 26 '14 at 15:50
  • 1
    This question appears to be off-topic because it is about a photoshop operation, so best for that group – chuqui Nov 26 '14 at 17:45
  • @chuqui There's a Photoshop group? – Saaru Lindestøkke Nov 26 '14 at 18:38
  • http://graphicdesign.stackexchange.com/questions/tagged/adobe-photoshop – chuqui Nov 26 '14 at 21:08
  • @chuqui note that this is not graphic design, and AFAIK this is highly relevant to photo manipulation – clabacchio Nov 27 '14 at 07:49
  • you can youse online tools such as https://blur-photo.com/ for that – Lukáš Irsák Apr 17 '22 at 20:18

1 Answers1

4

What I usually do (in GIMP), but is maybe overkill, is to duplicate the layer, apply a gaussian blur and then use transparency masks to limit the blur to the interested area. You can use the same trick with any other filter, like the mosaic; or you can also get creative and use some other effect, like white noise.

You can achieve more or less the same effect by using selection tools like the lazo.

Blur:

enter image description here

Image taken from Wikimedia.

Pixelation

enter image description here

I think that in Photoshop you can directly create a filter layer.

Of course, you can also paint it with the background color of the plate, if it's uniform enough it will look realistic:

enter image description here

Slightly more complicated, but can be funny: replace the car plate with a text of your choice, paying attention to colors, perspective, sharpness and so on.

Super-quick option: with the smudge tool of GIMP, you can make a couple of passes and get this:

enter image description here

Community
  • 1
clabacchio
  • 2,219
  • 3
  • 17
  • 31
  • That would be a riot :) Thanks for the info. I will probably mark your answer as the answer but I want to wait and see if anyone adds anything else :) – L.B. Nov 26 '14 at 15:52
  • 2
    Far fetched but still: In this answer it is mentioned that when the blurring function is known the blurred image can be almost fully restored. Isn't this the case when using a Guassian blur? I think that function is only dependent on the radius, for which different values can be tried in the reversal algorithm until a satisfactory result is found. Or is this much harder to do than I imagine (impossible?)? – Saaru Lindestøkke Nov 26 '14 at 16:21
  • @BartArondson you need to know what pixels were mapped there, which is not much different from giving you the original image. You can still probably reconstruct the first, but surely not the second and third I believe. I'm not even able to recover the firs one, by the way – clabacchio Nov 26 '14 at 16:24
  • 3
    The Gaussian blur can definitely be recovered. It won't be pretty but to obscure license plates this way is not safe. Here's an example where something similar was done. With license plates it is even easier. – Unapiedra Nov 26 '14 at 17:37
  • @Unapiedra I'd say that it really depends on the amount, but indeed it's a pretty simple filter from a mathematical point of view – clabacchio Nov 26 '14 at 18:25
  • 1
    @clabacchio, it really doesn't depend on the amount, the math is straight forward. However, discretesation will cause problems, and possibly more so if the blur is large. In practise, I've just tried to reconstruct the first image and couldn't do so in GIMP in 5 minutes. – Unapiedra Nov 26 '14 at 18:37
  • @Unapiedra I also tried with the unsharp mask, I guess that by deconvolution and trying to guessing the gaussian parameters it would be possible to do something, but honestly I'm not sure that it's always possible – clabacchio Nov 26 '14 at 18:44
  • 1
    You can find the point spread function by considering what you think should be a sharp high contrast boundary which is now blurred. You then zoom in into the boundary to take one small straight part and then it's not all that difficult to calculate the point spread function (here we assume that this is a rotationally symmetric function that only depends on the radius). This point spread function can then be used to deconvolve the image, obviously details smaller than one pixel cannot be restored. What is also important is that after the blurring no nonlinear mappings were applied. – Count Iblis Nov 26 '14 at 23:38