Deblurring images for OSINT — Part 2

Somdev Sangwan
3 min readJul 12, 2019


Let’s reverse Pixelize blur.

Note: GIMP is used as the photo editing tool in this tutorial.
There’s no way to retrieve 100% of the details from an image once it’s distorted.

In the previous article, we talked about how mathematical operations are used to apply filters to images and how we can deblur images that are blurred using Gaussian blur.

In this article, we will talk about another popular blurring method known as Pixelize. Images blurred using this method can be identified by visible “boxes”. Take a look at the following image for example


The Pixelize blur works by dividing an image into a number of squares and then replacing all the pixels in each square by average of their values.
For example, if a square has 7 black pixels and 7 white pixels, all the pixels in that square will become gray.

Let’s start the restoration of the example image by applying Gaussian blur.

Go to Filters>Blur>Gaussian Blur and keep increasing the blur size (radius) until the boxes disappear and click on Apply.

Name: B__E _A__ON

It looks much better already! Now duplicate the layer by pressing Shift+Ctril+D.

Now, go to Filters>Edge detect>Image Gradient and click on Apply. The canvas will end up looking like this

Now, go to Colors>Brightness-Contrast and gradually increase the contrast until you get refined edges


In the bottom-right of your GIMP window, find the following tab and click on the \/ icon and choose Soft Light from the list of options.

Blend mode

Doing so will add contrast to the edges and make the colors pop. The effect of this operation can be adjusted using the opacity slider

Adjusting opacity

Adjust it to a degree of your choice and you will have an slightly enhanced image like this


That’s all we can do. It is possible to make the edges slightly more sharper by using the brush tool manually but this tutorial is intended for people that do not have experience in photo editing.

Anyways, here’s the comparison between the blurred image and fixed image

____ ____ vs B__E LARSON

The face in result image is now recognizable as demonstrated by Yandex

I hope you will find this method of blurring helpful. Thanks for reading.

Have a nice day, stay hydrated.



Somdev Sangwan

I make stuff, I break stuff and I make stuff that breaks stuff.