When I talk about image filtering i'm usually referring to something different than Instagram filters, so today i'll share with you some information i gathered from the internet and some books about image processing in general and low level filters and i'll take blur as an example.
to create a blurring effect, the image processing software uses something called Gaussian Blur( Gaussian smoothing ) which is technically using the Gaussian function to change the value of each pixel, this technique results in a smooth blur resembling the one in the picture above.
This algorithm can also be used in computer vision algorithms as a pre-processing stage in order to enhance image structures at different scales.
Mechanics of this method :
let's say you want to apply this algorithm to an image:
instead of using an image i'll use a representation of a picture with each number representing a pixel value
yes i suck when it comes to using paint :D anyhow let's say you want to apply the Gaussian method to the pixel with the value 3
First you need to create a Kernel smaller than the picture usually we use a 3x3 kernel
we put the pixel we want to transform in the middle and after the we calculate the sum of the values of the pixels around it and divide everything by the sum of the number of pixels we used.
So for the first pixel the equation will be x=(3+5+4+9)/4
Repeat the same process for all the pixels; the drawbacks of this algorithms is that it takes time to apply to all the pixels and the bigger the image the longer it takes.
This was a small thing i learned today and i thought i would share it with, thank you for reading and have a nice day :)
- 10:30
- 1 Comments
