Convolution
Updated: 07/31/2022 by Computer Hope
![Convolution matrix math](/jargon/m/matrix-math.png)
In image processing, convolution is a method of modifying an image using a matrix (or kernel) to create new image data. Sharpening, blurring, edge detection, and embossing can all be done using a convolution matrix.
How does it work?
For each pixel, the values of the kernel are multiplied by pixels in the kernel's overlaid area, and then are summed. To normalize the resulting brightness/darkness, the result may be divided by the number of kernel elements. Adjustments can then be applied, resulting in a new pixel. How the kernel is applied when the overlay is out of the image's bounds must be accounted for as there's no image data present.