Image Processing Reference
The Filter Selective Gaussian Blur
The Selective Gaussian Blur window with the flattened image section in the preview window.
Next to it is the main window frame with the corresponding image.
For this example, we will open noise.png again and apply the Selective Gaussian
Blur filter ( Filters > Blur > Selective Gaussian Blur ). This filter doesn't work on the
entire surface area of the picture or the selection as other blur filters do. It is
applied only to the pixels that deviate in color (however slightly) from the
neighboring pixels by a defined delta value (brightness or color difference).
The effect is that edges with a hard color or brightness contrast remain while
surfaces are blurred—or blended together when pixel colors are similar. The
result is that the image is flattened without losing details and contour.
In this example, I chose a Blur Radius setting of 4 pixels. This value defines
how strongly the image will be blurred and, therefore, how the details are
preserved. The higher you adjust the value, the smoother, with less noise,
your image becomes. Depending on your setting, the disturbing pixel noise
disappears, but so does the sharpness of the image. But keep in mind that
the higher you set the value, the longer it takes for the program to render the
I set the Max. delta value at 150. The lower this value is, the finer the details
remain in the light/dark contrasts, but as a result, more noise contamination
remains in the image.