Could anyone explain to me the difference between these scaling algorithms? i.e. Which ones are better for upscaling or downscaling, which are better for photos and which are better for 2-bit images, and the relative speed of each, etc...
bicubic
bilinear
box
data dependent triangulation
nearest neighbor
Thank you!
I have some large 2-bit images that are a bit pixelated and I want to know which scaling algorithms I can use to de-pixelate them, perhaps by downsampling then upsampling (or vice-versa) using different algorithms.
Bicubic Bicubic is the type of interpolation to be used. It trys to fit a cubic polynomial to your known pixel. This polynomial is then used to calculate the color of unknown pixel.
The cubic polynomial has the advantage of a smooth colorchanges but it is much harder to calculate then all the others.
Bilinear Same goes for bilinear interpolation except that it assumes a linear coor change. The leads to colorchanges not as smooth as bicubic interpolation but it is much easier to calculate.
Box I am not quite sure but i would assume they just use the pixel value of the top left known pixel. This would lead to a very pixelated image.
Nearest neighbour Every unknown pixel gets the color of the nearest known pixel. Should lead to very pixelated images.
This said. Every method has its pros and cons and the result depends very much on the scale of your upsampling.