e.g. Throw away every other row and column to create a 1/2 size image
reduce size of the image by sacrificing some detail (small scale structures)
Remove noise first by spatial averaging (blurring or smoothing), so that we require less pixels to represent the image, i.e. same information in fewer pixels
Spatial Averaging
a.k.a. Local Averaging
The idea is that the neighbourhood of a pixel is similar to the pixel
replace centre pixel value by average of the neighbourhood pixel value including the centre pixel
The higher the neighbourhood, more the noise goes down BUT the picture becomes more blurry/loss of sharpness/loss of spatial information/ loss of small scale structures
this happens because the principle spatial similarity does not hold true for edges