NEAREST NEIGHBOUR: This method takes the single input pixel nearest to the transformed point as the resampled output pixel. Advantages are: very low computational cost, and the fact that the input pixel grey levels are not altered by averaging (this is important if you are registering theme or classified data). Disadvantages are: the output image may be jagged and blocky in appearance if there is much rotation or scaling.
BILINEAR INTERPOLATION: This method takes a weighted average of the four input pixels around the transformed point. Advantages are: low computational cost and relatively smooth output images. Disadvantages are: the output image may appear slightly blurred.
CUBIC CONVOLUTION: This is the 4-point classic method. This variation of cubic convolution uses a 4-by-4 window of input pixels. This method is closer to the perfect sin(x)/x resampler than nearest neighbour or bilinear interpolation. Advantages are: output image which is smooth and sharp. Disadvantages are: high computational cost.
SIN(X)/X: REGPRO the sin(x)/x resampler using either an 8 x 8 or 16 x 16 window. Advantages are: output image which is smooth and sharp. Disadvantages are: very high computational cost. Using an 8 x 8 window should be about two to four times slower than CUBIC CONVOLUTION. Using a 16 x 16 window should be about four to sixteen times slower than CUBIC CONVOLUTION.