5. Spatial filtering, edge enhancement and equalization

5.1 Spatial linear image filtering

Filtering an image consists, for each pixel, in replacing the measured value (the gray level supplied by the detector) by a linear combination of the values measured on the surrounding pixels. This operation is the algebraic equivalent of a convolution (a multiplication in Fourier space of spatial frequencies). The surrounding pixels are chosen from a square whose side is an odd number of pixels 2N + 1, the central pixel being the one to which the result of the operation applies. The square is called the template, and the set of (2N + 1) ² pixels it covers is called the kernel. In each box of the template, a number indicates the weight of the underlying pixel in the filtering operation. Several types...