We will define affine functions more carefully in a future lecture. In short though, affine functions have the following form: $f(x_1,\cdots,x_n) = A_1 x_1 + \cdots A_n x_n + b$. In the one dimensional case, this reduces to $f(x) = a x + b$ (a line).
In this slide, we simply refer to affine changes as multiplying all pixels by a constant factor, or adding a constant value to all pixels.
A non-affine change is basically anything else. It might include applying this operation to only a subset (i.e., not all) of pixels, or squaring all pixel values (a non-affine transformation). The Harris corner response is not necessarily invariant in these cases.
What is a non-affine intensity change?
We will define affine functions more carefully in a future lecture. In short though, affine functions have the following form: $f(x_1,\cdots,x_n) = A_1 x_1 + \cdots A_n x_n + b$. In the one dimensional case, this reduces to $f(x) = a x + b$ (a line).
In this slide, we simply refer to affine changes as multiplying all pixels by a constant factor, or adding a constant value to all pixels.
A non-affine change is basically anything else. It might include applying this operation to only a subset (i.e., not all) of pixels, or squaring all pixel values (a non-affine transformation). The Harris corner response is not necessarily invariant in these cases.