Usually affine transform matrix (in 2D) is represented like
where block A is responsible for linear transformation (no translation) and block B is responsible for translation.
Block D is always zero and block C is always one.
What if I put some values into blocks D and C I will affect only third (bottom) component of 2D vector, which should be always 1 and usually plays no role.
But can it? Are there some generalizations, where third component plays some role and consequently, bottom row of transform matrix also does?
And one more question: does this decomposition is actual only for 2D? Is it the same for 2D and xD?
Engineers (especially in the field of image processing and computer vision) used to represent an affine transformation in the stated form so that one can perform an affine transform using a single matrix-vector multiplication: $$ \begin{pmatrix}x'\\y'\\1\end{pmatrix} =\begin{pmatrix}a_{11}&a_{12}&t_1\\a_{21}&a_{22}&t_2\\0&0&1\end{pmatrix} \begin{pmatrix}x\\y\\1\end{pmatrix}. $$ Another merit of this representation is that you can compose two affine transformations using simply by multiplying their matrices together.
This representation obviously generalizes to higher dimensions as well, and it is a widely adopted convention to represent affine transformations in 2 or 3 dimensions in this way. Whether such a convention is good or not is a matter of taste, but engineers and computer scientists seem to think that this is convenient.