How do I interpret 2D transformation matrix correctly.

I have performed a 2D registration between a reference image 2.fig (512x512 with 0.4102mm/pixel spacing) and an input image 1.fig(512x512 with 0.7960mm/pixel spacing. With known shifts and no rotations The registered image and the figure showing both the registered image and the reference image are shown in 3.fig and 4.fig respectively. From my images I can see that my registration is done correctly and I use inbuilt matlab functions;tformAffine = imregtform(moving,fixed,'similarity',optimizer,metric);andmovingRegisteredRigid = imwarp(moving,tformAffine,'OutputView',imref2d(size(fixed))); I understand the tformAffine is my 2D transformation matrix. The images i registered were AP(anterior posterior images i.e. X,Y plane) my know shifts in the X and Y planes were 0.06cm and 3.50cm. My tformAffine from MATLAB is 1.29311976065651 -0.0593993246222778 0 0.0593993246222777 1.29311976065651 0 -81.5281146404199 23.5130403514696 1 so I know that the last line is my translation matrix. However when I multiply 0.4102mm/pixel or 0.04102mm/pixel with the translations I do not recover my known shifts. Does anyone know what I might be doing wrong? I really need help on this.