Eight SV-PSFs associated with each block (one for each block vertex location) are computed using the $N$-interface PSF model,^{32} which models light propagation through $N$ stratified layers within a block. SV-PSFs are calculated at discrete locations, $(xo,yo,zo)=(Xm,Yn,Zk)$, marking block vertices, using imaging conditions including thickness and RI of the sample at these unique locations. These SV-PSFs can be represented using a few principal components (PCs), thereby reducing not only the memory required but also the number of convolutions in the forward SV imaging model.^{33} The PCA formulation used in the SV imaging model is an extension of the PCA approach developed previously to represent DV-PSFs,^{25} in that it uses SV weighting coefficients, $cj(xo\u0332)$, instead of the DV ones, $cj(zo)$. Each SV-PSF can be written in terms of the PCA as follows: Display Formula
$h(x\u0332o,x\u0332i)=\u2211j=0RPj(x\u0332i)cj(x\u0332o)\u2248\u2211j=0JPj(x\u0332i)cj(x\u0332o),$(4)
where $P0(xi\u0332)$ is the mean of the SV-PSFs, $Pj(xi\u0332)$ is the $j$’th PC, and $cj(xo\u0332)$ are the corresponding SV coefficients, with $c0(xo_)=1$ for all $x_o$. The SV weighting coefficient is computed by the following inner product: $cj(xo\u0332)=\u2329Pj(x\u0332i),[h(x\u0332o,x\u0332i)\u2212P0(xi\u0332)]\u232a$. Each location $(xo,yo,zo)=(Xm,Yn,Zk)$ is associated with a unique $cj(xo\u0332)$ for each PC. As $j$ increases the value of $cj(xo\u0332)$ becomes close to zero [see Fig. 2(h)], allowing the use of fewer components ($J<R$) in the approximation of Eq. (4).