Figure 14.

Relationship between patch size ${\mathbf{S}}_p$ and both error and invalid pixel ratios for various prior window sizes ${\mathbf{S}}_w$ at a fixed intensity threshold $I_{th}=0.1$. (a) shows how smaller values of ${\mathbf{S}}_w$ and larger values of ${\mathbf{S}}_p$ tend to minimize error, with an optimal configuration observed around ${\mathbf{S}}_w=17\times 17$ and ${\mathbf{S}}_p=24\times 4$. (b) demonstrates that smaller values of ${\mathbf{S}}_w$ generally lead to higher invalid pixel ratios, while smaller values of ${\mathbf{S}}_p$ help in reducing invalid pixel ratios. This indicates a trade-off in parameter selection between minimizing error and reducing invalid pixel ratios.