Algorithm 2 Sampling
1:	$\mathrm{Input}: \mathrm{target} \mathrm{image} X_1$$; \mathrm{velocity} \mathrm{model} v_{\theta }$$: R^d\to R^d$ with parameter $\theta$; $\mathrm{Physics-Based} \mathrm{Noise} \mathrm{Model} C(X)$$; t_{best}=0$$; M_{best}=0$.
2:	Repeat
3:	$\mathrm{Sample}: X_0\sim N\left(0,1\right)$
4	$\mathrm{Sample}: T=C\left(X_1\right)$
5:	$X_2=v_{\theta }\left(X_0,{\rm T},0\right)+X_0$
6:	$\mathrm{for} t=s,2s,3s\dots ,ns$$\mathrm{do}; 0<s<1$$, 0<ns<1$
7:	$X_{\mathrm{t}}=t{X}_2+\left(1-t\right)X_0$
8:	$X_M=v_{\theta }\left(X_t,{\rm T},t\right)+X_0$
9:	$M$$=P\left(X_M,X_1\right)$ $P\left(X_M,X_1\right)$$\mathrm{means} \mathrm{computing} \mathrm{the} \mathrm{PSNR} \mathrm{between} X_M$$\mathrm{and} X_1$
10:	If $M$$> M_{best}$;
11:	$M_{best} =$$M$
12:	$t_{best} =$$t$
13:	$\mathrm{Return} t_{best}$