Figure 5.

Aligned reconstruction convergence against effective iterations j defined as a single application of the encoding or decoding operator for a single motion state and coil channel at full resolution. A, Different encoding orders, number of segments M ∈ {4, 64} (rows), motion levels $\theta \in \{2^{\circ },5^{\circ },{10}^{\circ }\}$ (columns), $j_{\max }=20000$. B, Different encoding orders and number of multiresolution levels L, number of segments M ∈ {16, 256} (rows), motion levels $\theta \in \{5^{\circ },{10}^{\circ },{20}^{\circ }\}$ (columns), $j_{\max }=200000$. C, Different encoding orders, L = 2, acceleration factors R ∈ {1 × 1,2 × 2} with matched number of segments M ∈ {64, 16} (rows), motion levels $\theta \in \{5^{\circ },{10}^{\circ },{20}^{\circ }\}$ (columns), $j_{\max }=200000$