Fig. 7. Methods comparison on Office Scene: Exposure time per each pixel measurement from first row to last row is 1 ms, 5 ms, 10 ms, 20 ms, 1000 ms (note that the 1000 ms Office Scene dataset was acquired with slight differences in the object location).

The total acquisition time from first row to last row is 23 s, 117 s, 4 min, 8 min, 390 min. The width of the reconstruction cube size in each dimension is 3 m as given along with other reconstruction parameters in Supplementary Table 4. Each column shows the reconstruction with different methods. The first two columns stand for our proposed RSD based solver with one or two spatial sections. The circle in the first column is actually the size of the farthest reconstruction plane which is the one with the largest region that is calculated with the same distance shift B₁. All planes in front of this one have a smaller reconstruction area; due to the maximum operation along the depth dimension, the circle size is defined by the largest one in the back. The third column is the Direct Integration (back-projection solver) as a comparison for the first two columns. The last two columns refer to the approximation method⁹ which approximate non-confocal by confocal data and solve it through the scanning-based solver (LCT: forth column, FK-migration: fifth column). For the last two columns, each small image shows the results from midpoint approximation⁹ in order to approximate confocal data from non-confocal measurements. The respective larger image results from zero-padding applied to the input data to show the same reconstruction volume as the first three columns.