Flattening of the detection images (Sect. 2.8). The entire 1°  ×  1° simulated star-forming region at 350 μm is shown (upper left), the central area of which was used to illustrate getsources in this paper. The image of the converged footprints ℱ_λ somewhat expanded by convolution (upper middle) is used to interpolate all detected sources off the image to obtain the clean background image (upper right). The image of the standard deviations (lower left) is computed in a very small (3 × 3 pixels) sliding window and further median-filtered using a 21 × 21 pixels sliding window to reduce effects of possible artifacts. The smoothed scaling image ℐ_λF (lower middle) is obtained by convolving the image in the previous panel with a circular Gaussian beam larger than the maximum size of sources extracted at 350 μm in the initial extraction. The resulting flattened image ℐ_λDF (lower right) is obtained by dividing the original image ℐ_λD by the scaling image ℐ_λF. For better visibility, the values displayed in the panels are somewhat limited in range; the color coding is a linear function of intensity in MJy/sr.