Reconstruction methods use the positions of samples and sometimes their estimated surface orientations to infer how a shape should connect. Different methods make different assumptions about smoothness, gaps and sample quality.
The result is an estimate rather than proof of the original object's exact geometry. Sparse or noisy samples can create missing regions, unwanted connections or invented surfaces that require inspection.
ELI5
Surface reconstruction turns a set of scattered shape samples into an estimated outer shell. The software has to make choices about the spaces between the samples.
For example, dots sampled around a ball can be used to build a curved shell. If a large patch has no dots, the software may guess its shape incorrectly.
Does surface reconstruction always need normals?
No. Requirements vary by method. Some approaches use point orientations, while others work from positions or estimate the information they need.
Is a reconstructed mesh automatically ready for use?
No. It may need inspection, cleanup or refinement for the intended application, especially when samples are incomplete or noisy.
