The restoration of paintings on elements in cultural heritage buildings (fundamentally, churches) involves two structural problems: capturing the geometry of the construction element and its development. In many cases, the geometries are regular (e.g., cylinders, spheres, elliptical domes). However, there are cases in which the elements cannot be adapted to any known geometry, much less one that can be mathematically developed. The development of surfaces becomes essential for the restoration of paintings over "flat elements" (over which work is performed on the ground) that are subsequently transferred to the real surface (ceilings). The mathematical transformations that allow regular geometries to be developed are widely known (cartographic projections). However, when the geometry is irregular, there is no development. This study presents a new methodology based on differential rectification and its application for the development of oculi in the Los Santos Juanes Church (Valencia), whose geometry is completely irregular both in shape and as a result of construction defects (and damage caused by fire). The present study focuses on the restoration of paintings damaged by fire.
Reverse engineering applied to architectural restoration for the reconstruction of structural surfaces depends on metric precision. Sometimes there are elements on these surfaces whose value is even higher than the building itself. This is the case for many churches whose ceilings have pictorial works of art. Reconstruction requires the existence of some identifiable remainder and/or a surface geometry that enables mathematical development. In our case, the vault has an irregular hemispherical geometry (without possible mathematical development), and there are no significant remains of the painting (which was destroyed by a fire). Through the 3D modelling of the irregular vault and two historic frames with a camera of unknown geometry, an inverse methodology is designed to project the original painting without metric deformations. For this, a new methodology to locate the camera positions is developed. After, a 3D virtual mathematical model of the complete image on the vault is calculated, and from it, partial 3D virtual images are automatically calculated depending on the variable unknown positions of the video cannons (distributed along the upper corridor of the apse) that will project them (visually forming a perfect complete 3D image).