Digital imaging systems are ubiquitous and impact many fields of research and industry, from (biomedical) electron microscopy to radio astronomy. Nearly all systems use some form of digital image restoration or reconstruction algorithm (DRA) to either improve image quality or as an essential part of their operation. The large impact of DRAs has enabled new applications, such as light field cameras, heterogeneous sensor devices, microwave tomography or HDR and 3D television. It is evidenced by the recent Nobel Prize for (the resolving power of) STED fluorescence microscopy. As DRA applications grow in number and become more diverse and complex, so does DRA design. This is exacerbated by the fact that developing better-quality DRAs often means tailoring to specific application use cases (which are generally not even known in full detail yet). To solve these limitations to digital imaging, we propose a new methodology for the design of DRAs. The main idea of this methodology is to allow a human to directly input application-specific information in a general DRA. A human can provide the DRA with very accurate models about noise, the acquisition system and the type of image that is expected for a very specific use case because the human can base those pieces of information on his expertise and current user demands. By comparison, existing DRAs are often based on powerful, yet less specific models by necessity because a DRA designer cannot accomodate all possible use case details.