Deep-learning Based Restoration of Moiré Patterns by Phase Features

In this paper, we focus on the restoration of a 3D height to inspect the components of a printed circuit board. The majority of automatic optical inspectors encounter issues with reflections and shadows caused by the lighting on the components. These conditions obstruct accurate measurements of shapes and heights. To efficiently restore these shadowed and reflected areas, we first define the regions of reflection and shadow based on the pixel values in an image. Then, we propose a network that incorporates a loss function by utilizing the characteristics of moiré patterns for restoration. The effectiveness of our proposed method is demonstrated through experimental results.