We attempt to realize a method that enables 3D structure estimation of a room represented by parameters of a cuboid based on a segmentation result. By this method, obtain surface information with 3D structure of the target room required for interactive virtual preview can be obtained. To improve our previous method, this study proposes a novel approach for semantic edge extraction based solely on logical operations. The proposed method enables independent edge extraction in a 2D image corresponding to a cuboid representing the target room’s 3D structure. Accordingly, two color patterns are assigned to the output of semantic segmentation to generate various edge images. Subsequently, logical operations performed on five types of edge images the generate of the eight semantic edges. Experimental results using CG-based datasets demonstrated that parameters corresponding to each semantic edge were accurately computed, even when inference results of semantic segmentation were used as input. Specifically, slope and distance errors were typically less than \(1^\circ \) and 1 pixel, respectively.