Image House employs the images as the solid model, the graph and the blob as space model for generating floor plans. The discontinuity in image color is mapped to the partition walls of the plan; the graph models the topology of the rooms and the blob of pixels models the shape of the room.  The images impose the patterns of composition of the plan, while the other two models address the requirements of rooms.

The graph is employed the model the topology of the rooms (represented by blob of pixels). The topological requirements are predefine by the designer.　The blob represents the rooms, the boundary of the blob is a part of the edges detected in the image. Two basic tools are employed to evaluate how good(or bad) a blob is for a room. The first tool is depth analyser, the second places a standard circle at the center of the blob in oder to evaluate the shape of the blob in a proximate way.

The topological and dimensional requirements of the rooms are predefined. The task is to assign the rooms to the blobs resulted from the image; There are two main goals: first, the selected blobs should satisfy the topological requirements of the rooms. Second, each of these blob should have proper shape. An optimization process is used to minimize the errors of assignment.

Programmed in Java

November 2011

video: