Segmentation Method of Building Surface Images

We are studying a coding process of 3-dimensional urban maps that require large amount of memory space, in order to let the maps be available with less system resources. As the first step, we developed a segmentation method of building surface images, which provides an extraction of repetition patterns that will be important in efficient coding of the images. Utilizing the characteristics of the building surface images, we designed a method including the repetition extraction based on “1D image" that is made from vertical or horizontal averaging of an image pixel array. We applied the method to several images and obtained adequate results.