We propose a method of corner detection algorithm using an adaptive octants filter paying attention to a characteristic of the edge pixel distribution around the corner. Changing the octants boundaries adaptively, we analyze the edge distribution among 8 octants for every edge pixel, and select some octants possible to contain edge segments. Furthermore, we extract some edge pixels as the corner candidates by evaluating the angles between the edge segments in the selected octants. The candidates are excessively extracted around the corner, and therefore, we find out the true corner using a cornerness measure.