Morphological hit-or-miss transformation for shape recognition

Abstract

In this paper, the morphological hit-or-miss transformation is analyzed for use in the recognition of both perfect shapes and imperfect shapes. Shape recognition is achieved by locating the objects to be recognized within the image. The shape recognition problem is discussed in the following two aspects. First, a theorem is presented in which the hit-or-miss operations employ the structuring elements [thetav]A and [thetav](W [subset of] Ac), instead of using sets A and (W [subset of] Ac). Here A is the boundary of A, and [thetav](Q [subset of] Ac) is the containing A, [thetav]A is the boundary of A, and [thetav](W [subset of] Ac) is the boundary of (W [subset of] Ac). Second, the recognition of imperfect shapes due to indeterminate variation of an object shape is studied here. Our method employs a priori known shape information as a basis for structuring elements and constructs structuring elements which are then used in a hit-or-miss transformation to find the location of the shape to be recognized. Each occurrence of a target shape is represented either by one point or by a small cluster of points within a calculated range according to the associated structuring elements. Finally, we present the hit-or-miss operation without window restrictions as a technique for recognizing both perfect and imperfect shapes, thereby making the method more flexible.