||Many feature matching algorithms compare square areas around feature points, which implies that the orientation of each feature must be obtained first. In the work presented, circular areas are used, which mean that the area is always the same regardless of orientation. Two different approaches will briefly be presented that both are invariant to rotation. One is extremely fast and compact, with the drawback of being sensitive to scale changes. The other is slower, but have the advantage of not being sensitive to scale, within certain limits. This is a work in progress and comments and suggestions are very much welcome!