A new method to count the number of objects into an image via the skeleton of the image is presented. The technique can be applied to the case of convex or non-convex elongated or not elongated objects with a smooth contour without holes, different length, but similar thickness, and in the presence of occlusions. The number of objects or blobs into an image can be obtained as a function of two quantities: The number of terminal points (points with just one neighbor) and the number of three-edge-points (points with just three neighbors) in the skeleton of the image. The technique has been tested with a collection of one hundred syntactically generated images. In practically all cases, the technique has successfully given the desired result. Fails in the method are due to noise in the contour of the object.