Iterated Function Systems (IFS) seem to be used best to represent objects in the nature, because many of them are self similar. An IFS is a set of affine and contractive transformations. The union (so-called collage) of the subimages generated by transforming the whole image produces the image again - the self similar attractor of these transformations, which can be described by a binary image. For a fast and compact representation of those images, it would be desirable to calculate the transformations (the IFS-Codes) directly from the image that means to solve the inverse IFS-Problem. The solution presented in this paper will directly use the features of the self similar image. Subsets of the entire image and the subimage to be calculated are identified by the computation of the set difference between the pixels of the original and a rotated copy. The rotation and the scale factor of the transformation can be computed by the mapping of this two subsets onto each other, if the translation part - the fixed point - is predefined. The calculation of the transformation has to be repeated for each subimage. It will be proved, that with this method the IFS-Codes can be calculated for not convex, undistorted, and self similar images as long as the fixed point is known. An efficient algorithm for the identification of these fixed points within the image is introduced. Different ways to achieve this solutions are presented. In the conclusion the class of images, which can be coded by this method is defined, the results are pointed out, the advantages resp. the disadvantages of the method are evaluated, and possible ways to extend the method are discussed.