New fractal coding scheme using reference images

We present a new fractal coding scheme to find more optimal transformations using reference images, which are employed as seeds for obtaining candidat es for the optimal transformations. Each transformation that minimizes the distance between an original image and a reference image is a candidate. It is actually impossible to find optimal transformation due to heavy computa tion. Thus, instead of considering all of the allowable transformations, we select a few transformations as candidates for the optimal transformation, and thereafter select the best from that group. Our scheme can be consider ed a "generalized collage coding scheme," since its process for each refere nce image is similar to the collage coding process. That is, the collage co ding scheme is a special case in our scheme, with only one reference image. At first, in a simple case where the optimal transformation can be obtaine d, our scheme is experimentally evaluated, as compared with the optimal one . In general cases where the optimal one is unavailable, our scheme is also evaluated compared to conventional schemes. (C) 2001 Society of Photo-Opti cal Instrumentation Engineers.