In this work an efficient algorithm for estimation of the Fractal Dimension (FD) of images is presented. At first, the approach is tested on the synthetic images. It is expected that the PD range is 2.0 – 3.0. A good method, as this approach, should reflect this desirable feature. The utilization of such algorithm on textile imperfection identification is investigated. The FD is a feature proposed recently to characterize roughness, self-similiarity and the complexity degree in a picture. This characteristic has been used in textures segmentation and classification, shape analysis and other problems. However, its utilization on image change characterization is a new feacture. Experiments has been done, not only on synthetic images, but also on real textile. The relation of a picture scanned at various different orientation and relative rotation of digital images are also discussed.