Rahman, Arif and Widyantoro, Dwi H. (2007) Fractal-Code Based Image Retrieval System. [Artikel Dosen]

Text fractal-ir-arif.pdf Download (5MB)

Abstract

Fractal code is a code from image encoding process using Partitioned Iterated Function System (PIFS) method. The characteristic of PIFS is self-similarity searching between image parts based on pixel intensities value. Thus it can be assumed that similarity of fractal code represents similarity of pixel intensities variance pattern which represents the image texture. In this paper, fractal code based image retrieval system called SPECKTRAL is proposed. Fractal code in this system consists of two components that invariant to translation, scale and rotation, i.e. contrast and brightness coefficients. Fractal code similarity between two images measured with the mean value of minimum Euclidean distance between each contrast and brightness coefficients. If average distance smaller, then the image considered has more similar texture. The experiment results on Brodatz and Vistex image collections show that system can retrieve most images that have relevant texture to the query in early retrieval process i.e. when 25% of images in the collection retrieved and performs better for textural image collection.

Actions (login required)

View Item