The Research On The Sub-Pixel Edge-Extraction Based On Fractal

The geometry of fractal give man a more effective and objective tool to describe theworld compared with the traditional geometry,and it has been widily used in natural andscience.After study the theory of the geometry of fracta,this paper make a few changeson the edge detection algorithm based on the traditional geometry of fractal.The mainmassions of this article are as follws:1.Introduce sereval classic fractal dimensions,and analyzes their adventages anddisadvantages.The fractual dimensions are divided into used for theory study and usedfor the image progressing by function.In each class,there are different dimensions by thedifferent ways of calculate,Firsty this paper introduce the base principles and thecalculate methods.of several fractual dimensions.and the scope of their application.2.Used the image amplifier technology based on fractual dimension to enhance theimage information.Image amplifier technology are divide into linear interpolation andnon-linear interaplation.The methord we used in this paper is a kind of non-linearinteraplation technology.This mehtord use the self-similarity quality of and informationof the image block.In one way,it can keep the self-similarity quality.of the image, inanother way,it can enhance the image information.3.This paper introduce the direction of pixel into the progress of the calculation ofthe fractual dimension.So get the new fractual dimension reflect the neighborhood of hepixels.The new fractual dimension with direction,make use of the quality that,differentedge have different neighborhood structure.to decide weather the pixel on the edge.aftrethat it use the self-simility quality and IFS to delete the pixels which maybe is relex ordefined by mistake.that is the innovation of this paper.Experimental results shows thathe algorithm can remove the doble border,and achive the sub-pixel edge.The is paper also propose a new methord,which do the fuzzy clustering on thespectrum of the image,in a way this mehord enhance the robustness of the conventionalalgorithms.