Abstract:
Wavelet transform can decompose images into various multi resolution sub bands. In these sub bands the correlation exists. Most edge detectors are based on- local measurements of the image variations and edges are generally defined as 'block' where the image intensity has a maximum variation[1]. The image variations can be examined by derivative operators, such as gradient, Laplacian, Sobel and Roberts operators. In this paper we have used a standard method of edge detection Sobal, Perwit, Robert, Kirsch and Robinson. In addition we have used the wavelet transform in three levels, in each of these level the HL, LH return to the original size. Also,a hybrid method (wave late transform + perwit) to improve the image edge detection have been used.More over the standard deviation to the (2*2) block of the (LL)in the level (3) has been applied. The threshold to the block of the of HL, LH,HH of different size in the level 3,2,1 has been computed compared with the standard deviation results of the third (LL) band. And when the comparison completed we shall construct the image and finally apply the Perwit operator to the constructed image. |