Wavelet based Texture Classification for Remotely Sensed Data - IEEE Paper

0
   

ABSTRACT:
In this paper the application of nonseparable gabor wavelet transform for texture classification is investigated.  The effect of applying the discrete cosine transform is compared as a traditional method with gabor wavelet for texture extraction of the  remotely sensed data (IRS LISS III digital image) of Madurai metropolitan area Tamil Nadu, India. The methodology consists of five stages:
1) Applying gabor wavelet transform on the data  
2) Extracting the features 
3) Classifying the feature vectors   
4) Comparing the results with discrete cosine transform output 
5) Accuracy assessment. As a first stage of this process, Gabor wavelet is applied on the image. The wavelet decomposed image is given as an input for the feature extraction stage, where the necessary features are extracted for classification. The  feature vectors are then subjected to K-means clustering algorithm, thereby obtaining the classified image. The classified image contains different classes such as urban, vegetation, water body, hilly region and wasteland. 

Attachment:
.pdf   Wavelet based Texture Classification for.pdf (Size: 342.71 KB / Downloads: 364)