影像紋理辨識之研究
碩士 === 國防大學理工學院 === 電子工程碩士班 === 99 === In this thesis, our major purpose is to extract the texture information from the images. These images are embedded with some common texture information in each group. We first analysis the test image and extract its texture features. The feature vector will be...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | zh-TW |
| Published: |
2010
|
| Online Access: | http://ndltd.ncl.edu.tw/handle/59478241884382156220 |
| Summary: | 碩士 === 國防大學理工學院 === 電子工程碩士班 === 99 === In this thesis, our major purpose is to extract the texture information from the images. These images are embedded with some common texture information in each group. We first analysis the test image and extract its texture features. The feature vector will be build. The images are also stored with its corresponding texture features. We will calculate the feature distance with the test and database images to find the most similar image as the query image. Our main algorithms are NGLDM and GLCM which are used to calculate the two dimensional statistical information form the image.
We will apply the wavelet analysis method with NGLDM and GLCM algorithms to extract the more representative features. Besides the space correspondences, the frequency correspondences are also extracted. The one dimensional statistical information is also joined in the feature to form a more accurate feature vector.
We compare the texture differences with two methods. One is one-vector-based and the other is clustering-based. The one-vector-based method is comparing directly the difference of two vectors. The clustering-based is first classifying the vectors into several groups then comparing the features in each group.
|
|---|