Abstract
This paper proposes a novel local descriptor, local texton center symmetric texture matrix (LTCSTM) for texture classification on wavelet domain. The proposed LTCSTM extracts i) structural features from texton representation ii) Local texton center symmetric pattern (LTCSP) code iii) integrates the above two features with gray level co-occurrence matrix (GLCM) features. The texture classification is performed using machine learning classifiers. Initially the raw image is transformed in to wavelet based image. The LL-1 image is sub divided in to local regions of size 2 x 2 and each region is replaced with texton index. The LTCSP is derived on texton index image. The LTCSP code replaces the center pixel of the 3x3 window. The derivation of co-occurrence matrix on this LTCSP coded image derives the proposed LTCSTM. The GLCM features on LTCSTM are used for texture classification. The proposed LTCSTM is compared with state-of-art of texton based methods and local descriptors of LBP on five popular databases. The experimental evidence clearly indicates the efficiency of the proposed method over the rest of the state-of-art methods.
Original language | English |
---|---|
Pages (from-to) | 440-445 |
Number of pages | 6 |
Journal | International Journal of Innovative Technology and Exploring Engineering |
Volume | 8 |
Issue number | 2S |
Publication status | Published - 01-01-2018 |
All Science Journal Classification (ASJC) codes
- Computer Science(all)
- Civil and Structural Engineering
- Mechanics of Materials
- Electrical and Electronic Engineering