The Study On Texture Features Extraction Of SAR Image In Wavelet Domain

Texture analysis plays an important role in remote sensing analysis, medical image processing, pattern recognition and great database retrieval. The description and extraction of SAR image texture feature is important to texture segmentation.A feature extraction method based on wavelet and gray-level run-length matrix is presented in this dissertation, with a description of local pattern and array rules by run-length, and a combination with scale-feature of wavelet. Our method with a combination of wavelet and run-length, a construction of gray-level run-length matrix, and an introduction of new statistic is applied to feature extraction and segmentation of SAR image texture. The experimental results suggest that our method denoise images effectively and make a good performance.Using the multi-scale feature of WT, a new method is proposed based on wavelet domain and scale space for image texture extraction. According to the relation of features among different scales, the algorithm constructs a scale-based run-length matrix in wavelet domain, and then extracts the features for segmentation. Experiments show that the algorithm can get more accurate results for segmentation. Compared with the traditional methods, it can describe the scale characteristic of texture more general and accurate. The dimension of this method is low, that is easy to segment.