Modeling for high resolution SAR image data

To resolve the problem of performance degeneration of K-S test for non-independent samples, one enhanced version of K-S test based on data per-treatment is proposed. Before K-S test, the samples are extracted with an interval which is decided by independence test in this enhanced method for ensuring independence. Considering the fact of the absence of statistical analysis about land image data of high-resolution SAR, the data of grassland is modeled by studying the data´s compatibility with SIRV model. To resolve the problem that the higher order statistical characteristic is mathematically tractable in SIRV model, the definition of “Coherent Region” is proposed, which means the image is processed locally, according to the viewpoint that the texture component is slowly-varying in SIRV model. The experiments show that the enhanced method of K-S test has perfect robustness for non-independent samples. The data for the grassland here are compatible with SIRV model. By analysis of grassland data, the rationality of the concept of “Coherent Region” is proved and the space scale of “Coherent Region” for the data is found here.