Texture Feature Extraction And Analysis In Remote Sensing Image

With the rapid development of space and information technology, the quality of planet remote sensing and sensors'imaging both make a great process. Remote sensing image becomes one of the most important information sources for earth observation. It is widely used in national defense, detection of earth ecology and environment, agriculture, mining, surface engineering, city planning and detection and handling of major disaster, and so on.Texture reflects the periodic territorial feature of an image. It plays an important role in image processing as a method of feature description of image. Texture analysis is one method that studies the texture feature to process image and it plays an important role in digital processing. Especially in remote sensing image, texture analysis, which is based on the territorial property of image, becomes more important as a result that the objects in remote sensing are regarded as a study area.This dissertation is supported by the project Intelligent Methods of Grid Computing, Information Interpreting of High Resolution Sensing Image. This paper analyzed image rotation invariance and object based texture extraction method deeply after reviewing texture study achievements and new methods of image processing home and abroad. Firstly, this paper classifies the present texture methods of image processing into statistic, image processing, modeling and structure method. It details their principles, current situations and applications. And all methods are implemented on ELU, which is one platform of remote sensing image developed by Remote Sensing Lab of Shanghai Jiao Tong University.Secondly, the issue of rotation of texture is produced which is one of the most important problem in remote sensing image analysis. And more emphases are on Gabor and Local Binary Pattern methods in rotation invariant texture image.In the end, a new method of texture analysis– Gabor based texture motifs modeling methods is produced in this paper. It based on a reality that objects in remote sensing image are always composed of some special texture motifs, which is a representative spatial pattern in different instances of objects of remote sensing image. The location and orientation of texture motifs in different instances of the same objects are uncertain. So object can be described by texture motifs and it will simplify the description of spatial object. Results show that this method performance well in texture segmentation and building extraction of remote sensing image.