The Research Of Remote Sensing Texture Analysis Based On The Clifford Wavelet

Texture is one of the basic features of images.Texture feature,which is an important element in remote sensing images describing and analyzing,reflects the distribution model and spatial relations of greys in images.Texture analyzing is one of the key techniques of computer visual,image processing,image analysis,image retrieving and so on.Wavelet transform,which is one of the most important methods in signal processing,has been wildly applied to texture image analysis owing to its good spatial-frequent analyzing ability and multi-scale analyzing ability.However,there are some drawbacks of traditional wavelet methods like translation sensibility and lacking of phase information restricting the application of wavelet methods to texture image analysis.DT-CWT is approximately translation invariant and has more directions and limited redundancy.However,there is only one phase parameter in DT-CWT,which reflects the position information of pixels and is disabled to describe the texture structure.On this basic,the Clifford wavelet transform is proposed,which is a new multi-scale spatial and frequent analyzing tool.It not only is approximately translation invariant,but also can provide one magnitude and three phase parameters in every scale.More important,the third phase parameter is related to texture structure.In this paper,on the basic of research of theories and properties of Clifford wavelet transform,the texture feature parameters are constructed based on the magnitude and phase information and applied to texture classification and texture segmentation of images.The main works in this paper are as follows:1.Analyzing and researching the basic theories of Clifford wavelet transform,including the definition and properties of quaternion algebra,the definitions of 1-D,2-D and quaternion analytic signals,the definition and some properties of Clifford Fourier transform and so on.Moreover,the Mallat algorithm and design of filters of Clifford wavelet transform have also been analyzed.2.By studying the properties of magnitude and phase parameters of Clifford wavelet transform,the new texture feature parameters are constructed based on magnitude and the third phase to describe and distinguish different texture structures.3.The translation,rotation and scaling invariant texture classification algorithm has been proposed based on the Clifford wavelet transform and Log-Polar transform.The new texture feature parameters are also used in texture extraction.The experiments are performed to compare the performance of the texture classification algorithm proposed in this paper with other algorithms.The results show that our algorithm is more efficient in recognizing texture structures with geometric distortions.Another experiment is designed to compare the new texture feature parameters with those traditional texture parameters.And the results show that the 'new texture feature parameter is more powerful in distinguishing different texture structures.Finally,our algorithm is applied to classification of complex remote sensing images and obtains good results.4.The theories and methods of Clifford Gabor wavelet are studied.And then the Clifford Gabor wavelet based multi-scale and multi-direction texture segmentation algorithm has been proposed.The algorithm not only considers the complementarity of texture features of the same texture on different scales,but also truly utilizes the directional difference between textures.Besides,our algorithm also takes the position features of spatial data into account by adding spatial constraint characteristics.The first experiment is performed to compare our algorithm with traditional texture segmentation algorithm.The results show that our algorithm has better performance in texture segmentation.The second experiment is performed to compare the new texture feature parameters with traditional texture parameters.And our new texture feature parameters are also powerful and efficient in texture segmentations of images.