The Methodological Research To Enhance Edges Of Remotely Sensed Imagery Based On Cluster Structure Information In Feature Space

One of the most fundamental and most important characteristics of images isincluded in their edges, accurately and effectively extracting information of which isconducive to image analysis, interpretation and application. Edges often becomeblurred duo to the impact of various external factors in capturing and transmittingimages. So it has been an important research topic in digital image processing toeliminate blur and highlight the structural information of image by edge enhancementtechniques. Most edge enhancement methods, such as Unsharp Mask method,gradient method, etc. proposed in current literature are only suitable for grayscaleimages although they are now relatively mature. However, with the progress oftechnology, color images are gradually popular in most application fields. More andmore attentions are drawn to the research on edge enhancement of color images,which is still in exploration without breakthrough. Considering the disadvantages ofresearch on color image edge enhancement methods, we proposed a method toenhance edges of remotely sensed imagery based on cluster structure information infeature space, and performed a case study by taking Quickbird imagery of Fuzhou assample data.The main content is listed as below:(1) The history and present situation on studying image edge enhancement arereviewed. On the basis of systematically analyzing the advantages and disadvantagesof presently available research methods, a new idea to enhance edges of remotelysensed imagery based on cluster structure information in feature space was proposedfor their improvement.(2) It was examined that the effect of edge enhancement resulted respectivelyfrom synthetic methods such as Lower Upper Middle and Laplacian, as well as vectormethods such as Multichannel Edge Enhancing Filter and Sharpening Vector Median,which shows that the outcome of edge enhancement from the vector methods isoverall superior to that from the synthetic methods.(3) In the process of analyzing and extracting structure information from remotely sensed imagery data, a clustering method based on geometric probabilitywas introduced and extended to three-dimensional feature space, by means of whichthe method to extract cluster structure information from remotely sensed imagery dataand to perform edge enhancement based on the extracted information has beendesigned, the related algorithm has been implemented as well.(4) The multi-spectral remotely sensed imagery data of experimental zone hasbeen handled by the feature-space-cluster-structure-information-based algorithm foredge enhancement proposed and implemented in this thesis with the support of VisualStudio2005development platform. The relative advantages of the method proposedin the thesis on edge enhancement of remotely sensed imagery data were analyzed bycomparing experimental results from the method proposed in the thesis with thosefrom methods having been introduced in presently available literature.