Coastline Detection In Sar Images Based On Stationary Wavelet Transform

Synthetic Aperture Radar (SAR), as an active microware imaging radar, is an effective tool to observe the coastline with its characteristics of all-weather, all-time, high resolution and wide coverage. So far, lots of SAR satellites have been launched and three of them belong to our country. This results huge amount of SAR data needed to be classified and processed, especially in coastline monitoring and detection. Due to the image noise and low resolution factors, it is difficult to automatically detect and classify coastline in SAR images. In this thesis, a coastline detection method based on Stationary Wavelet Transform (SWT) is developed, and wavelet analysis and some image processing algorithms are combined to analyze the available SAR images data. This accurate coastline feature detection method can also reduce the Speckle noise.In this thesis, three parts are discussed and stated as follows.(1) SAR coastline images preprocessing:Firstly, we use 10*10 pixel size window to compress SAR images aiming to reduce the image processing complexity without affecting the coastline features; Secondly, we detect and remove the black boundary lines or areas automatically to avoid the false line features; Thirdly, we apply gray level histogram equalization method and contrast stretching method to enhance the coastline images gray contrast. Experiments show that this method not only improves image gray contrast but also reduces post-processing computation cost, and also limits the Speckle noise.(2) SAR images Speckle noise filter algorithm methods:including the traditional filter algorithm, local statistical adaptive filter algorithm and wavelet filter algorithm. Based on this three filter methods, an improved wavelet low pass filter method is introduced. This method used the multi-scale analysis of wavelet transform features, and combined with the advantages of the enhance Lee filter in the local statistical adaptive filter algorithm to further suppress SAR images Speckle noise. From the experimental results, the improved method can effectively remove the SAR images Speckle noise, through removing high frequency information and keeping low frequency information of wavelet transform.(3) A coastline detection method based on Stationary Wavelet Transform in SAR images:In this method, SWT is applied to process SAR images and its coefficients are used to calculate and generate the Wavelet Gradient Information (WGI). The coastline detection is obtained as edges by searching the module maximum according to WGI. The morphological thinning algorithm and threshold technique are also used for edge refinement to suppress the non-coastline features such as the isolate Speckle noise. The experimental results obtained from ENVISAT-1 SAR images show that the proposed method is efficient in SAR images coastline detection.