The Sonar Image Filtering Based On DTCWT

Sonar image filtering is the first step of sonar image preprocessing, which plays an important role in the image processing,because the result will directly affect the image segmentation, feature extraction and other important links, so the sonar image filter is of great significance.There are many methods for sonar image filtering, this paper use the average filtering, median filtering, Lee filtering, Kuan filtering, Frost filtering and wavelet threshold filtering to process the Lena image with noise and two sonar images. At first, through analysising each method with different parameters, and comparing the effect of each method depending on the peak signal-to-noise radio?equivalent noise level and the processing time, than get the parameters when the effect achieve better. After a lot of experiments, The hard threshold wavelet filtering method has the best effect.Dual tree complex wavelet transform is presented on the basis of wavelet transform, which not only inherites the time-frequency partial analysis features, but also has an approximate shift invariance and directional selectivity. This paper presents a new method based on dual tree complex wavelet transform and adaptive median filtering image filtering method, this method will combine the excellent characteristics of dual tree complex wavelet transform and the advantages of median filtering to remove impulse noise effectively, which can filter out the speck noise well in the sonar image. According to the experiment analysis, it is concluded that the new method, is better than the hard threshold wavelet filtering method and the dual tree complex wavelet threshold method.