Filter Selection for Speckle Noise Reduction

Abstract

This work provides the knowledge about adaptive and anisotropic diffusion techniques for speckle noise removal from different types of images like synthetic, photographic, ultrasound and Synthetic Aperture Radar images and the various filters based on anisotropic diffusion for the removal of speckle are also being discussed. A comparative study is made on the performance of the Lee filter, Kuan filter, Frost filter, SRAD1 filter, AD filter, and SRAD2 filter in removing the speckle from the image and in preserving the edges. Finally an algorithm is developed which performs all the filtering techniques on the input image and the statistical parameters are calculated for the output images obtained from all the filters. These statistical parameters are weighted on the scale of 10 and compared, the images corresponding to the best statistical value are displayed along with the filter name and corresponding value of the statistical parameters. For the evaluation of the performance of above mentioned filters statistical parameters like Signal to Noise Ratio, Root Mean Square Error, Peak Signal to Noise Ratio, Correlation of Coefficient, and Mean Structural Similarity Index Measure are used and the MATLAB codes required in calculating these parameters are developed. These parameters are used to calculate the image quality of the output image obtained from above mentioned filters, based on the values of these parameters the performance of the filters in terms of speckle removal and edge preservation is discussed. Usually each filter gives optimal result for a particular type of image at a particular parametric value like SRAD1 gives good results for ultrasonic images and AD gives good results for synthetic images. Hence the proposed algorithm relieves the human from taking the decision regarding which filter is to be used for an image type since this algorithm selects the best output among the different filter outputs and displays the optimal results among the different results produced by different filtering techniques