|
Sythetic Aperture Radar (SAR) systems employ coherent techni- qes to generate images of terrain in the microwave region of the spectrum. The speckle noise is due to the coherent nature of the SAR image-formation process. Since the presence of speckle in an image reduces the resolution of the image and the dectability of the target, speckle reduction is an important work for SAR image processing. A simple method of speckle reduction is multilook and the con- cept of the typical multilook is only the look-averaging. A dir- ect application of an averaging spatial filter is inadequate,be- cause it smoothes not only the speckle but also the textured fe- ature. Thus, many adaptive spatial filters have been developed for speckle reduction, such as Lee and MAP filter. In addition, Wu and MAITRE have proposed the MHR filter. The idea of the MHR filter is using the local standard deviation and the evolution of the local standard deviation with window growing to test the homogeneity of a window around each pixel and to find the most suitable window to perform filtering. The filtering it-self may be any smoothing filtering (Mean, Lee, MAP filter). Homogeneity testing by the local standard deviation has some limitations, so that we have the movitation to improve the MHR filter. We employ different local statistics(the local contrast and the normalized evolution of the local standard deviation) to implement the idea of the MHR filter, named as the Modified MHR (MMHR) filter. We also propose the method of the estimation of the Equivalent Number of Look(ENL) by the simulated SAR image, such that the decision of the filter's parameters becomes eas- ier. Experimental results show that the MMHR filter has better performance than the MHR filter. A method for the accomplishment of the idea of the MHR filter using Fuzzy theory is also propos- ed, named as Fuzzified MHR(FMHR) filter. The FMHR filter is more flexible in fulfilling the desired filter performance.
|