SAR speckle reduction model based on 2D two-channel filter bank

In order to improve the capability of synthetic aperture radar (SAR) speckle reduction, the speckle reduction model of a 2D two-channel filter bank multiscale structure is presented. Mcclellan transformation is implemented for 1D two-channel filter bank to obtain the corresponding 2D two-channel filter bank. Because the mapping function of Mcclellan transformation is zero-phase, the phase and frequency response of 1D prototype filter is easy to be reserved. Then, the 2D two-channel filter bank multiscale structure is constructed. A nonsubsampled method is utilized in multiscale decomposition to make this multiscale structure shift-invariant. Finally, the corresponding SAR speckle reduction model is established, in which the original SAR image with multiplicative noise is transformed into that with additive noise by homomorphic transform (i.e. a logistic operation), the image is decomposed by 2D two-channel multiscale structure, and threshold method is used to separate the noise and signal in high frequency section. The simulative experiments show a comparison between the proposed model and the wavelet denoising method. For a remote sensing image with multiplicative noise, the improvement of signal to noise ratio (SNR) of the proposed method is 1 dB. Por an original SAR image, the equivalent number of looks (ENL) of both methods keeps balance and the improvement of peak signal to noise ratio (PSNR) of the proposed method is 4.5 dB. The results show that the performance of 2D two-channel filter bank is superior in both speckle reduction and detail reservation to the 2D wavelet transform.