SAR Image Noise Reduction Based on Wavelet Transform and DnCNN

Research on SAR image denoising method based on feature extraction

“…In SAR image, the relative phase between the scattering points in each resolution unit is closely related to the radar viewing angle, speckle noise arises from the coherent superposition of echoes of many scattered points randomly distributed in the same resolution unit in a scene. Studies have proved that the speckle noise of SAR images is a kind of multiplicative noise, and the following multiplicative noise model is used to describe the SAR image noise [15]:…”

“…In SAR image, the relative phase between the scattering points in each resolution unit is closely related to the radar viewing angle, speckle noise arises from the coherent superposition of echoes of many scattered points randomly distributed in the same resolution unit in a scene. Studies have proved that the speckle noise of SAR images is a kind of multiplicative noise, and the following multiplicative noise model is used to describe the SAR image noise [15]: $$$$\begin{equation} z(x,y) = F(x,y) \cdot N(x,y) \end{equation}$$$$where $$z(x,y)$$ is the noisy image, $${{\bf x}} = (x,y)$$ is a two‐dimensional pixel location, and the image is projected at different scales $$r$$ with gradient norm $$\nabla z$$, $$F(x,y)$$ is the clean image, and the amplitude signal of the multiplicative noise $${{\bf N}}=N(x,y)$$ in the SAR image obeys the gamma distribution [16]: $$$$\begin{equation} \rho ({{\bf N}}) = \frac{{{L^L}{{{\bf N}}^{L - 1}}{{\rm {e}}^{ - L{{\bf N}}}}}}{{\Gamma (L)}} \end{equation}$$$$where $$L<1$$, $${{\bf N}} > 0$$, …”