Image denoising using boundary discriminated switching bilateral filter with highly corrupted universal noise

There has been extensive research on removing universal noise from corrupted images using various kinds of non-linear filters. Median and bilateral filters are basic nonlinear filters which have been modified over the years to improve the quality of images. Median filters are good at removing Impulse noise, while these filters fail terribly in removing Gaussian noise. For Gaussian noise removal bilateral filter is preferred and successful over the years. But bilateral filter is not effective in removing impulse noise. Switching Bilateral Filter (SBF) works good for removing universal noise which involves noise detection followed by filtering. Detection plays vital role in image denoising of highly corrupted images. Boundary Discriminated Noise Detection (BDND) algorithm is very efficient in detection of noise even in highly corrupted images. Boundary Discriminated Switching Bilateral Filter (BDSBF) is proposed which employs BDND detection followed by SBF for better accuracy in noise detection even at higher noise conditions. Proposed filter also changes the size of the window based on the noise density computed from noise statistics after the BDND detection. Proposed filter shows superior results over SBF under high noise densities and high noise standard deviations.