Multiclass CNN-based adaptive optimized filter for removal of impulse noise from digital images

Multiclass CNN-based window adaptive optimized filter has been proposed in this research work to remove impulse noise from colored images. Instead of applying adaptive filter on the whole images, selective adaptive window has been applied on the test image so that the effective computational complexity of the system is reduced. In this article, multiclass CNN has been used for the choosing the range of selective window. After that, vectored minimum mean value-based detection is being applied on the pixel under operation on a particular kernel of image for detection of noise and defining the appropriate size of the window surrounding the pixel under operation. An adaptive vector median filter incorporated with particle swarm optimization (AVMPSO) is passed over the corrupted pixel if the pixel is determined as corrupted after detection of noisy and noise-free pixels. The performance comparison of the proposed AVMPSO has been carried out along with the available state-of-the-art filters in terms of peak signal-to-noise ratio (PSNR), mean square error (MSE), structural similarity index matrix (SSIM), and feature similarity index matrix (FSIMc). A large set of images has been considered to conclude how AVMPSO outperforms the existing methods at all levels of noise by similar to 1.5 to 3 dB (PSNR) at least.