Image De-Noising With Machine Learning: A Review

Images are susceptible to various kinds of noises, which corrupt the pictorial information stored in images. Image de-noising has become an integral part image processing workflow. It is used attenuate noises and accentuate specific within. Machine learning important tool image-de-noising workflow terms its robustness, accuracy, time requirement. This paper explores numerous state-of-the-art machine-learning-based de-noisers like dictionary models, convolutional neural networks generative adversarial for a range Gaussian, Impulse, Poisson, Mixed Real-World noises. The motivation, algorithm framework different machine analyzed. These compared using PSNR as quality assessment metric on some benchmark datasets. best results noise types discussed along with future prospects. Among Gaussian de-noisers, GCBD, BRDNet PReLU network prove be promising. CNN+LSTM, MC2RNet most suitable CNN-based Poisson de-noisers. For impulse removal, Blind CNN, CNN+PSO perform well. mixed WDL, EM-CNN, SDL, CNN prominent. De-noisers GRDN DDFN show accurate domain real-world de-noising.