A two-pass method to impulse noise reduction from digital images based on neural networks

One of the most research challenges in image processing is image enhancement and reducing impulse noise from digital images. There are various methods for impulse noise reduction such as median based filters or nonlinear filters, but these methods more or less cause images to blur and to remove important details from images, as in high noise ratio in that noise reduction will destroy vital information such as edges and high amount of noise causes the image information be destroyed. Some ways are proposed to impulse noise reduction using soft computing that has a good performance. This paper presents an efficient method in two passes for reduction of impulse noise. At the first pass impulse noise detection using ANFIS, and at the second pass the impulse noise estimation, that corrupted noise pixel replaced with new value based on ANN. Our method is experimented on some popular grayscale test images and is compared to other methods using subjective and objective measures. Results show that our proposed method is efficient in impulse noise reduction and works better than the other compared methods.