Communications of the ACM
Digital Image Processing
| Hi-index | 0.00 |
In this paper a novel filtering scheme has been proposed for the removal of random valued impulsive noise from images. The scheme is based on a second order differential impulse detection mechanism followed by a recursive median filter. The salient feature of the impulse detector is that it utilizes an adaptive threshold derived from an artificial neural network. The neural network predicts the threshold using inputs from the noisy image characteristics and achieves generalization. It is observed from exhaustive computer simulation results that the proposed scheme out-performs the existing reported strategies in terms of noise rejection and detail preservation. In addition, as the proposed scheme uses a selective filter, the computation overhead is less in comparison to others.