A New Flat Pattern Oriented Order Statistic Filter for Impulse Noise Reduction from Highly Corrupted Images

When the image is highly corrupted, such as images photographed by the darkness regard camera, the detection of flat patterns is easier than that of edges. In this paper, a new adoptive order statistic filter by using flat pattern detection is proposed, and an approach to optimize the parameters of the order statistic filter is introduced. The proposed system is composed of the flat pattern detector, estimator of mean pixel value in flat pattern region, and the adopted order statistic filter. To evaluate the performance of the noise reduction by the proposed approach, several experiments are applied to reduce the impulse noise by this approach. These results show that the MSE by the proposed filter is the least in case that the image is highly corrupted.