A fast method for reducing noise in digital color images using anomaly detection and interpolation

The interest of this paper is impulse noise reduction in digital color images. The most popular noise reduction schemes are the vector median filter and its many variants that operate by minimising the aggregate distance from one pixel to every other pixel in a chosen window. The major bottleneck in these methods is the high computational complexity involved due to sequential filtering over the chosen windows. This impedes the use of vector median filters for real time applications despite their superior performance. This paper presents a fast method for reducing impulse noise in digital color images. The key idea here is to consider each row in each color component as a univariate data slice. The impulse noise appears as anomalies at the corrupted pixel indices within the data slice. These indices are identified using thresholding and then the values therein are interpolated across the two pixels left and right of the anomaly(s). This process is performed across all the slices and along all the color components. This ensures that the impulse noise is smoothed out. Since the proposed method smooths out the impulse noise by way of interpolation, it does not result in as much reduction in noise as do the conventional median filters but scales super efficiently in terms of time complexity.