Histogram-based Method for Restoration of Critical Details in Noise Cleaned Images

摘要

In this paper a method for restoring details, which are lost due to noise cleaning with rank order niters, is discussed. A disadvantage of the rank order niters like the median filter is that particular image details can be misinter-preted as noise and thus be removed. These details usually consist of thin lines, circles, letters etc. and may contain important image information. In this paper three differ-ent approaches for the restoration of these de-tails are discussed. Two are based on the usage of standard image processing methods such as edge and contour detecting filters. With a histogram-based method, only those image points are inspected, which are removed by the noise cleaning filters. For this purpose a dif-ference image is constructed. The points are analysed by calculation and comparison of histograms for each image section. The extracted details are restored in the smoothed image.