Data mining based noise-type identification and filtering on road surface image

Noise cancellation is the basic step of road surface image processing, such as road roughness detection, road crack detecting. The 2-D data will always be corrupted by noise in the process of road image sampling and digitalizing. Conventional filtering techniques such as median filter and mean filter are designed to dispel single type of noise. While the image data are always contaminated by mixed noise in the practical situation, because of vibration and disturbance. To solve the problem we develop a data mining approach for noise type identification, and further proposes a fuzzy filter combined with the characteristic parameter in the noise type identification. In the process of noise type identification, we extract some portions of image data by using data mining technique. The speed of the identification can be promoted obviously by these methods. In the step of mixed noise image filtering, the fuzzy median-mean filter proposed in the article will adjust filter parameter adaptively according to noise mixability and perform a favorable effect. From our experimental results, we can draw a conclusion that our proposed fuzzy median-mean filter outperforms existing filters in particular for dealing with images corrupted by Gaussian noise plus salt and pepper noise.