Repeating Average Filter For Noisy Texture Classification

In this paper, it is shown that repeating average filter increases the uniform patterns of noisy textures and consequently increases the classification accuracy of textures. In other words, for noisy textures,first, an average filter such as 3x3 mean filter is applied to each image then a feature extraction method such as LBP is used to extract features of filtered image. The more value of noise the more repeating ofaverage filter should be applied to textures. It is true that repeating average filter decreases the variance of noisy image. However, in this paper it is shownthat by repeating average filter for textures the variance of texture decreases then increases. So,average filter must be repeated while the variance of image decreases and until the variance is increased, it must be stop. Using convolution to apply average filter for an image takes so much time, therefore a simple technique is proposed in this paper that increases the speed of average filtering significantly.After noise reduction, by using LBP operator,features of texture areextracted for classification. Implementationson Outex, CUReT and UIUC datasets determine that the performance of proposed method is higher than some advanced noise resistant LBP variants such as BRINT and CRLBP.