A Dynamic Transform Noise Resistant Uniform Local Binary Pattern (DTNR-ULBP) for Age Classification

Local binary pattern (LBP) captures the local information of a texture effectively and thus plays a dominant role in many image processing applications. A LBP with (8, 1) or (8, 2) derives 2= 256 different patterns, to reduce this dimensionality uniform local binary patterns (ULBP) are derived and it has been proved that they capture fundamental properties of texture. The major disadvantage of LBP is they are prone to noise and this noise may convert a ULBP into Non-ULBP (NULBP) and this degrades the overall performance. To overcome noise problem in NULBP windows the present paper proposes Dynamic transform noise resistant ULBP (DTNR-ULBP). The DTNR-ULBP identifies floating NULBP windows. A NULBP window falls into floating window, if the absolute difference of one or more neighboring pixels with central pixel falls within the range of threshold. A floating window is considered as a noisy window. The DTNRULBP complements one or more floating bits to transforms the floating window into ULBP. The proposed model is dynamic because it transforms one or more floating bits depending on the need instead of all floating bits thus reduces the variation of LBP code. The proposed DTNR-ULBP recovers from the distorted image patterns. The proposed DTNR-ULBP overcomes the noise effect and improves overall age classification rate, and it is tested on FG-NET aging database.