Face Recognition Based On Multi-feature Fusion Of Local Directional Pattern

Because face recognition is friendly and non-invasive,face recognition had always been a hot research area for researchers.At present,face recognition technology is relatively mature,and has been widely used in finance,transportation,security,medical care,education and other fields.The face recognition feature extraction algorithm are constantly improving from global features to local features.For local feature extraction,the operator of Local Binary Pattern feature extraction is good and simple.As a modified Local Directional Pattern(LDP)operator,it not only inherits all the advantages of the algorithm of Local Binary Pattern,but also is more robust to illumination and noise.Based on the LDP operator,the paper proposes several face recognition algorithms with multi-feature fusion.The paper is mainly introduced from the following aspects:(1)A face feature extraction algorithm based on Double Operation Local Directional Pattern is proposed.Firstly,the preprocessed face image is extracted from the 3×3 neighborhood pixel block,and convolved with 8 different directions and fixed template radius of the Kirsch operator to obtain the corresponding 3×3 size edge response value.Then,the neighboring edge response values are countered and summed in counterclockwise directions to obtain two sets of eight-direction edge response differences and sums,and the two sets of values are taken as absolute values.Finally,the directions of the maximum values of the two sets of edge response values are encoded into a two-digit octal number to form a DOLDP code.The experimental results on YALE,ORL,AR and CAS-PEAL face database show that the proposed method combines the sum space and the difference space face feature information,and achieves a better recognition effect.Compared with the intensity space,the sum space face feature information plays a smooth role and shows stronger robustness to light,expression,occlusion and noise.(2)A multi-level Local Directional Pattern face recognition algorithm is proposed.Firstly,Firstly,the preprocessed face image is extracted from the 3×3neighborhood pixel block,and first convolved with 8 different directions and fixed template radius of the Kirsch operator to obtain the corresponding 3×3 size edge response value,and then the edge response value and 8 Kirsch templates are calculated to obtain another edge response values.Then,the two sets of edge responsevalues are encoded according to the point size of the central pixel,and an 8-bit binary is formed.Forming an MLLDP code,The uniform invariant mode is used to reduce the number of modes to 59.The experimental results on YALE and ORL face database show that this method achieves better recognition results by multi-level face features through the way of operator fusion without increasing the feature dimension.