In face recognition, alignment of the face images has been a known open issue. This thesis proposes a displacement based local aligning scheme to construct a structural descriptive image template for comparison. To conquer the registration difficulties caused by the non-rigidity of human face images, a block displacement strategy is introduced to apply the regional voting scheme to face recognition field. Local Binary Pattern (LBP) is adopted to construct this block LBP displacement-based local matching approach, we name LBP-DLMA. Experiments are performed and have demonstrated the outstanding performances of this LBP-DLMA over the original LBP approach. It is expected and shown by experiments that this approach applies to both large and small sized images, and that it also applies to descriptor approaches other than LBP. --Leaf ii.
展开▼
