Image Retrieval based on LBP Transitions

One of the current theoretically significant, simple and very effective texture descriptor that describe local structure efficiently and precisely is the 'Local Binary Pattern' (LBP). Today LBP and its variants are applied in many areas. One of the disadvantage with LBP is it derives a total of 256 patterns out of which 58 are the Uniform LBP (ULBP) and remaining are Non Uniform LBP (NULBP). The ULBP holds the fundamental characteristic and most of the textures predominantly contain ULBP . The disadvantage with ULBP is one should consider 58 pattern features for any classification or retrieval etc. The ULBP approaches completely ignored the NULBP and grouped them into mislenious class. This leads to lot of complexity. To overcome this, present paper designed a new method for retrieval based on histogram of transitions from 0 to 1 or 1 to 0 on LBP. LBP contains only 5 such transitions (0 or 2 or 4 or 6 or 8). The proposed method is experimented on various images collected from Google data base. The experimental result indicates the efficiency of the proposed method over the various methods.