| 摘要 |
Present invention discloses a saliency extraction method for an image which comprises conducting permeability function (PF(I(x, y)) of the image by conducting horizontal update operations (LRsum(x, y), RLsum(x, y)) and vertical update operations (TBsum(x, y), BTsum(x, y)) for each pixel of concern (P(x, y)) of said image, for different values of a permeability adjustment constant (Ã) and obtaining saliency map (SM(x, y)) of said image by summing centre - surround differences (CSD(x, y)), which are calculated by differencing normalized permeability functions (I F (x, y)) for different values of said permeability adjustment constant (Ã 1 , Ã 2 , Ã 3 , ... , Ã N ). The method enables extraction of a saliency map with high accuracy which preserves edge transitions. |
