In
a group photograph, the computersystem by itselfsegregating theficice image
area fi-oni am, whel image segment is too dflicult. In this article, a novel
method.for detecting humanfacesfroni group images based on skin filtering using
elliptical mode is described Individual candidates are extracted from
thefillered (DCT) and classWed asioce and non-face by, detecting eyes. Found
eye locations are used to re-orient thefaces vertically As distance proportion
between thefbund eyes is use to approximateface region. Discrete cosine
transformation (DCT) is used in-feature extraction. In DCTa subset ofthe transform
coefficients are necessarY to preser most important facial features. A class
(fier is used to ncognizeface features into appropriate classes. This approach
achieves 99% accurac-v inj ace detection in group images and92%, in face
recognition This approach outperforms in various conditions involving/aces with
a wide range offacial variations such its color; position scale, orientation
and complex background
