A novel discrete symmetry transform (DST) algorithm of location of eyes in facial images was presented based on the value of discrete symmetry. The pixels on regular neighbor zone were removed by detecting the gray uniformity on neighbor zone of each pixel, and the circle mask of symmetry operator was replaced by an annular mask. Better accuracy and shorter computing time could be obtained by using DST than using generalized symmetry transform. Experimental results prove that this algorithm can locate the eyes in facial images automatically without using symmetry angle and geometric feature of human face.