An efficient sub-pixel level edge detection algorithm for CT brain images is presented in this paper, which is based on Sobel operator, Zernike moments operator, and derived limited non-optimum suppression (LNOS) scheme. Sobel operator is firstly used to extract potential edge points in pixel level, and then Zernike moments operator, together with derived limited non-optimum suppression approach, is utilized to relocate the edges to sub-pixel level. The experiments on CT brain images are conducted to validate the usage of Sobel operator for pixel-level edge operator, and demonstrate that the proposed method is efficient to achieve sub-pixel edge detection for CT brain images, which tends to locate edges more accurately and preserve desired texture details.