The conventional algorithms for edge-detection are based on the gray gradient in an image

and it is inevitable to get wide edges . Generalized Fuzzy Operator ( GFO ) is a new efficient algorithm for edge-detection [1]. By use of GFO

a high quality edge image with very low noise can be obtained. This paper presents an improved algorithm of GFO

Before the implement of GFO

a Neighborhood-weighted pre-processing is realized to original images

which can help GFO to solve some problems such as discontinuous lines of edges. Our experiments show that this method appears preferable to others.