The traditional image edge detectors can only get multiple-pixel-wide edges usually, thus it is difficult to operate by post-processing methods. The paper presents a novel edge thinning algorithm named as connectivity preserved edge thinning algorithm, which combines the detecting strategy of the connectivity critical point with the estimating method of the edge direction. It can preserve the origin edge information such as connection and direction. Moreover, it gets ideal or acceptable single-pixel-wide edges with lower computing cost.