A new four steps edge detection method was proposed.First,the threshold was calculated in natura illumination according to fuzzy logic between sum of the largest differences of gray level values of column pixels and mean gray values of the image.Second,a compromise between even and median filter was introduced to filter noises of image.Meanwhile,one simple approach,which was based on paeth'method,was employed to obtain group of medians in order to decrease the operation time.Third,the gradient between two pixels that have certain distance was calculated in three color components respectively,and then the maximum of gradients of the three was attained.And if absolute value of the maximum gradients exceeds of red,green and blue a threshold that calculated in first step,a candidate for edges is said to exist.Finally,the candidate is converted to edge according to zero-crossing of second derivatives.The algorithm we proposed is applied to images in various weather conditions and times in a day to detect edges,and experimental results indicate that edges detected are continuous and effective.