The edge’s orientation obtained with the method of the gray gradient orientation is quite different from the real edge’s orientation. Therefore

it causes the deterioration of the high pass filtering performance

if we place the filter according to the estimated angle obtained with the gray gradient orientation. The influences of the angle of edge’s orientation and filter’s long axis on the filtering performance have been studied in this paper. Based on the analysis of the curve and combined with a coordinate transformation

an improved algorithm on the method of the gray gradient orientation is proposed. Because of the algorithm making full use of the pixels in the region for a better estimation

then the optimized estimated angle is selected

which is closer to that of the real edge’s orientation. If the orientation of the filter’s long axis is strictly consistent with that of the real edges

the filtering performance and its material experimental data are regarded as a performance criterion. The experimental result shows that using the algorithm

both the filtering performance and its material experimental data are close better to the performance criterion mentioned above.