Study on Extraction of Image Feature Points

Feature points are an important local feature of an image. Due to the poor performance in noise endurance of traditional gray weighted centroid algorithm, an improved algorithm is proposed. By dividing the object region into inner pixel region and edge pixel region, all inner gray pixels are equalized to restrain the noise. At the same time, all feature points are marked by connectivity to improve the velocity of extraction. Finally, simulation experimental results prove that the improved algorithm can extract the feature points effectively with a better noise immunity performance.