In this article, image-processing methods to extract thin stains on the surface of industrial products are described. When an image background is noisy, as with satin finished surfaces, it is difficult to extract stains exclusively. In this article, Fourier transform image processing methods to accomplish extraction of thin stains are presented. A low frequency transmitting filter, which has normal distribution transmission characteristics, was applied on a Fourier-transformed plane. Then, experiment to extracted thin stains on satin finished surfaces was performed by changing the bandwidth of the filter. The experiment produced good reconstructed images. It is shown that the light and dark distribution of the images corresponds appropriately to the frequency characteristics of the applied filter.
A torus frequency-transmitting filter to extract thin stains is proposed. Experiment is performed by changing the bandwidth of the torus filter. The torus filter demonstrated the ability to extract thin stains over a range of sizes which correspond to the frequency characteristics of the applied torus filter.
View full abstract