Читать книгу Rethinking Prototyping - Группа авторов - Страница 119

The bright dots in Fig. 4 (right) represent high magnitudes. That means, these frequencies contribute a lot to the contrast distribution of the image, the darker dots only are responsible for little variations. By applying a rectangular low-pass filter (50 columns wide, 10 rows high or approximately 0.8 % of the original data) and reconstructing the image by only using the values within the red rectangle in Fig. 4 (right) and discarding all the rest, an image containing all the large scale features of the original input image is retrieved (Fig. 5, left). The corresponding high pass filtered image (Fig. 5, center) is globally much more monochromatic and contains all small-scale details such as the cut vessels. A rebuilt image with a low pass filter of the same size but inversed dimensions applied is shown in Fig. 5 (right). The image is much more blurred and contains less relevant details. This is typical for wood cut parallel to the grain. Features like quilt or flame on the other hand are only recognized when including these spectrums as well (Fig. 8). Usually, the term low-pass filter refers to a circular region centered around the origin. Due to the predominantly vertical orientation of most of the wood textures, an asymmetric rectangular region has been chosen. In the custom tool, dimensions and position of the filter can be adjusted freely.

A reduction to the same amount of data points (ca. 500 * 3) but without the prior shift to the frequency domain is shown in Fig. 13. The down sampled 23 by 23 pixels version only vaguely preserves the main traces of the original input image.

Fig. 5 Reconstructed image using a low-pass filter (1st) and the corresponding high-pass filter (2nd) / filter dimensions inversed to ten columns wide and 50 rows high (3rd)

The reconstruction of the image is done by first setting all the values outside of the red rectangle (low pass, inside for the high pass) to zero and then calculating the inverse Fourier transformation (IFFT) of the red, the green and the blue matrix.