Some demo results

Upper row from left to right*:

• (1) Original equalized noisy image from hand scanner.
• (2) Fourier Transform (FFT frequency component) image of (1), after log operation for viewing. Strong frequencies can be observed along the horizontal axis corresponding to the virtical bands in the original image.
• (3) Mask created in Microsoft Paint and loaded into Alternate Image in demo program. It consists of 0-valued pixels across the two central rows to blank all vertical bands. All other pixels, along with the four center pixels (the DC component region) are level 255, corresponding to a 1.0 floating point value. The effect is that all components except exclusively horizontal frequencies are masked away.
• (4) Inverse Fourier Transform (IFFT) after Scale By Alternate selected. Note the demo uses only the quantized FFT images (to illustrate that of quantization reduces storage costs at the expense of fidelity), so that some quantization effects are also expected. The VSK itself can hold the FFT components either as a floating point images or as quantized to 8 bits.

Lower row:

• (5) Hough Transform image of (7)
• (6) Warped version of (4).
• (7) Median3x3, Mean3x3, Filter3x3 (a Laplacian), Inverse, Slice and then Median3x3 applied to (4).
• (8) Template matching applied to (7), with the central 16x16 square used as the template. The central pixels should appear brightest, corresponding to the closest match.

* The original images where 256 x 256, while those displayed here are 128 x 128. The demo displays the images with an accurate palette, whereas these images may appear dithered. Also, if the images are not seen in grayscale, try enlarging the viewer to full screen size (to limit palette competition).

Back to Cogimatic home page