I have written a script using GD::Image that produces a number of small grayscale images (32x32) to use in simple edge detection in a machine learning course. My problem is with the 'ground truth' images. The edges are marked with one single line of pixels, each image rotated in 8 different angles. For some images, the rotated ones are printed with just one pixel in each row, which is how I want it. On others, the line are made of pairs of 2 pixels in each row, making the line double size. Do anyone know why this is, and perhaps how I can fix it?
Examples of the images:
http://home.halden.net/piazava/perlmonks/ramp_4_gt.png
http://home.halden.net/piazava/perlmonks/ramp_4_gt_45.png
http://home.halden.net/piazava/perlmonks/ridge_4_gt.png
http://home.halden.net/piazava/perlmonks/ridge_4_gt_45.png
http://home.halden.net/piazava/perlmonks/roof_0_gt.png
http://home.halden.net/piazava/perlmonks/roof_0_gt_45.png
Thanks for your attention :)In reply to Why does GD draw thicker lines on rotate? by Medri
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |