6 hqx is a fast, high-quality magnification filter designed for pixel art.
10 NOTE: DevIL library and development headers are required.
15 For more information refer to INSTALL.
19 hqx -s scaleBy input output
20 Where scaleBy is either 2, 3 or 4
23 hqx -s 4 test.png out.png
30 uint32_t *src; // Pointer to source bitmap in RGB format
31 size_t width, height; // Size of source bitmap
34 * Code to init src, width & height
37 uint32_t *dest = (uint32_t *) malloc(width * 4 * height * 4 * sizeof(uint32_t));
39 hq4x_32(src, dest, width, height);
43 The first step is an analysis of the 3x3 area of the source pixel. At first, we
44 calculate the color difference between the central pixel and its 8 nearest
45 neighbors. Then that difference is compared to a predefined threshold, and these
46 pixels are sorted into two categories: "close" and "distant" colored. There are
47 8 neighbors, so we are getting 256 possible combinations.
49 For the next step, which is filtering, a lookup table with 256 entries is used,
50 one entry per each combination of close/distant colored neighbors. Each entry
51 describes how to mix the colors of the source pixels from 3x3 area to get
52 interpolated pixels of the filtered image.
54 The present implementation is using YUV color space to calculate color
55 differences, with more tolerance on Y (brightness) component, then on color
56 components U and V. That color space conversion is quite easy to implement if
57 the format of the source image is 16 bit per pixel, using a simple lookup table.
58 It is also possible to calculate the color differences and compare them to a
59 threshold very fast, using MMX instructions.
61 Creating a lookup table was the most difficult part - for each combination the
62 most probable vector representation of the area has to be determined, with the
63 idea of edges between the different colored areas of the image to be preserved,
64 with the edge direction to be as close to a correct one as possible. That vector
65 representation is then rasterised with higher (3x) resolution using
66 anti-aliasing, and the result is stored in the lookup table.
68 The filter was not designed for photographs, but for images with clear sharp
69 edges, like line graphics or cartoon sprites. It was also designed to be fast
70 enough to process 256x256 images in real-time.