3 ---------------------------------------------------------------------
5 wave or noise like patterns around sharp edges
6 bad: sinc, lanczos (with high filter length)
7 acceptable: lanczos (with low filter length), cubic, spline
8 ok: area, (fast)blinear, gauss, point
11 loss of detail / high frequency
12 bad: gauss with high variance
13 acceptable: (fast)bilinear, area
16 aliasing (only downscale)
17 straight lines look like staircases
18 areas of high detail look worse
19 regular patterns turn into a funny looking mess (moire patterns)
20 bad: fast bilinear, point, gauss with low variance
25 looks like the image is made of rectangular blocks like a mosaic
29 recommendations: (based upon my personal opinion many ppl might disagree ...)
30 ---------------------------------------------------------------------
32 the recommended scalers for upscaling:
33 fast_bilinear, point if speed is important
34 cubic, spline, lanczos if quality is important
36 the recommended scalers for downscaling:
37 fast_bilinear, point if speed is important
38 gauss, bilinear if quality is important
39 cubic, spline, lanczos if a sharper picture is important
41 note: when encoding at a limited amount of bits (not constant quantizer)
42 then a slightly blurred input might look better after decoding than a slightly
43 sharpened one, especially for lower bitrates
45 sws numbers / names / artifact types
46 -------------------------------------------------------------------------
49 a aliasing (downscale only)
50 m mosaic (blocky) (upscale only)
52 -sws 0 fast bilinear bA
54 -sws 2 bicubic high sharpness r
56 -sws 3 experimental ????
57 -sws 4 nearest neighbour AM
59 -sws 6 luma bicubic / chroma bilinear r
60 -sws 7 gauss low sharpness B
63 -sws 9 lanczos long filter R
65 -sws 10 bicubic spline r
68 area upscale is identical to bilinear