Using ImageScaler for resizing images

[maforget]

Just a test that uses ImageScaler for resizing images. Is the performance better? No more annoying crash due to AccessViolationException?

Are the Interpolation Setting correct?

FastAndUgly => NearestNeighbor
FastBilinear => Linear
FastBicubic => Cubic
BilinearHQ => Quadratic
GdiPlus => Hermite
GdiPlusHQ => Lanczos

[github-actions[bot]]

Build Successful! You can find a link to the downloadable artifact below.

Name	Link
Commit	1e676bdfe6a8324f66f3561ae09a96c723f441c2
Logs	https://github.com/maforget/ComicRackCE/actions/runs/9534701062
Download	https://github.com/maforget/ComicRackCE/suites/24954685024/artifacts/1605865074

[maforget]

Some info how MagicScaler resizing relates to GDI Interpolation: https://photosauce.net/blog/post/image-scaling-with-gdi-part-4-examining-the-interpolationmode-values

Usually the usage of each interpolation in ComicRack are:

• Default when not specified is BilinearHQ
• Thumbnail are FastAndUgly (recently) or FastBilinear (was the default)
• Syncing uses GdiPlus
• Exporting uses GDIPlusHQ
• 3D Comic uses GDIPlusHQ
• ComicBookNavgator uses FastBicubic
• Software Filter uses GDIPlusHQ
• Recent Files uses BilinearHQ

Here is the benchmark of ComicRack Resizer vs MagicScaler (when not specified Hybrid is Off, for better comparaison):

So for now I changed some setting based on the blog page (the default when not speficied on the previous settings was FavorQuality):

FastAndUgly => NearestNeighbor, FavorSpeed
FastBilinear => Linear, FavorSpeed
FastBicubic => CatmullRom, FavorSpeed
BilinearHQ => Quadratic, FavorQuality
GdiPlus => Linear, FavorQuality
GdiPlusHQ => Cubic, Off

Some info about HybridMode: https://docs.photosauce.net/api/PhotoSauce.MagicScaler.HybridScaleMode.html#fields

• FavorQuality => Allow lower-quality downscaling to a size at least 3x the target size. Use high-quality scaling to reach the final size.
• FavorSpeed => Allow lower-quality downscaling to a size at least 2x the target size. Use high-quality scaling to reach the final size.
• Turbo => Allow lower-quality downscaling to the nearest power of 2 to the target size. Use high-quality scaling to reach the final size.
• Off => Do not allow hybrid scaling. Use the high-quality scaler exclusively.

[maforget]

After testing, Legacy is still way faster. Even considering the benchmark values. I believe is the multiple conversion from byte[] => Bitmap => byte[] => Bitmap are to blame.

I've started trying to implement a Custom PixelSource instead to pass to MagicScaler, but the current implementation gives corrupted images.