Table of Contents

Benchmarks

How fast Egami is next to ImageSharp, SkiaSharp and Magick.NET on the same tasks and inputs. This page has 260 comparisons (each case against each library timed on it). Egami is faster in 234 of them, within 5% in 16, and more than 5% slower in 10, led by lossless WebP at effort 4 on screen content, where SkiaSharp's native libwebp is about 10 times faster. Against ImageSharp, Egami is typically 1.5 to 3 times faster.

Performance has the tips that keep your own code fast and the benchmarks that ship with Egami.

Summary

Each library has its own column with its typical time relative to Egami's on the same task and input (the geometric mean over the cases). Egami is the 1.00× baseline: 2.00× means that library takes twice as long, and below 1.00× it is faster than Egami. An empty cell means that library was not measured for the area. Cases counts the scenarios in the area; a scenario timed against two libraries makes two comparisons. Areas marked Oct 30 were rerun on 30 October 2026 and JPEG encode on 3 November; the rest come from runs between 9 and 22 October.

Area Cases Egami ImageSharp SkiaSharp Magick.NET
JPEG decode (Oct 30) 24 1.00× 1.47× 2.11×
JPEG encode (Nov 3) 4 1.00× 2.45×
JPEG probe, EXIF and thumbnail 6 1.00× 3.03×
PNG decode (Oct 30) 6 1.00× 2.27× 2.09×
PNG probe 6 1.00× 38.09× 1.44×
PNG encode (Oct 30) 8 1.00× 2.48× 2.74×
WebP lossless decode (Oct 30) 2 1.00× 3.50× 1.53×
WebP lossy decode (Oct 30) 2 1.00× 1.61× 1.01×
WebP lossy encode, effort 4 (Oct 30) 2 1.00× 3.48× 0.99×
WebP lossless encode, effort 0 (Oct 30) 2 1.00× 7.34× 1.37×
WebP lossless encode, effort 4 (Oct 30) 2 1.00× 2.03× 0.31×
WebP at matched file size (Oct 30) 4 1.00× 1.62×
Simple formats decode 2 1.00× 1.46×
Simple formats encode (BMP, PNM, QOI, TGA) 18 1.00× 3.25×
GIF and APNG 24 1.00× 2.57×
Reduced-size decode 12 1.00× 1.09×
Operations (resize, blur, rotate, fill and more) 36 1.00× 3.39×
Blend modes (DrawImage) 18 1.00× 1.43× 3.36×
Compositor blend 3 1.00× 1.06×
Scoped operations (rectangle, mask) 4 1.00× 5.40× 35.46×
Pixel conversion 18 1.00× 2.43×
Color space conversion 2 1.00× 1.68×
Channel extract, combine, shuffle 6 1.00× 5.44×
Color keying 10 1.00× 10.82×

Where Egami is slower

The one large gap left is lossless WebP at effort 4 on screen content, about 10× slower than libwebp; at effort 0 Egami matches libwebp there and writes a smaller file. The rest are within 15% or single cases.

Case Egami ImageSharp SkiaSharp
WebP lossless encode, effort 4, screenshot, all cores (Oct 30) 118 ms 208 ms 11.9 ms
Pixel conversion, Rgba8 to GrayAlpha16 8.95 ms 2.80 ms
Reduced-size decode, TGA 32-bit, all cores 8.03 ms 5.91 ms
DrawImage DestinationIn, 1 thread 615.8 µs 524.1 µs 1.22 ms
WebP lossless encode, matched size, screenshot, 1 thread (Oct 30) 14.1 ms 12.0 ms
Reduced-size decode, TGA 32-bit, 1 thread 7.60 ms 6.68 ms
Pixel conversion, Argb8 to Rgba8 97.1 µs 86.2 µs
Reduced-size decode, PNG screenshot, all cores 7.29 ms 6.56 ms
WebP lossless encode, effort 4, photo, all cores (Oct 30) 1,039 ms 2,437 ms 964 ms
WebP lossy decode, photo, all cores (Oct 30) 23.7 ms 40.5 ms 22.1 ms

At 1 thread, lossless WebP at effort 4 on the photo is also about 2.3× slower than libwebp (Egami 2,283 ms).

The JPEG quality 90 case is not like for like: Egami writes optimized Huffman tables there, ImageSharp does not.

How it was measured

The numbers come from a comparison harness kept apart from the Egami repository, so that Egami itself never references the other libraries. Most areas were run between 9 and 22 October 2026 against Egami's main branch on that day. JPEG and PNG were rerun on 30 October against Egami cd0e6b41 (decode, after the speed work in #208) and 35d45808 (encode), and JPEG encode again on 3 November against c7adaa2e (after the encode speed work in #244 to #246); WebP was rerun on 30 October against 6b8b07b9. Compare libraries within one run only: absolute times differ a little between runs.

  • Machine: AMD Ryzen 9 3900X, 12 cores and 24 threads, Windows 10 22H2, .NET 10.0.12 x64 (AVX2).
  • Libraries: ImageSharp 3.1.12, SkiaSharp (native Skia, libpng and libwebp), Magick.NET Q16 (native ImageMagick).
  • Job: BenchmarkDotNet ShortRun: 3 warmups and 3 iterations. Error bars are wide on some rows, so treat differences under about 10% as a tie.
  • Threads: "1 thread" sets Egami's MaxParallelism and ImageSharp's MaxDegreeOfParallelism to 1; "All cores" leaves them at the processor count. SkiaSharp and Magick.NET choose their own threading.
  • Left out: drawing, text, SVG, barcode and video speed are timed separately and are not on this page.

Full results

Every comparison row, codecs first, then operations. Times are BenchmarkDotNet means; the last columns are the other library's time divided by Egami's.

JPEG decode

Rerun Oct 30 against Egami cd0e6b41, the JPEG and PNG speed work merged as #208. Compare libraries within a run only: absolute times differ between runs for all libraries. Quarter and eighth size use each library's reduced decode; SkiaSharp was timed at full size only.

Case File Threads Egami ImageSharp SkiaSharp ImageSharp ÷ Egami SkiaSharp ÷ Egami
Full size 420 All cores 33.83 ms 61.16 ms 93.74 ms 1.81× 2.77×
Full size, fast upsampling 420 All cores 31.53 ms
Quarter size 420 All cores 40.11 ms 45.34 ms 1.13×
Eighth size 420 All cores 29.47 ms 37.85 ms 1.28×
Full size 420 1 thread 48.73 ms 60.86 ms 93.27 ms 1.25× 1.91×
Full size, fast upsampling 420 1 thread 49.72 ms
Quarter size 420 1 thread 39.99 ms 45.86 ms 1.15×
Eighth size 420 1 thread 30.54 ms 37.50 ms 1.23×
Full size 444 All cores 36.85 ms 78.44 ms 106 ms 2.13× 2.88×
Full size, fast upsampling 444 All cores 38.62 ms
Quarter size 444 All cores 47.81 ms 59.70 ms 1.25×
Eighth size 444 All cores 39.80 ms 42.43 ms 1.07×
Full size 444 1 thread 62.58 ms 78.11 ms 107 ms 1.25× 1.71×
Full size, fast upsampling 444 1 thread 60.74 ms
Quarter size 444 1 thread 49.65 ms 61.57 ms 1.24×
Eighth size 444 1 thread 37.87 ms 42.52 ms 1.12×
Full size gray All cores 29.24 ms 48.04 ms 73.14 ms 1.64× 2.50×
Full size, fast upsampling gray All cores 28.09 ms
Quarter size gray All cores 32.88 ms 38.63 ms 1.17×
Eighth size gray All cores 27.73 ms 32.27 ms 1.16×
Full size gray 1 thread 36.22 ms 47.41 ms 71.50 ms 1.31× 1.97×
Full size, fast upsampling gray 1 thread 36.11 ms
Quarter size gray 1 thread 32.50 ms 38.27 ms 1.18×
Eighth size gray 1 thread 27.34 ms 33.87 ms 1.24×
Full size progressive All cores 67.38 ms 148 ms 168 ms 2.20× 2.50×
Full size, fast upsampling progressive All cores 67.27 ms
Quarter size progressive All cores 84.55 ms 128 ms 1.52×
Eighth size progressive All cores 18.32 ms 127 ms 6.95×
Full size progressive 1 thread 136 ms 147 ms 168 ms 1.09× 1.24×
Full size, fast upsampling progressive 1 thread 136 ms
Quarter size progressive 1 thread 130 ms 129 ms 0.99×
Eighth size progressive 1 thread 28.43 ms 119 ms 4.19×

JPEG encode

Rerun Nov 3 against Egami main c7adaa2e, after the JPEG encode speed work (#244 to #246). Quality 90 is timed both plain, like ImageSharp, and with Egami's optimized Huffman tables, which ImageSharp does not offer.

Case Threads Egami ImageSharp ImageSharp ÷ Egami
Quality 75, half chroma All cores 15.96 ms 59.22 ms 3.71×
Quality 90, full chroma All cores 38.08 ms 102 ms 2.67×
Quality 90, full chroma, optimized All cores 50.69 ms
Quality 75, half chroma 1 thread 28.22 ms 57.44 ms 2.04×
Quality 90, full chroma 1 thread 56.06 ms 99.79 ms 1.78×
Quality 90, full chroma, optimized 1 thread 81.45 ms

Against libjpeg-turbo's cjpeg (3.0.3, with its SIMD), egami-compare's 24-megapixel gate encodes on one thread, from a PPM in memory to a null output for both: Egami takes 0.91 to 1.07 times cjpeg's time at quality 75 4:2:0 and 0.55 to 0.58 times at quality 90 4:4:4 with optimized tables (cjpeg -optimize). The files are byte for byte cjpeg's.

JPEG probe, EXIF and thumbnail

Run Oct 9.

Case Threads Egami ImageSharp ImageSharp ÷ Egami
Probe 50 files All cores 390.9 µs 824.3 µs 2.11×
EXIF 50 files All cores 497.2 µs 5.73 ms 11.52×
Thumbnail All cores 32.72 ms 38.71 ms 1.18×
Probe 50 files 1 thread 393.9 µs 832.1 µs 2.11×
EXIF 50 files 1 thread 474.7 µs 5.13 ms 10.80×
Thumbnail 1 thread 32.41 ms 38.53 ms 1.19×

PNG decode and probe

Decode rerun Oct 30 against Egami cd0e6b41 (#208); the probe rows are from Oct 18. Egami's probe reads only the header; ImageSharp's reads further into the file.

Method Scenario Egami ImageSharp SkiaSharp ImageSharp ÷ Egami SkiaSharp ÷ Egami
Decode Adam7-RGBA8 29.15 ms 41.88 ms 58.24 ms 1.44× 2.00×
Decode Gray1 5.45 ms 23.74 ms 20.64 ms 4.36× 3.79×
Decode Palette8 1.26 ms 8.15 ms 2.51 ms 6.48× 1.99×
Decode Photo-RGBA8 71.18 ms 102 ms 103 ms 1.43× 1.45×
Decode RGB16 49.91 ms 74.05 ms 61.60 ms 1.48× 1.23×
Decode Screenshot-RGB8 1.09 ms 1.75 ms 3.33 ms 1.61× 3.06×
Probe (Oct 18) Adam7-RGBA8 2.0 µs 483.7 µs 2.3 µs 241.53× 1.13×
Probe (Oct 18) Gray1 880 ns 2.9 µs 2.1 µs 3.33× 2.36×
Probe (Oct 18) Palette8 2.1 µs 4.0 µs 2.3 µs 1.94× 1.11×
Probe (Oct 18) Photo-RGBA8 1.8 µs 1.67 ms 1.7 µs 939.16× 0.95×
Probe (Oct 18) RGB16 1.6 µs 793.2 µs 1.7 µs 497.18× 1.08×
Probe (Oct 18) Screenshot-RGB8 1.0 µs 4.4 µs 3.0 µs 4.19× 2.89×

PNG encode

Rerun Oct 30 against Egami main 35d45808.

Scenario Level Egami ImageSharp SkiaSharp ImageSharp ÷ Egami SkiaSharp ÷ Egami
Photo-RGBA8 1 305 ms 905 ms 900 ms 2.97× 2.95×
Photo-RGBA8 5 1,118 ms 1,898 ms 1,670 ms 1.70× 1.49×
Photo-RGBA8 6 2,213 ms 4,694 ms 3,679 ms 2.12× 1.66×
Photo-RGBA8 9 3,057 ms 14,567 ms 7,481 ms 4.76× 2.45×
Screenshot-RGB8 1 18.25 ms 33.55 ms 67.46 ms 1.84× 3.70×
Screenshot-RGB8 5 16.31 ms 49.85 ms 84.68 ms 3.06× 5.19×
Screenshot-RGB8 6 16.75 ms 52.64 ms 94.48 ms 3.14× 5.64×
Screenshot-RGB8 9 79.55 ms 127 ms 129 ms 1.59× 1.62×

WebP lossy decode

Rerun Oct 30 against Egami main 6b8b07b9 (ShortRun). All cores; 1 thread was within a few percent.

Picture Egami ImageSharp SkiaSharp ImageSharp ÷ Egami SkiaSharp ÷ Egami
photo 23.7 ms 40.5 ms 22.1 ms 1.71× 0.93×
screenshot 24.2 ms 36.6 ms 26.5 ms 1.51× 1.10×

WebP lossy encode

Rerun Oct 30 against Egami main 6b8b07b9. Quality 75, effort 4, all cores; 1 thread was within a few percent. SkiaSharp runs native libwebp.

Picture Egami ImageSharp SkiaSharp ImageSharp ÷ Egami SkiaSharp ÷ Egami
photo 248 ms 884 ms 241 ms 3.56× 0.97×
screenshot 167 ms 567 ms 169 ms 3.40× 1.01×

WebP lossless encode

Rerun Oct 30 against Egami main 6b8b07b9, all cores. SkiaSharp runs libwebp at its own default, so its time is the same in both effort rows; Egami's effort sets how hard it searches. At 1 thread Egami takes 671 ms (effort 0) and 2,283 ms (effort 4) on the photo and 127 ms on the screenshot at effort 4.

Picture Egami effort Egami ImageSharp SkiaSharp ImageSharp ÷ Egami SkiaSharp ÷ Egami
photo 0 537 ms 1,612 ms 961 ms 3.00× 1.79×
screenshot 0 11.3 ms 203 ms 11.8 ms 17.96× 1.04×
photo 4 1,039 ms 2,437 ms 964 ms 2.35× 0.93×
screenshot 4 118 ms 208 ms 11.9 ms 1.76× 0.10×

WebP lossless decode

Run Oct 30 against Egami main 6b8b07b9, all cores.

Picture Egami ImageSharp SkiaSharp ImageSharp ÷ Egami SkiaSharp ÷ Egami
photo 38.4 ms 95.6 ms 45.5 ms 2.49× 1.18×
screenshot 1.71 ms 8.40 ms 3.37 ms 4.91× 1.97×

WebP at matched file size

Run Oct 30, 1 thread. SkiaSharp (libwebp at its defaults) against Egami at the lowest effort whose file is no larger than libwebp's. ImageSharp is not in this suite.

Case Egami effort Egami size SkiaSharp size Egami SkiaSharp SkiaSharp ÷ Egami
photo, lossless 0 2,474,194 bytes 2,931,846 bytes 683 ms 1,154 ms 1.69×
photo, lossy q75 0 134,692 bytes 192,232 bytes 48.9 ms 245 ms 5.01×
screenshot, lossless 0 2,564 bytes 3,522 bytes 14.1 ms 12.0 ms 0.85×
screenshot, lossy q75 4 366,354 bytes 374,590 bytes 171.5 ms 169.0 ms 0.99×

Simple formats decode

Run Oct 9.

Threads Egami ImageSharp ImageSharp ÷ Egami
All cores 709.5 µs 1.03 ms 1.45×
1 thread 688.3 µs 1.02 ms 1.48×

Simple formats encode (BMP, PNM, QOI, TGA)

Run Oct 9. Times were the same at 1 thread and all cores, so all cores is shown.

Case Egami ImageSharp ImageSharp ÷ Egami
BMP 24-bit 464.6 µs 730.3 µs 1.57×
BMP 32-bit 180.7 µs 937.7 µs 5.19×
TGA 32-bit RLE 3.62 ms 109 ms 29.99×
TGA 32-bit raw 293.7 µs 954.9 µs 3.25×
QOI photo 14.42 ms 63.58 ms 4.41×
QOI screenshot 3.05 ms 14.04 ms 4.61×
PNM P6 196.2 µs 369.3 µs 1.88×
PNM P5 wide 184.8 µs 221.4 µs 1.20×
PNM P3 4.03 ms 4.32 ms 1.07×

GIF and APNG

Run Oct 18.

Case Threads Egami ImageSharp ImageSharp ÷ Egami
GIF decode, animated All cores 38.85 ms 54.56 ms 1.40×
GIF decode, animated 1 thread 38.22 ms 54.72 ms 1.43×
GIF decode, still All cores 12.34 ms 17.36 ms 1.41×
GIF decode, still 1 thread 12.58 ms 17.37 ms 1.38×
GIF decode, first frame All cores 734.6 µs 1.16 ms 1.57×
GIF decode, first frame 1 thread 748.6 µs 1.17 ms 1.57×
GIF encode, photo dithered All cores 98.17 ms 212 ms 2.16×
GIF encode, photo dithered 1 thread 81.64 ms 203 ms 2.49×
GIF encode, photo undithered All cores 46.91 ms 91.07 ms 1.94×
GIF encode, photo undithered 1 thread 44.30 ms 78.00 ms 1.76×
GIF encode, exact colors All cores 54.79 ms 97.90 ms 1.79×
GIF encode, exact colors 1 thread 43.99 ms 88.56 ms 2.01×
GIF encode, animated, global palette All cores 193 ms 342 ms 1.77×
GIF encode, animated, global palette 1 thread 253 ms 343 ms 1.36×
GIF encode, animated, palette per frame All cores 128 ms 2,898 ms 22.73×
GIF encode, animated, palette per frame 1 thread 370 ms 2,888 ms 7.81×
APNG decode All cores 27.28 ms 50.23 ms 1.84×
APNG decode 1 thread 26.11 ms 48.86 ms 1.87×
APNG encode All cores 172 ms 385 ms 2.25×
APNG encode 1 thread 157 ms 298 ms 1.90×
Animated thumbnail All cores 92.64 ms 185 ms 2.00×
Animated thumbnail 1 thread 176 ms 195 ms 1.11×
Probe All cores 162.3 µs 4.40 ms 27.11×
Probe 1 thread 159.0 µs 4.37 ms 27.46×

Reduced-size decode

Run Oct 11.

Scenario Threads Egami ImageSharp ImageSharp ÷ Egami
bmp-24 All cores 5.67 ms 5.38 ms 0.95×
bmp-24 1 thread 5.49 ms 5.55 ms 1.01×
jpeg-photo All cores 41.87 ms 47.98 ms 1.15×
jpeg-photo 1 thread 43.61 ms 55.08 ms 1.26×
png-photo All cores 126 ms 178 ms 1.41×
png-photo 1 thread 126 ms 138 ms 1.09×
png-screenshot All cores 7.29 ms 6.56 ms 0.90×
png-screenshot 1 thread 7.30 ms 7.38 ms 1.01×
qoi-rgb All cores 21.47 ms 32.41 ms 1.51×
qoi-rgb 1 thread 21.13 ms 31.65 ms 1.50×
tga-32 All cores 8.03 ms 5.91 ms 0.74×
tga-32 1 thread 7.60 ms 6.68 ms 0.88×

Operations (resize, blur, rotate, fill and more)

Run Oct 18.

Scenario Threads Egami ImageSharp ImageSharp ÷ Egami
Blur, radius 25 All cores 48.10 ms 79.65 ms 1.66×
Blur, radius 25 1 thread 119 ms 1,183 ms 9.97×
Blur, radius 3 All cores 2.48 ms 9.78 ms 3.94×
Blur, radius 3 1 thread 19.44 ms 280 ms 14.42×
Color chain All cores 4.44 ms 8.84 ms 1.99×
Color chain 1 thread 34.99 ms 80.51 ms 2.30×
DrawImage All cores 106.4 µs 305.3 µs 2.87×
DrawImage 1 thread 362.9 µs 571.2 µs 1.57×
Fill All cores 5.11 ms 20.01 ms 3.92×
Fill 1 thread 15.92 ms 20.77 ms 1.30×
Flip horizontal All cores 2.06 ms 2.57 ms 1.25×
Flip horizontal 1 thread 4.43 ms 4.41 ms 1.00×
Resize, bicubic All cores 3.22 ms 31.80 ms 9.86×
Resize, bicubic 1 thread 23.21 ms 30.76 ms 1.33×
Resize, bilinear All cores 2.75 ms 24.16 ms 8.79×
Resize, bilinear 1 thread 15.43 ms 23.45 ms 1.52×
Resize, box All cores 2.00 ms 21.44 ms 10.73×
Resize, box 1 thread 12.39 ms 20.02 ms 1.62×
Resize, Lanczos3 All cores 4.81 ms 38.51 ms 8.01×
Resize, Lanczos3 1 thread 33.38 ms 40.60 ms 1.22×
Resize, linear light All cores 7.59 ms 96.50 ms 12.72×
Resize, linear light 1 thread 67.99 ms 89.25 ms 1.31×
Resize, Mitchell All cores 3.64 ms 33.80 ms 9.29×
Resize, Mitchell 1 thread 23.99 ms 32.71 ms 1.36×
Resize, nearest All cores 116.1 µs 293.5 µs 2.53×
Resize, nearest 1 thread 446.2 µs 810.3 µs 1.82×
Rotate 15° All cores 4.85 ms 82.14 ms 16.92×
Rotate 15° 1 thread 47.15 ms 1,053 ms 22.33×
Rotate 90° All cores 9.18 ms 20.23 ms 2.20×
Rotate 90° 1 thread 17.64 ms 153 ms 8.68×
Sharpen All cores 1.87 ms 5.01 ms 2.68×
Sharpen 1 thread 14.52 ms 96.69 ms 6.66×
Thumbnail All cores 115 ms 142 ms 1.24×
Thumbnail 1 thread 116 ms 143 ms 1.23×
Upscale All cores 2.46 ms 26.98 ms 10.96×
Upscale 1 thread 16.45 ms 26.32 ms 1.60×

ImageSharp's resize times barely change between 1 thread and all cores in this run, so most of the all-cores resize gap (8 to 13×) comes from Egami's parallel resize.

Blend modes (DrawImage)

Run Oct 11. ImageSharp has no ColorDodge, SoftLight, Hue or Luminosity mode.

Threads Mode Egami ImageSharp SkiaSharp ImageSharp ÷ Egami SkiaSharp ÷ Egami
All cores SourceOver 240.9 µs 302.0 µs 1.25 ms 1.25× 5.17×
All cores DestinationIn 178.1 µs 274.1 µs 1.29 ms 1.54× 7.27×
All cores Multiply 149.7 µs 284.0 µs 1.40 ms 1.90× 9.35×
All cores Screen 154.0 µs 278.3 µs 1.38 ms 1.81× 8.99×
All cores Overlay 165.8 µs 361.3 µs 1.52 ms 2.18× 9.19×
All cores ColorDodge 1.04 ms 2.38 ms 2.30×
All cores SoftLight 400.8 µs 3.15 ms 7.86×
All cores Hue 736.5 µs 3.74 ms 5.08×
All cores Luminosity 580.2 µs 3.39 ms 5.85×
1 thread SourceOver 350.0 µs 612.0 µs 1.03 ms 1.75× 2.93×
1 thread DestinationIn 615.8 µs 524.1 µs 1.22 ms 0.85× 1.97×
1 thread Multiply 580.9 µs 638.4 µs 1.05 ms 1.10× 1.80×
1 thread Screen 576.0 µs 678.2 µs 1.15 ms 1.18× 1.99×
1 thread Overlay 585.3 µs 764.0 µs 1.39 ms 1.31× 2.38×
1 thread ColorDodge 2.98 ms 2.99 ms 1.00×
1 thread SoftLight 2.00 ms 3.70 ms 1.84×
1 thread Hue 2.79 ms 4.19 ms 1.50×
1 thread Luminosity 2.65 ms 3.61 ms 1.36×

Compositor blend

Run Oct 11.

Mode Egami ImageSharp ImageSharp ÷ Egami
SourceOver 8.56 ms 8.79 ms 1.03×
Multiply 8.57 ms 8.78 ms 1.02×
Screen 7.89 ms 8.97 ms 1.14×

Scoped operations (rectangle, mask)

Run Oct 11. Egami (whole image) is Egami's plain run over the full image, for reference. Rectangle blur is compared with ImageSharp, the mask with Magick.NET's write mask.

Scenario Threads Egami Egami (whole image) ImageSharp Magick.NET Other ÷ Egami
Blur inside a rectangle All cores 2.70 ms 19.52 ms 7.35 ms 2.72×
Blur inside a rectangle 1 thread 13.20 ms 123 ms 142 ms 10.72×
Color change inside a mask All cores 15.48 ms 8.97 ms 1,420 ms 91.71×
Color change inside a mask 1 thread 94.28 ms 57.92 ms 1,293 ms 13.71×

Pixel conversion

Run Oct 10. Converting one image to and from Rgba8 (ImageSharp's Rgba32).

Pixel type Direction Egami ImageSharp ImageSharp ÷ Egami
Abgr8 From Rgba8 90.6 µs 96.2 µs 1.06×
Abgr8 To Rgba8 89.8 µs 86.0 µs 0.96×
Alpha8 From Rgba8 471.1 µs 1.66 ms 3.53×
Alpha8 To Rgba8 795.1 µs 6.36 ms 8.00×
Argb8 From Rgba8 89.2 µs 87.9 µs 0.99×
Argb8 To Rgba8 97.1 µs 86.2 µs 0.89×
Bgr565 From Rgba8 429.3 µs 6.46 ms 15.05×
Bgr565 To Rgba8 414.0 µs 7.28 ms 17.58×
Bgra4444 From Rgba8 4.54 ms 7.02 ms 1.55×
Bgra4444 To Rgba8 2.08 ms 7.34 ms 3.53×
GrayAlpha16 From Rgba8 8.95 ms 2.80 ms 0.31×
GrayAlpha16 To Rgba8 1.65 ms 1.59 ms 0.97×
Rgb10A2 From Rgba8 4.52 ms 6.84 ms 1.51×
Rgb10A2 To Rgba8 4.52 ms 9.71 ms 2.15×
Rgb16 From Rgba8 323.4 µs 1.34 ms 4.15×
Rgb16 To Rgba8 819.6 µs 1.83 ms 2.23×
RgbaF16 From Rgba8 1.96 ms 9.93 ms 5.07×
RgbaF16 To Rgba8 2.85 ms 14.58 ms 5.11×

Color space conversion

Run Oct 10.

Case Egami SkiaSharp SkiaSharp ÷ Egami
Display P3 to sRGB 142 ms 255 ms 1.79×
Adobe RGB to sRGB 141 ms 223 ms 1.58×

Channel extract, combine, shuffle

Run Oct 11.

Scenario Threads Egami Magick.NET Magick.NET ÷ Egami
Combine All cores 129 ms 177 ms 1.37×
Combine 1 thread 117 ms 204 ms 1.75×
Extract All cores 79.08 ms 156 ms 1.98×
Extract 1 thread 77.61 ms 161 ms 2.07×
Shuffle All cores 3.34 ms 550 ms 164.63×
Shuffle 1 thread 34.36 ms 548 ms 15.95×

Color keying

Run Oct 11.

Scenario Threads Egami Magick.NET Magick.NET ÷ Egami
Flood fill All cores 232 ms 250 ms 1.08×
Flood fill 1 thread 224 ms 252 ms 1.12×
Replace color All cores 6.70 ms 378 ms 56.40×
Replace color 1 thread 54.39 ms 388 ms 7.14×
Stack mean All cores 13.72 ms 482 ms 35.13×
Stack mean 1 thread 131 ms 438 ms 3.34×
Stack median All cores 28.45 ms 2,537 ms 89.19×
Stack median 1 thread 259 ms 2,506 ms 9.68×
Make color transparent All cores 6.75 ms 409 ms 60.61×
Make color transparent 1 thread 56.23 ms 412 ms 7.33×
Edit this page