`blurry`

fast, high peformance image processing libary.

blurry provides image processing algorithms with halide-lang backend.
implements optimized processor for amd64 CPUs on Linux/macos

Benchmarks

Halide JIT benchmarks

This is the result of using halide's benchamrk.
darwin/amd64 Intel(R) Core(TM) i7-8569U CPU @ 2.80GHz

src 320x240
BenchmarkJIT/cloneimg                      : 0.00767ms
BenchmarkJIT/convert_from_argb             : 0.02335ms
BenchmarkJIT/convert_from_abgr             : 0.03712ms
BenchmarkJIT/convert_from_bgra             : 0.02362ms
BenchmarkJIT/convert_from_rabg             : 0.03165ms
BenchmarkJIT/convert_from_yuv_420          : 0.02848ms
BenchmarkJIT/convert_from_yuv_444          : 0.02538ms
BenchmarkJIT/convert_to_yuv_444            : 0.15939ms
BenchmarkJIT/rotate0                       : 0.00776ms
BenchmarkJIT/rotate90                      : 0.02621ms
BenchmarkJIT/rotate180                     : 0.00739ms
BenchmarkJIT/rotate270                     : 0.02624ms
BenchmarkJIT/crop                          : 0.05598ms
BenchmarkJIT/scale                         : 0.04718ms
BenchmarkJIT/scale_box                     : 0.07704ms
BenchmarkJIT/scale_linear                  : 0.07382ms
BenchmarkJIT/scale_gaussian                : 0.09483ms
BenchmarkJIT/blend_normal                  : 0.07561ms
BenchmarkJIT/blend_sub                     : 0.07654ms
BenchmarkJIT/blend_add                     : 0.07503ms
BenchmarkJIT/blend_diff                    : 0.07699ms
BenchmarkJIT/grayscale                     : 0.03702ms
BenchmarkJIT/invert                        : 0.03566ms
BenchmarkJIT/brightness                    : 0.04141ms
BenchmarkJIT/gammacorrection               : 0.07959ms
BenchmarkJIT/contrast                      : 0.01459ms
BenchmarkJIT/boxblur                       : 0.11553ms
BenchmarkJIT/gaussianblur                  : 0.18224ms
BenchmarkJIT/blockmozaic                   : 0.26195ms
BenchmarkJIT/erosion                       : 0.12033ms
BenchmarkJIT/dilation                      : 0.12586ms
BenchmarkJIT/morphology_open               : 0.10494ms
BenchmarkJIT/morphology_close              : 0.10196ms
BenchmarkJIT/morphology_gradient           : 0.08061ms
BenchmarkJIT/emboss$1                      : 0.04667ms
BenchmarkJIT/laplacian                     : 0.03224ms
BenchmarkJIT/highpass                      : 0.03940ms
BenchmarkJIT/gradient                      : 0.03376ms
BenchmarkJIT/edgedetect                    : 0.02602ms
BenchmarkJIT/sobel                         : 0.06334ms
BenchmarkJIT/canny                         : 0.28707ms
BenchmarkJIT/canny_dilate                  : 0.36491ms
BenchmarkJIT/canny_morphology_open         : 0.39379ms
BenchmarkJIT/canny_morphology_close        : 0.39387ms
BenchmarkJIT/match_template_sad            : 5.70232ms
BenchmarkJIT/match_template_ssd            : 4.55996ms
BenchmarkJIT/match_template_ncc            : 8.41311ms
BenchmarkJIT/prepared_match_template_ncc   : 6.15969ms
BenchmarkJIT/match_template_zncc           : 12.15370ms
BenchmarkJIT/prepared_match_template_zncc  : 10.93124ms

AOT benchmarks

Calling a library compiled by AOT(ahead-of-time) via cgo.
In cgo, due to the overhead of ffi calls(e.g.), more complex operations will be optimized for CPU and become faster.
Also, the execution speed may be reduced by the overhead of multiple calls.

Blur

/D is DisablePool, i.e. the benchmark when BufferPool is off.

goos: darwin
goarch: amd64
pkg: github.com/octu0/blurry/benchmark
cpu: Intel(R) Core(TM) i7-8569U CPU @ 2.80GHz
BenchmarkBlur
BenchmarkBlur/bild/blur/Box
BenchmarkBlur/bild/blur/Box-8         	     154	   7812824 ns/op	  640402 B/op	      11 allocs/op
BenchmarkBlur/bild/blur/Gaussian
BenchmarkBlur/bild/blur/Gaussian-8    	     333	   3486751 ns/op	 1262485 B/op	      21 allocs/op
BenchmarkBlur/imaging/Blur
BenchmarkBlur/imaging/Blur-8          	     786	   1520193 ns/op	  793698 B/op	      45 allocs/op
BenchmarkBlur/stackblur-go
BenchmarkBlur/stackblur-go-8          	     231	   5147219 ns/op	  925937 B/op	  153609 allocs/op
BenchmarkBlur/libyuv/ARGBBlur
BenchmarkBlur/libyuv/ARGBBlur-8       	    1861	    642486 ns/op	10182722 B/op	       3 allocs/op
BenchmarkBlur/blurry/Boxblur
BenchmarkBlur/blurry/Boxblur-8        	    7257	    178086 ns/op	      88 B/op	       2 allocs/op
BenchmarkBlur/blurry/Gaussianblur
BenchmarkBlur/blurry/Gaussianblur-8   	    5367	    222615 ns/op	     146 B/op	       2 allocs/op
BenchmarkBlur/blurry/Boxblur/D
BenchmarkBlur/blurry/Boxblur/D-8      	    6093	    201573 ns/op	  311361 B/op	       2 allocs/op
BenchmarkBlur/blurry/Gaussianblur/D
BenchmarkBlur/blurry/Gaussianblur/D-8 	    4629	    257483 ns/op	  311361 B/op	       2 allocs/op

Edge

goos: darwin
goarch: amd64
pkg: github.com/octu0/blurry/benchmark
cpu: Intel(R) Core(TM) i7-8569U CPU @ 2.80GHz
BenchmarkEdge
BenchmarkEdge/bild/EdgeDetection
BenchmarkEdge/bild/EdgeDetection-8         	     643	   1858350 ns/op	  631257 B/op	      10 allocs/op
BenchmarkEdge/blurry/Edge
BenchmarkEdge/blurry/Edge-8                	   10000	    100695 ns/op	  311513 B/op	       3 allocs/op

Rotate

goos: darwin
goarch: amd64
pkg: github.com/octu0/blurry/benchmark
cpu: Intel(R) Core(TM) i7-8569U CPU @ 2.80GHz
BenchmarkRotate
BenchmarkRotate/bild/Rotate/90
BenchmarkRotate/bild/Rotate/90-8         	     612	   2080543 ns/op	 1237046 B/op	  115685 allocs/op
BenchmarkRotate/bild/Rotate/180
BenchmarkRotate/bild/Rotate/180-8        	     480	   2355424 ns/op	 1540311 B/op	  153605 allocs/op
BenchmarkRotate/bild/Rotate/270
BenchmarkRotate/bild/Rotate/270-8        	     520	   2061518 ns/op	 1236932 B/op	  115685 allocs/op
BenchmarkRotate/imaging/90
BenchmarkRotate/imaging/90-8             	    7918	    130736 ns/op	  314181 B/op	       6 allocs/op
BenchmarkRotate/imaging/180
BenchmarkRotate/imaging/180-8            	    9654	    138252 ns/op	  313542 B/op	       6 allocs/op
BenchmarkRotate/imaging/270
BenchmarkRotate/imaging/270-8            	    6972	    163349 ns/op	  314165 B/op	       6 allocs/op
BenchmarkRotate/libyuv/ARGBRotate/90
BenchmarkRotate/libyuv/ARGBRotate/90-8   	   13423	     81131 ns/op	  311360 B/op	       2 allocs/op
BenchmarkRotate/libyuv/ARGBRotate/180
BenchmarkRotate/libyuv/ARGBRotate/180-8  	   34771	     34425 ns/op	  311361 B/op	       2 allocs/op
BenchmarkRotate/libyuv/ARGBRotate/270
BenchmarkRotate/libyuv/ARGBRotate/270-8  	   15904	     78290 ns/op	  311361 B/op	       2 allocs/op
BenchmarkRotate/blurry/Rotate/90
BenchmarkRotate/blurry/Rotate/90-8       	   10000	    109336 ns/op	  311514 B/op	       3 allocs/op
BenchmarkRotate/blurry/Rotate/180
BenchmarkRotate/blurry/Rotate/180-8      	   13102	     89067 ns/op	  311514 B/op	       3 allocs/op
BenchmarkRotate/blurry/Rotate/270
BenchmarkRotate/blurry/Rotate/270-8      	   10000	    124949 ns/op	  311514 B/op	       3 allocs/op

Sobel

goos: darwin
goarch: amd64
pkg: github.com/octu0/blurry/benchmark
cpu: Intel(R) Core(TM) i7-8569U CPU @ 2.80GHz
BenchmarkSobel
BenchmarkSobel/bild/Sobel
BenchmarkSobel/bild/Sobel-8         	     208	   5707148 ns/op	 2196784 B/op	      32 allocs/op
BenchmarkSobel/libyuv/ARGBSobel
BenchmarkSobel/libyuv/ARGBSobel-8   	   16557	     72320 ns/op	  311361 B/op	       2 allocs/op
BenchmarkSobel/blurry/Sobel
BenchmarkSobel/blurry/Sobel-8       	    9255	    140586 ns/op	  311515 B/op	       3 allocs/op

Other Benchmarks

See _benchmark for benchmarks of other methods and performance comparison with libyuv.

Installation

$ go get github.com/octu0/blurry

Examples

original image

Rotate

rotation 0/90/180/270 clockwise

img, err := blurry.Rotate(input, blurry.Rotate90)

`blurry.RotationMode`	Result
`blurry.Rotate90`
`blurry.Rotate180`
`blurry.Rotate270`

Crop

crop x,y with crop_width,crop_height

img, err := blurry.Crop(input, image.Pt(175, 40), crop_width, crop_height)

original	`x=175,y=40,cw=80,ch=50`

Scale

a.k.a. Resize resampling

img, err := blurry.Scale(input, scale_width, scale_height, blurry.ScaleFilterNone)

`blurry.ScaleFilter`	Result
`blurry.ScaleFilterNone`
`blurry.ScaleFilterBox`
`blurry.ScaleFilterLinear`
`blurry.ScaleFilterGaussian`

Grayscale

img, err := blurry.Grayscale(input)

Invert

img, err := blurry.Invert(input)

Brightness

img, err := blurry.Brightness(input, 1.5)

Gamma

img, err := blurry.Gamma(input, 2.5)

Contrast

img, err := blurry.Contrast(input, 0.525)

BoxBlur

img, err := blurry.Boxblur(input, 11)

GaussianBlur

img, err := blurry.Gaussianblur(input, 5.0)

BlockMozaic

img, err := blurry.Blockmozaic(input, 10)

Erode

img, err := blurry.Erosion(input, 5)

Dilate

img, err := blurry.Dilation(input, 8)

Morphology

Morphology repeats Erode and Dilate N times.

size := 5
N := 2
img, err := blurry.Morphology(input, MorphOpen, size, N)

`blurry.MorphologyMode`	Result
`blurry.MorphologyOpen`
`blurry.MorphologyClose`
`blurry.MorphologyGradient`

Emboss

img, err := blurry.Emboss(input)

HighPass

img, err := blurry.Highpass(input)

Laplacian

img, err := blurry.Laplacian(input)

Gradient

img, err := blurry.Gradient(input)

Edge

a.k.a. Edge Detection

img, err := blurry.Edge(input)

Sobel

img, err := blurry.Sobel(input)

Canny

a.k.a. Canny Edge Detection

img, err := blurry.Canny(input, 250, 100)

`max:250 min:100`	`max:400 min:10`

Canny with Dilate

img, err := blurry.CannyWithDilate(input, 250, 100, 3)

`max:250 min:100 dilate:3`	`max:250 min:150 dilate:4`

Morphology Canny with Dilate

Pre-process morphology before applying Canny process.

mode := blurry.CannyMorphologyClose
morph_size := 5
dilate_size := 3
img, err := blurry.MorphologyCannyWithDilate(input, 250, 100, mode, morph_size, dilate_size);

`blurry.CannyMorphologyMode`	Result
`blurry.CannyMorphologyOpen`
`blurry.CannyMorphologyClose`

Template Matching

SAD(Sum of Absolute Difference), SSD(Sum of Squared Difference), NCC(Normalized Cross Correlation) AND ZNCC(Zero means Normalized Cross Correlation) methods are available for template matching.

SAD

scores, err := blurry.MatchTemplateSAD(input, template, 1000)

filter	input	template	Result
`none`
`grayscale`
`sobel`
`canny dilate:3 morph:open`

SSD

scores, err := blurry.MatchTemplateSSD(input, template, 1000)

filter	input	template	Result
`none`
`grayscale`
`sobel`
`canny dilate:3 morph:open`

NCC

scores, err := blurry.MatchTemplateNCC(input, template, 0.1)

filter	input	template	Result
`none`
`grayscale`
`sobel`
`canny dilate:3 morph:open`

Prepared NCC

Improve processing speed by pre-calculating part of NCC process.

p, err := blurry.PrepareNCCTemplate(template)
if err != nil {
  panic(err)
}
defer blurry.FreePreparedNCCTemplate(p)

for _, img := range images {
  scores, err := blurry.PreparedMatchTemplateNCC(img, p, 0.1)
  if err != nil {
    panic(err)
  }
}

ZNCC

scores, err := blurry.MatchTemplateZNCC(input, template, 0.1)

filter	input	template	Result
`none`
`grayscale`
`sobel`
`canny dilate:3 morph:open`

Prepared ZNCC

Improve processing speed by pre-calculating part of ZNCC process.

p, err := blurry.PrepareZNCCTemplate(template)
if err != nil {
  panic(err)
}
defer blurry.FreePreparedZNCCTemplate(p)

for _, img := range images {
  scores, err := blurry.PreparedMatchTemplateZNCC(img, p, 0.1)
  if err != nil {
    panic(err)
  }
}

Blend

Blend input1 on input0.

img, err := blurry.Blend(input0, input1, image.Pt(76, 36), blurry.BlendNormal)

`blurry.BlendMode`	Result
`blurry.BlendNormal`
`blurry.BlendSub`
`blurry.BlendAdd`
`blurry.BlendDiff`

Convert

blurry supports reading ARGB, ABGR, BGRA, YUV420 and YUV444.
It also supports YUV444 output.

Read: RGBA Color Model

img, err := blurry.ConvertFromARGB(input)

ColorModel	Method
ARGB	`blurry.ConvertFromARGB(*image.RGBA)`
ABGR	`blurry.ConvertFromABGR(*image.RGBA)`
BGRA	`blurry.ConvertFromBGRA(*image.RGBA)`
RABG	`blurry.ConvertFromRABG(*image.RGBA)`

Read: YUV Chroma Subsampling

img, err := blurry.ConvertFromYUV420(ycbcr)

Subsampling	Method
420	`blurry.ConvertFromYUV420(*image.YCbCr)`
444	`blurry.ConvertFromYUV444(*image.YCbCr)`

or byte slice can also be specified

var y,u,v []byte
var strideY,strideU,strideV int
var width, height int

img, err := blurry.ConvertFromYUV420Plane(y, u, v, strideY, strideU, strideV, width, height)

Subsampling	Method
420	`blurry.ConvertFromYUV420Plane(y,u,v []byte, int,int,int, w,h int)`
444	`blurry.ConvertFromYUV444Plane(y,u,v []byte, int,int,int, w,h int)`

Write: YUV Chroma Subsampling

ycbcr, err := blurry.ConvertToYUV444(rgba)

Subsampling	Method
444	`blurry.ConvertToYUV420(*image.RGBA)`

CLI usage

Run it via docker.
Use docker run -v to specify where to load the images and where to output them (/tmp will be used as a temporary file).

$ mkdir myimagedir
$ mkdir myimageout
$ cp /from/img/path.png myimagedir/src.png

# grayscale
$ docker run --rm -it \
  -v $PWD/myimagedir:/img \
  -v $PWD/myimageout:/tmp \
  blurry:1.0.0 grayscale -i /img/src.png

Help

NAME:
   blurry

USAGE:
   blurry [global options] command [command options] [arguments...]

VERSION:
   1.18.4

COMMANDS:
     blend
     blockmozaic
     boxblur
     brightness
     canny
     clone
     contrast
     convert
     convert_from_yuv
     convert_to_yuv
     crop
     dilation
     edge
     emboss
     erosion
     gamma
     gaussianblur
     gradient
     grayscale
     highpass
     invert
     laplacian
     morphology
     match_template
     rotate
     scale
     sobel
     help, h           Shows a list of commands or help for one command

GLOBAL OPTIONS:
   --debug, -d    debug mode
   --verbose, -V  verbose. more message
   --help, -h     show help
   --version, -v  print the version

Build

When building, create a docker container with Halide(clang, llvm, etc). installed as the build environment.

$ make build-generator

Compile libruntime.a and all kinds lib*_osx.a or lib*_linu.a to make static link.

$ make generate

Finally, generate a docker image if necessary.

$ make build

Develop

Set up configuration for macos to be able to run image filtering directly through Halide.

setup Halide on local

$ make setup-halide-runtime

generate and run

genrun package allows you to export images to temporary file and run image filtering directly.

$ go run cmd/genrun/main.go benchmark

License

MIT, see LICENSE file for details.

xjtdy888/blurry

blurry

Benchmarks

Halide JIT benchmarks

AOT benchmarks

Blur

Edge

Rotate

Sobel

Other Benchmarks

Installation

Examples

Rotate

Crop

Scale

Grayscale

Invert

Brightness

Gamma

Contrast

BoxBlur

GaussianBlur

BlockMozaic

Erode

Dilate

Morphology

Emboss

HighPass

Laplacian

Gradient

Edge

Sobel

Canny

Canny with Dilate

Morphology Canny with Dilate

Template Matching

SAD

SSD

NCC

Prepared NCC

ZNCC

Prepared ZNCC

Blend

Convert

Read: RGBA Color Model

Read: YUV Chroma Subsampling

Write: YUV Chroma Subsampling

CLI usage

Help

Build

Develop

setup Halide on local

generate and run

License

`blurry`