lookup: github.com/deluan/lookup Index | Examples | Files

package lookup

import "github.com/deluan/lookup"

Package lookup implements a nice, simple and fast library which helps you to lookup objects on a screen. It also includes OCR functionality. With Lookup you can do OCR tricks like recognizing any information in a screenshot from your Robot application. Which can be useful for debugging or automating things.

This library is a port of the Java Lookup library (https://gitlab.com/axet/lookup) to GoLang. Details on NCC (Normalized Cross Correlation) used by this library can be found in the original library's docs folder (a lot of math).


// Load full image
img := loadImage("testdata/cyclopst1.png")

// Create a lookup for that image
l := lookup.NewLookup(img)

// Load a template to search inside the image
template := loadImage("testdata/cyclopst3.png")

// Find all occurrences of the template in the image
pp, _ := l.FindAll(template, 0.9)

// Print the results
if len(pp) > 0 {
    fmt.Printf("Found %d matches:\n", len(pp))
    for _, p := range pp {
        fmt.Printf("- (%d, %d) with %f accuracy\n", p.X, p.Y, p.G)
} else {
    println("No matches found")


Found 1 matches:
- (21, 7) with 0.997942 accuracy


// Create an OCR object with an accuracy of 0.7
ocr := lookup.NewOCR(0.7)

// Load a fontSet
_ = ocr.LoadFont("testdata/font_1")

// Load an image to recognize
img := loadImageFromFile("testdata/test3.png")

// Recognize text in image
text, _ := ocr.Recognize(img)

// Print the results
fmt.Printf("Text found in image: %s\n", text)


Text found in image: 3662



Package Files

doc.go fonts.go helpers.go image_binary.go integral_image.go lookup.go ncc.go ocr.go ocr_parallel_find.go

type GPoint Uses

type GPoint struct {
    X, Y int
    G    float64

GPoint represents a match of a template inside an image.

type Lookup Uses

type Lookup struct {
    // contains filtered or unexported fields

Lookup implements a image search algorithm based on Normalized Cross Correlation. For an overview of the algorithm, see http://www.fmwconcepts.com/imagemagick/similar/index.php

func NewLookup Uses

func NewLookup(image image.Image) *Lookup

NewLookup creates a Lookup object. Lookup objects created by these function will do a gray scale search of the templates. Note that if the image or any template is not GrayScale, it will be converted automatically. This could cause some performance hits

func NewLookupColor Uses

func NewLookupColor(image image.Image) *Lookup

NewLookupColor creates a Lookup object that works with all color channels of the image (RGB). Keep in mind that Lookup objects created with this function can only accept color templates as parameters to the FindAll method

func (*Lookup) FindAll Uses

func (l *Lookup) FindAll(template image.Image, threshold float64) ([]GPoint, error)

FindAll searches for all occurrences of template inside the whole image.

func (*Lookup) FindAllInRect Uses

func (l *Lookup) FindAllInRect(template image.Image, rect image.Rectangle, threshold float64) ([]GPoint, error)

FindAllInRect searches for all occurrences of template only inside a part of the image. This can be used to speed up the search if you know the region of the image that the template should appear in.

type OCR Uses

type OCR struct {
    // contains filtered or unexported fields

OCR implements a simple OCR based on the Lookup functions. It allows multiple fontsets, just call LoadFont for each fontset.

If you need to encode special symbols use UNICODE in the file name. For example if you need to have '\' character (which is prohibited in the path and file name) specify %2F.png as a image symbol name.

Sometimes you need to specify two different image for one symbol (if image / font symbol vary too much). To do so add unicode ZERO WIDTH SPACE symbol (%E2%80%8B) to the filename. Ex: %2F%E2%80%8B.png will produce '/' symbol as well.

func NewOCR Uses

func NewOCR(threshold float64, numThreads ...int) *OCR

NewOCR creates a new OCR instance, that will use the given threshold. You can optionally parallelize the processing by specifying the number of threads to use. The optimal number varies and depends on your use case (size of fontset x size of image). Default is use only one thread

func (*OCR) LoadFont Uses

func (o *OCR) LoadFont(fontPath string) error

LoadFont loads a specific fontset from the given folder. Fonts are simple image files containing a PNG/JPEG of the font, and named after the "letter" represented by the image.

This can be called multiple times, with different folders, to load different fontsets.

func (*OCR) Recognize Uses

func (o *OCR) Recognize(img image.Image) (string, error)

Recognize the text in the image using the fontsets previously loaded. If a SubImage is received, the search will be limited by the boundaries of the SubImage

Package lookup imports 11 packages (graph). Updated 2020-03-20. Refresh now. Tools for package owners.