Skip to content

Ezaldeen99/BackgroundRemoval

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

25 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

BackgroundRemoval

Removal results

License Platform Swift

Description

Removal Icon

We depends heavily on Images these days, and sometimes developers need to make processing on these files and one of these things that give us headache is backgorund removal, I really hate that there are so few open source background removal models and libraries so I decided to make this library open source so people can use it, If this suits your need, give it a try 😉

Background removal can give you two different outouts

  • Mask Image: The mask for your input image (black & white) so you can do any processing that you needs
  • Output image: ( default case) your output image without its background

And, it has zero dependencies 😎

Usage

Output image

To start the background removal on your image you just need to pass a UIImage component to the library and get your output image, only that

let image = UIImage(named: "child")
let backgroundRemoval = BackgroundRemoval()
do {
    outputImage.image = try backgroundRemoval.removeBackground(image: image!)
} catch {
    print(error)
}

The output will always be a UIImage in all cases.

Mask Image

In case you wanted only The mask for your input image (black & white), then you case pass maskOnly arg to the library

let image = UIImage(named: "child")
let backgroundRemoval = BackgroundRemoval()
do {
    segmentedImage.image = try backgroundRemoval.removeBackground(image: image!, maskOnly: true)
} catch {
    print(error)
}

Save the Image

Many users have already reported issues while saving the image without background to the device,

the solution below should work.

let resultImage = BackgroundRemoval().removeBackground(image: image!)

let newImage2 = drawImage(resultImage)
let newImage = UIImage(data: newImage2!.pngData()!)

outputImage.image = newImage

where the draw function is

func drawImage(_ image: UIImage) -> UIImage?
    {
        guard let coreImage = image.cgImage else {
            return nil;
        }
        UIGraphicsBeginImageContext(CGSize(width: coreImage.width, height: coreImage.height))
        image.draw(in: CGRect(x: Int(0.0), y: Int(0.0), width: coreImage.width, height: coreImage.height))
        let resultImage = UIGraphicsGetImageFromCurrentImageContext()
        UIGraphicsEndImageContext()
    
        return resultImage;
    }

Improve results

you may see some shadows on the edges after you remove the background, you can add some filters on your mask before you mask the input image to get the output result, here is an example of a good tested workaround that gave me a better resutls

let image = UIImage(named: "child")
let scaledOut = BackgroundRemoval.init().removeBackground(image: image!, maskOnly: true)


/// post processing to get rid of image blur
let imageSource = BBMetalStaticImageSource(image: scaledOut)

// Set up some filters for mask post processing
let contrastFilter = BBMetalContrastFilter(contrast: 3)
let sharpenFilter = BBMetalSharpenFilter(sharpeness: 1)

// Set up filter chain
// Make last filter run synchronously
imageSource.add(consumer: contrastFilter)
 //   .add(consumer: lookupFilter)
    .add(consumer: sharpenFilter)
    .runSynchronously = true

// Start processing
imageSource.transmitTexture()

// Get filtered mask
let filteredImage = sharpenFilter.outputTexture?.bb_image

iOS version

The model supports macOS = 11, iOS = 14. however the library currently supports only iOS systems.

Installation

Background Removal is available through Package Manager. To install it, simply copy the url file and click add package in your xcode and paste it:

https://github.com/Ezaldeen99/BackgroundRemoval.git

Author

Ezaldeen sahb, let us connect https://www.linkedin.com/in/ezzulddin-abdulzahra-17918b137

Contributing

We would love you to contribute to Background Removal, there is still a huge room for improvment.

Credits

The library wouldn't be done without the effort of open source community specially.

License

Project is published under the Apache 2.0 license. Feel free to clone and modify repo as you want, but don't forget to add reference to authors :)