OpenCV-SPM

Use OpenCV in your Swift project in a more elegant way.

This Swift package simplifies the process of importing the prebuilt opencv2.xcframework into your project, eliminating the need for manual building. It monitors release events in the OpenCV Github Project and automatically generates new releases using Github Actions.

Installation

Add https://github.com/yeatse/opencv-spm.git to your package dependencies

!add dependency

Choosing a build

OpenCV ships a lot of modules, and bundling all of them makes the binary large even though most apps only use a fraction. This package therefore offers three prebuilt tiers as separate products. Pick exactly one — they all expose the same opencv2 module, so import opencv2 and your code stay identical no matter which one you choose.

| Product | Modules included | Use when | | --- | --- | --- | | OpenCV | Everything (incl. dnn, ptcloud) | You need deep-learning inference or the full feature set | | OpenCV-NoDNN | Everything except dnn | Classic computer vision without neural-network inference (dnn is by far the largest module) | | OpenCV-Core | core, imgproc, imgcodecs, geometry | Basic image loading and processing only |

If you're unsure, start with OpenCV-NoDNN; move down to OpenCV-Core if you only need image processing, or up to OpenCV if you need dnn.

Note: stitching and highgui have no Objective-C/Swift bindings and are not

callable from Swift in any tier.

Usage

Import opencv2 and use it as documented in opencv.org. For example, a swift version of Extract horizontal and vertical lines by using morphological operations:

import opencv2
// Show source image
let src = Mat(uiImage: image)
// Transform source image to gray if it is not already
let gray: Mat
if (src.channels() == 3) {
    gray = Mat()
    Imgproc.cvtColor(src: src, dst: gray, code: .COLOR_BGR2GRAY)
} else {
    gray = src
}
// Apply adaptiveThreshold at the bitwise_not of gray, notice the ~ symbol
let notGray = Mat()
Core.bitwise_not(src: gray, dst: notGray)
let bw = Mat()
Imgproc.adaptiveThreshold(src: notGray, dst: bw, maxValue: 255, adaptiveMethod: .ADAPTIVE_THRESH_MEAN_C, thresholdType: .THRESH_BINARY, blockSize: 15, C: -2)
// Create the images that will use to extract the horizontal lines
let horizontal = bw.clone()
let vertical = bw.clone()
// Specify size on horizontal axis
let horizontalSize = horizontal.cols() / 30
// Create structure element for extracting horizontal lines through morphology operations
let horizontalStructure = Imgproc.getStructuringElement(shape: .MORPH_RECT, ksize: .init(width: horizontalSize, height: 1))
// Apply morphology operations
Imgproc.erode(src: horizontal, dst: horizontal, kernel: horizontalStructure, anchor: .init(x: -1, y: -1))
Imgproc.dilate(src: horizontal, dst: horizontal, kernel: horizontalStructure, anchor: .init(x: -1, y: -1))
// Specify size on vertical axis
let verticalSize = vertical.rows() / 30
// Create structure element for extracting vertical lines through morphology operations
let verticalStructure = Imgproc.getStructuringElement(shape: .MORPH_RECT, ksize: .init(width: 1, height: verticalSize))
// Apply morphology operations
Imgproc.erode(src: vertical, dst: vertical, kernel: verticalStructure, anchor: .init(x: -1, y: -1))
Imgproc.dilate(src: vertical, dst: vertical, kernel: verticalStructure, anchor: .init(x: -1, y: -1))
// Inverse vertical image
Core.bitwise_not(src: vertical, dst: vertical)
// Extract edges and smooth image according to the logic
// 1. extract edges
// 2. dilate(edges)
// 3. src.copyTo(smooth)
// 4. blur smooth img
// 5. smooth.copyTo(src, edges)
// Step 1
let edges = Mat();
Imgproc.adaptiveThreshold(src: vertical, dst: edges, maxValue: 255, adaptiveMethod: .ADAPTIVE_THRESH_MEAN_C, thresholdType: .THRESH_BINARY, blockSize: 3, C: -2)
// Step 2
let kernel = Mat.ones(rows: 2, cols: 2, type: CvType.CV_8UC1)
Imgproc.dilate(src: edges, dst: edges, kernel: kernel)
// Step 3
let smooth = Mat();
vertical.copy(to: smooth)
// Step 4
Imgproc.blur(src: smooth, dst: smooth, ksize: .init(width: 2, height: 2))
// Step 5
smooth.copy(to: vertical, mask: edges)
// Show final result
let result = vertical.toUIImage()