版本记录

前言

iOS 11+和macOS 10.13+ 新出了Vision框架，提供了人脸识别、物体检测、物体跟踪等技术，它是基于Core ML的。可以说是人工智能的一部分，接下来几篇我们就详细的解析一下Vision框架。感兴趣的看下面几篇文章。
1. Vision框架详细解析（一） —— 基本概览（一）
2. Vision框架详细解析（二） —— 基于Vision的人脸识别（一）
3. Vision框架详细解析（三） —— 基于Vision的人脸识别（二）
4. Vision框架详细解析（四） —— 在iOS中使用Vision和Metal进行照片堆叠（一）

源码

1. Swift

首先看下工程组织结构

下面看下sb中的内容

接着就看下代码了

1. RecordButton.swift

import UIKit

@IBDesignable
class RecordButton: UIButton {
  var progress: CGFloat = 0.0 {
    didSet {
      DispatchQueue.main.async {
        self.setNeedsDisplay()
      }
    }
  }
  
  override func draw(_ rect: CGRect) {
    // General Declarations
    let context = UIGraphicsGetCurrentContext()!
    
    // Resize to Target Frame
    context.saveGState()    
    
    context.translateBy(x: bounds.minX, y: bounds.minY)
    context.scaleBy(x: bounds.width / 218, y: bounds.height / 218)

    // Color Declarations
    let red = UIColor(red: 0.949, green: 0.212, blue: 0.227, alpha: 1.000)
    let white = UIColor(red: 0.996, green: 1.000, blue: 1.000, alpha: 1.000)
    
    // Variable Declarations
    let expression: CGFloat = -progress * 360
    
    // Button Drawing
    let buttonPath = UIBezierPath(ovalIn: CGRect(x: 26, y: 26, width: 166, height: 166))
    red.setFill()
    buttonPath.fill()
    
    
    // Ring Background Drawing
    let ringBackgroundPath = UIBezierPath(ovalIn: CGRect(x: 8.5, y: 8.5, width: 200, height: 200))
    white.setStroke()
    ringBackgroundPath.lineWidth = 19
    ringBackgroundPath.lineCapStyle = .round
    ringBackgroundPath.stroke()
    
    
    // Progress Ring Drawing
    let progressRingRect = CGRect(x: 8.5, y: 8.5, width: 200, height: 200)
    let progressRingPath = UIBezierPath()
    progressRingPath.addArc(withCenter: CGPoint(x: progressRingRect.midX, y: progressRingRect.midY), radius: progressRingRect.width / 2, startAngle: -90 * CGFloat.pi/180, endAngle: -(expression + 90) * CGFloat.pi/180, clockwise: true)
    
    red.setStroke()
    progressRingPath.lineWidth = 19
    progressRingPath.lineCapStyle = .round
    progressRingPath.stroke()
    
    context.restoreGState()
  }
  
  func resetProgress() {
    progress = 0.0
  }
}

2. AverageStacking.metal

#include <metal_stdlib>
using namespace metal;
#include <CoreImage/CoreImage.h>

extern "C" { namespace coreimage {
  float4 avgStacking(sample_t currentStack, sample_t newImage, float stackCount) {
    float4 avg = ((currentStack * stackCount) + newImage) / (stackCount + 1.0);
    avg = float4(avg.rgb, 1);
    return avg;
  }
}}

3. AverageStackingFilter.swift

import CoreImage

class AverageStackingFilter: CIFilter {
  let kernel: CIBlendKernel
  
  var inputCurrentStack: CIImage?
  var inputNewImage: CIImage?
  var inputStackCount = 1.0
  
  override init() {
    guard let url = Bundle.main.url(forResource: "default", withExtension: "metallib") else {
      fatalError("Check your build settings.")
    }
    
    do {
      let data = try Data(contentsOf: url)
      kernel = try CIBlendKernel(functionName: "avgStacking", fromMetalLibraryData: data)
    } catch {
      print(error.localizedDescription)
      fatalError("Make sure the function names match")
    }
    
    super.init()
  }
    
  required init?(coder aDecoder: NSCoder) {
    fatalError("init(coder:) has not been implemented")
  }

  func outputImage() -> CIImage? {
    guard
      let inputCurrentStack = inputCurrentStack,
      let inputNewImage = inputNewImage
      else {
        return nil
    }
    return kernel.apply(extent: inputCurrentStack.extent, arguments: [inputCurrentStack, inputNewImage, inputStackCount])
  }
}

4. CIImageExtension.swift

import CoreImage

extension CIImage {
  func cgImage() -> CGImage? {
    if cgImage != nil {
      return cgImage
    }
    return CIContext().createCGImage(self, from: extent)
  }
}

5. CameraViewController.swift

import AVFoundation
import UIKit

class CameraViewController: UIViewController {
  @IBOutlet var previewView: UIView!
  @IBOutlet var containerView: UIView!
  @IBOutlet var combinedImageView: UIImageView!
  @IBOutlet var recordButton: RecordButton!
  var previewLayer: AVCaptureVideoPreviewLayer!
  let session = AVCaptureSession()
  var saver: ImageSaver?
  let imageProcessor = ImageProcessor()
  var isRecording = false
  let maxFrameCount = 20
  
  override func viewDidLoad() {
    super.viewDidLoad()
    containerView.isHidden = true
    configureCaptureSession()
    session.startRunning()
  }
}

// MARK: - Configuration Methods

extension CameraViewController {
  func configureCaptureSession() {
    guard let camera = AVCaptureDevice.default(for: .video) else {
      fatalError("No video camera available")
    }
    do {
      let cameraInput = try AVCaptureDeviceInput(device: camera)
      session.addInput(cameraInput)
      try camera.lockForConfiguration()
      camera.activeVideoMinFrameDuration = CMTime(value: 1, timescale: 5)
      camera.activeVideoMaxFrameDuration = CMTime(value: 1, timescale: 5)
      camera.unlockForConfiguration()
    } catch {
      fatalError(error.localizedDescription)
    }
    // Define where the video output should go
    let videoOutput = AVCaptureVideoDataOutput()
    videoOutput.setSampleBufferDelegate(self, queue: DispatchQueue(label: "video data queue"))
    //videoOutput.alwaysDiscardsLateVideoFrames = true
    videoOutput.videoSettings = [kCVPixelBufferPixelFormatTypeKey as String: kCVPixelFormatType_32BGRA]
    // Add the video output to the capture session
    session.addOutput(videoOutput)
    let videoConnection = videoOutput.connection(with: .video)
    videoConnection?.videoOrientation = .portrait
    // Configure the preview layer
    previewLayer = AVCaptureVideoPreviewLayer(session: session)
    previewLayer.videoGravity = .resizeAspectFill
    previewLayer.frame = view.bounds
    previewView.layer.addSublayer(previewLayer)
  }
}

// MARK: - UI Methods

extension CameraViewController {
  @IBAction func recordTapped(_ sender: UIButton) {
    recordButton.isEnabled = false
    isRecording = true
    saver = ImageSaver()
  }
  
  @IBAction func closeButtonTapped(_ sender: UIButton) {
    containerView.isHidden = true
    recordButton.isEnabled = true
    session.startRunning()
  }

  func stopRecording() {
    isRecording = false
    recordButton.progress = 0.0
  }
  
  func displayCombinedImage(_ image: CIImage) {
    session.stopRunning()
    combinedImageView.image = UIImage(ciImage: image)
    containerView.isHidden = false
  }
}

// MARK: - Capture Video Data Delegate Methods

extension CameraViewController: AVCaptureVideoDataOutputSampleBufferDelegate {
  func captureOutput(_ output: AVCaptureOutput, didOutput sampleBuffer: CMSampleBuffer, from connection: AVCaptureConnection) {
    if !isRecording {
      return
    }
    guard
      let imageBuffer = CMSampleBufferGetImageBuffer(sampleBuffer),
      let cgImage = CIImage(cvImageBuffer: imageBuffer).cgImage()
      else {
        return
    }
    let image = CIImage(cgImage: cgImage)
    imageProcessor.add(image)
    saver?.write(image)
    let currentFrame = recordButton.progress * CGFloat(maxFrameCount)
    recordButton.progress = (currentFrame + 1.0) / CGFloat(maxFrameCount)
    if recordButton.progress >= 1.0 {
      stopRecording()
      imageProcessor.processFrames(completion: displayCombinedImage)
    }
  }
}

6. ImageProcessor.swift

import CoreImage
import Vision

class ImageProcessor {
  var frameBuffer: [CIImage] = []
  var alignedFrameBuffer: [CIImage] = []
  var completion: ((CIImage) -> Void)?
  var isProcessingFrames = false
  var frameCount: Int {
    return frameBuffer.count
  }
  
  func add(_ frame: CIImage) {
    if isProcessingFrames {
      return
    }
    frameBuffer.append(frame)
  }
  
  func processFrames(completion: ((CIImage) -> Void)?) {
    isProcessingFrames = true
    self.completion = completion
    let firstFrame = frameBuffer.removeFirst()
    alignedFrameBuffer.append(firstFrame)
    for frame in frameBuffer {
      let request = VNTranslationalImageRegistrationRequest(targetedCIImage: frame)
      do {
        let sequenceHandler = VNSequenceRequestHandler()
        try sequenceHandler.perform([request], on: firstFrame)
      } catch {
        print(error.localizedDescription)
      }
      alignImages(request: request, frame: frame)
    }
    combineFrames()
  }
  
  func alignImages(request: VNRequest, frame: CIImage) {
    guard
      let results = request.results as? [VNImageTranslationAlignmentObservation],
      let result = results.first
      else {
        return
    }
    let alignedFrame = frame.transformed(by: result.alignmentTransform)
    alignedFrameBuffer.append(alignedFrame)
  }
  
  func combineFrames() {
    var finalImage = alignedFrameBuffer.removeFirst()
    let filter = AverageStackingFilter()
    for (i, image) in alignedFrameBuffer.enumerated() {
      filter.inputCurrentStack = finalImage
      filter.inputNewImage = image
      filter.inputStackCount = Double(i + 1)
      finalImage = filter.outputImage()!
    }
    cleanup(image: finalImage)
  }
  
  func cleanup(image: CIImage) {
    frameBuffer = []
    alignedFrameBuffer = []
    isProcessingFrames = false
    if let completion = completion {
      DispatchQueue.main.async {
        completion(image)
      }
    }
    completion = nil
  }
}

7. ImageSaver.swift

import CoreImage

struct ImageSaver {
  var count = 0
  let url: URL
  
  init() {
    let uuid = UUID().uuidString
    let urls = FileManager.default.urls(for: .documentDirectory, in: .userDomainMask)
    url = urls[0].appendingPathComponent(uuid)
    try? FileManager.default.createDirectory(at: url, withIntermediateDirectories: false, attributes: nil)
  }
  
  mutating func write(_ image: CIImage, as name: String? = nil) {
    guard let colorSpace = CGColorSpace(name: CGColorSpace.sRGB) else {
      return
    }
    let context = CIContext()
    let lossyOption = kCGImageDestinationLossyCompressionQuality as CIImageRepresentationOption
    let imgURL: URL
    if let name = name {
      imgURL = url.appendingPathComponent("\(name).jpg")
    } else {
      imgURL = url.appendingPathComponent("\(count).jpg")
    }
    try? context.writeJPEGRepresentation(of: image,
                                         to: imgURL,
                                         colorSpace: colorSpace,
                                         options: [lossyOption: 0.9])
    count += 1
  }
}

后记

本篇主要讲述了在iOS中使用Vision和Metal进行照片堆叠，感兴趣的给个赞或者关注~~~