如何禁用NSImage的HDR绘制？解决窗口缩放卡顿问题

问题现象 #

• 使用从iPhone获取的常规NSImage，当窗口缩放至小于100px时绘制极度卡顿，CPU占用极高
• 窗口尺寸较大时绘制正常，CPU占用处于低水平
• 已尝试的优化方案均未解决小尺寸场景下的高CPU问题：
  • 使用kCGImageSourceDecodeToSDR解码HDR图像
  • 从CGImage重新创建NSImage
• 非HDR图像在相同小尺寸缩放场景下无绘制卡顿问题
• 补充信息：已知CGContext和NSImageView的存在，问题编号FB22297294（附视频示例、性能截图、相关代码）

可行解决方案 #

1. 预缩放HDR图像至常用小尺寸 #

提前将HDR图像解码并缩放到几个常用的小尺寸（如64px、32px等），窗口缩小时直接调用预缩放后的图像，避免实时缩放带来的高计算开销：

func preScaleHDRImage(_ image: NSImage, targetSizes: [NSSize]) -> [NSSize: NSImage] {
    var scaledImageMap = [NSSize: NSImage]()
    for size in targetSizes {
        let bitmapRep = NSBitmapImageRep(
            bitmapDataPlanes: nil,
            pixelsWide: Int(size.width),
            pixelsHigh: Int(size.height),
            bitsPerSample: 8,
            samplesPerPixel: 4,
            hasAlpha: true,
            isPlanar: false,
            colorSpaceName: .calibratedRGB,
            bytesPerRow: 0,
            bitsPerPixel: 0
        )!
        bitmapRep.size = size
        
        NSGraphicsContext.current = NSGraphicsContext(bitmapImageRep: bitmapRep)
        image.draw(in: NSRect(origin: .zero, size: size))
        NSGraphicsContext.current = nil
        
        let scaledImage = NSImage(size: size)
        scaledImage.addRepresentation(bitmapRep)
        scaledImageMap[size] = scaledImage
    }
    return scaledImageMap
}

2. 临时禁用颜色管理绘制小尺寸图像 #

小尺寸下HDR图像的颜色转换开销占比极高，绘制时临时切换到设备RGB颜色空间，减少颜色管理的计算量：

func drawHDRImageWithoutColorManagement(_ image: NSImage, in rect: NSRect) {
    guard let cgImage = image.cgImage(forProposedRect: nil, context: nil, hints: nil) else { return }
    guard let context = NSGraphicsContext.current?.cgContext else { return }
    
    let originalColorSpace = context.colorSpace
    context.setColorSpace(CGColorSpaceCreateDeviceRGB())
    context.draw(cgImage, in: rect)
    context.setColorSpace(originalColorSpace)
}

3. 用Core Image硬件加速缩放 #

利用Core Image的GPU加速能力处理HDR图像缩放，比CPU端的实时缩放效率更高：

func scaleHDRImageWithCoreImage(_ image: NSImage, targetSize: NSSize) -> NSImage? {
    guard let cgImage = image.cgImage(forProposedRect: nil, context: nil, hints: nil),
          let ciImage = CIImage(cgImage: cgImage) else { return nil }
    
    let scaleX = targetSize.width / image.size.width
    let scaleY = targetSize.height / image.size.height
    let scaledCI = ciImage.transformed(by: CGAffineTransform(scaleX: scaleX, y: scaleY))
    
    let ciContext = CIContext(options: [.useSoftwareRenderer: false])
    guard let scaledCGImage = ciContext.createCGImage(scaledCI, from: scaledCI.extent) else { return nil }
    
    return NSImage(cgImage: scaledCGImage, size: targetSize)
}

内容的提问来源于stack exchange，提问作者Marek H