iOS猫狗图像分类PoC迁移macOS：NSImage转CVPixelBuffer遇阻

错误原因解析 #

1. CGImage获取方式错误：NSImage没有iOS中UIImage的直接cgImage属性，需通过cgImage(forProposedRect:context:hints:)方法获取，该方法返回可选类型CGImage?，因此必须用可选绑定处理。
2. 坐标系与API差异：macOS的NSImage坐标原点在左下角，与CVPixelBuffer（和iOS一致）的左上角原点不同，直接绘制会导致图像颠倒；另外无需直接访问dataProvider，通过CGContext绘制即可完成像素数据写入。

修复后的完整代码 #

import Cocoa
import CoreVideo
import CoreGraphics

extension NSImage {
    func toCVPixelBuffer() -> CVPixelBuffer? {
        let imageSize = self.size
        let attributes = [
            kCVPixelBufferCGImageCompatibilityKey: kCFBooleanTrue,
            kCVPixelBufferCGBitmapContextCompatibilityKey: kCFBooleanTrue
        ] as CFDictionary
        
        var pixelBuffer: CVPixelBuffer?
        let status = CVPixelBufferCreate(
            kCFAllocatorDefault,
            Int(imageSize.width),
            Int(imageSize.height),
            kCVPixelFormatType_32BGRA,
            attributes,
            &pixelBuffer
        )
        
        guard status == kCVReturnSuccess, let buffer = pixelBuffer else {
            return nil
        }
        
        // 获取NSImage对应的CGImage
        var proposedRect = CGRect(origin: .zero, size: imageSize)
        guard let cgImage = self.cgImage(forProposedRect: &proposedRect, context: nil, hints: nil) else {
            return nil
        }
        
        CVPixelBufferLockBaseAddress(buffer, [])
        defer { CVPixelBufferUnlockBaseAddress(buffer, []) }
        
        // 创建CGContext并处理坐标系翻转
        guard let context = CGContext(
            data: CVPixelBufferGetBaseAddress(buffer),
            width: Int(imageSize.width),
            height: Int(imageSize.height),
            bitsPerComponent: 8,
            bytesPerRow: CVPixelBufferGetBytesPerRow(buffer),
            space: CGColorSpaceCreateDeviceRGB(),
            bitmapInfo: CGImageAlphaInfo.premultipliedFirst.rawValue
        ) else {
            return nil
        }
        
        // 翻转坐标系，匹配CVPixelBuffer的原点规则
        context.translateBy(x: 0, y: imageSize.height)
        context.scaleBy(x: 1.0, y: -1.0)
        
        // 将CGImage绘制到CVPixelBuffer
        context.draw(cgImage, in: CGRect(origin: .zero, size: imageSize))
        
        return buffer
    }
}

关键修改点 #

• CGImage获取：使用cgImage(forProposedRect:context:hints:)方法，通过可选绑定确保获取有效CGImage实例。
• 坐标系适配：通过CGContext的translateBy和scaleBy方法翻转坐标系，避免图像上下颠倒。
• 安全处理：所有可能返回nil的步骤均用guard let校验，保证代码稳定性。

内容的提问来源于stack exchange，提问作者Andrew P.