如何从CGImage获取3通道图像数据？解决CGContext创建报错

报错信息：

CGBitmapContextCreate: unsupported parameter combination: 7443 % 4!=0; bytes per row must be a multiple of bytes per pixel

运行代码：

// cgImage is defined
let width = cgImage.width
let height = cgImage.height
let bytesPerPixel = 3 // RGB
let bytesPerRow = bytesPerPixel * width

let rawData = UnsafeMutablePointer<UInt8>.allocate(capacity: height * bytesPerRow)

let bitmapInfo = CGBitmapInfo(rawValue: CGImageAlphaInfo.noneSkipLast.rawValue)
    .union(.byteOrder32Little).rawValue

guard let context = CGContext(data: rawData,
                              width: width,
                              height: height,
                              bitsPerComponent: 8,
                              bytesPerRow: bytesPerRow,
                              space: CGColorSpaceCreateDeviceRGB(),
                              bitmapInfo: bitmapInfo) else {
    fatalError("Unable to create bitmap context.")
}
context.draw(cgImage, in: CGRect(x: 0, y: 0, width: width, height: height))

我原本认为CGImageAlphaInfo.noneSkipLast会忽略Alpha通道，仅记录3字节每像素的RGB数据，但显然它仍尝试按4字节每像素编码。我是否存在误解？该如何从CGImage获取仅3通道的图像数据？

问题原因与解决方法 #

问题原因 #

你确实存在两处关键误解：

1. 字节序标志误用：.byteOrder32Little是针对4字节像素格式（如RGBA）的字节序配置，和3字节RGB格式完全不兼容，搭配后会导致Core Graphics误判像素格式。
2. 内存对齐要求：Core Graphics强制要求位图的每行字节数（bytesPerRow）必须是4的倍数，当width * 3无法被4整除时（比如你的7443），就会触发报错。

解决方法 #

方案1：修正参数，创建合法的3通道RGB上下文

去掉错误的字节序标志，同时将bytesPerRow调整为4的倍数，读取数据时跳过每行末尾的填充字节即可：

// cgImage is defined
let width = cgImage.width
let height = cgImage.height
let bytesPerPixel = 3 // RGB
// 计算对齐到4字节倍数的bytesPerRow
var bytesPerRow = bytesPerPixel * width
bytesPerRow = (bytesPerRow + 3) & ~3 // 向上取整到最近的4的倍数

let rawData = UnsafeMutablePointer<UInt8>.allocate(capacity: height * bytesPerRow)

// 仅使用noneSkipLast，不要搭配byteOrder32Little
let bitmapInfo = CGImageAlphaInfo.noneSkipLast.rawValue

guard let context = CGContext(data: rawData,
                              width: width,
                              height: height,
                              bitsPerComponent: 8,
                              bytesPerRow: bytesPerRow,
                              space: CGColorSpaceCreateDeviceRGB(),
                              bitmapInfo: bitmapInfo) else {
    fatalError("Unable to create bitmap context.")
}
context.draw(cgImage, in: CGRect(x: 0, y: 0, width: width, height: height))

// 读取RGB数据时跳过每行的填充字节
for y in 0..<height {
    let rowStart = rawData + y * bytesPerRow
    for x in 0..<width {
        let pixelOffset = x * bytesPerPixel
        let r = rowStart[pixelOffset]
        let g = rowStart[pixelOffset + 1]
        let b = rowStart[pixelOffset + 2]
        // 处理RGB数据
    }
}

方案2：先创建4通道RGBA上下文，再手动提取RGB通道

如果不想处理内存填充字节，可以先创建带Alpha通道的标准上下文，之后遍历数据时只保留RGB部分：

// cgImage is defined
let width = cgImage.width
let height = cgImage.height
let bytesPerPixel = 4 // RGBA
let bytesPerRow = bytesPerPixel * width

let rawData = UnsafeMutablePointer<UInt8>.allocate(capacity: height * bytesPerRow)

let bitmapInfo = CGImageAlphaInfo.premultipliedLast.rawValue | CGBitmapInfo.byteOrder32Little.rawValue

guard let context = CGContext(data: rawData,
                              width: width,
                              height: height,
                              bitsPerComponent: 8,
                              bytesPerRow: bytesPerRow,
                              space: CGColorSpaceCreateDeviceRGB(),
                              bitmapInfo: bitmapInfo) else {
    fatalError("Unable to create bitmap context.")
}
context.draw(cgImage, in: CGRect(x: 0, y: 0, width: width, height: height))

// 提取纯RGB数据，忽略Alpha通道
let rgbData = UnsafeMutablePointer<UInt8>.allocate(capacity: height * width * 3)
for y in 0..<height {
    let rgbaRowStart = rawData + y * bytesPerRow
    let rgbRowStart = rgbData + y * width * 3
    for x in 0..<width {
        let rgbaOffset = x * bytesPerPixel
        let rgbOffset = x * 3
        rgbRowStart[rgbOffset] = rgbaRowStart[rgbaOffset]     // R
        rgbRowStart[rgbOffset + 1] = rgbaRowStart[rgbaOffset + 1] // G
        rgbRowStart[rgbOffset + 2] = rgbaRowStart[rgbaOffset + 2] // B
    }
}

内容的提问来源于stack exchange，提问作者Soid