获取YpCbCr8BiPlanar格式CVPixelBuffer亮度数据时函数崩溃求助

我尝试获取CVPixelBuffer的averageLuminance（平均亮度），但函数在处理若干行后因EXC_BAD_ACCESS错误崩溃。我认为自己对苹果YpCbCr8BiPlanar格式的使用存在误解，以下是我的实现函数，请问问题出在哪里？

func averageLuminance(pixelBuffer: CVPixelBuffer) -> Float? {
    guard CVPixelBufferGetPixelFormatType(pixelBuffer) == kCVPixelFormatType_Lossy_420YpCbCr8BiPlanarFullRange else {
        print("Pixel buffer format mismatch.")
        return nil
    }
    
    CVPixelBufferLockBaseAddress(pixelBuffer, .readOnly)
    defer { CVPixelBufferUnlockBaseAddress(pixelBuffer, .readOnly) }
    
    guard let baseAddress = CVPixelBufferGetBaseAddressOfPlane(pixelBuffer, 0) else {
        print("Unable to get base address.")
        return nil
    }
    
    let lumaWidth = CVPixelBufferGetWidthOfPlane(pixelBuffer, 0)
    let lumaHeight = CVPixelBufferGetHeightOfPlane(pixelBuffer, 0)
    let lumaRowBytes = CVPixelBufferGetBytesPerRowOfPlane(pixelBuffer, 0)
    
    var totalLuminance: Float = 0.0
    
    for row in 0..<lumaHeight {
        let rowBaseAddress = baseAddress.advanced(by: row * lumaRowBytes)
        let rowPointer = rowBaseAddress.assumingMemoryBound(to: UInt8.self)
        
        var rowSum: Float = 0.0
        for col in 0..<lumaWidth {
            rowSum += Float(rowPointer[col])
        }
        
        let rowAverageLuminance = rowSum / Float(lumaWidth)
        totalLuminance += rowAverageLuminance
    }
    
    let averageLuminance = totalLuminance / Float(lumaHeight)
    
    return averageLuminance
}

问题根源 #

你误用了有损压缩的YUV格式kCVPixelFormatType_Lossy_420YpCbCr8BiPlanarFullRange：
这个格式的内存存储的是H.264/HEVC编码后的压缩数据，并非原始的逐像素亮度字节数组。直接把它当成原始Y平面数据访问，必然会超出有效内存范围，触发EXC_BAD_ACCESS错误。

你的内存访问逻辑本身是正确的，但只适用于无损原始YUV格式，比如kCVPixelFormatType_420YpCbCr8BiPlanarFullRange或kCVPixelFormatType_420YpCbCr8BiPlanarVideoRange。

解决方案 #

方案一：切换到无损原始YUV格式 #

修改格式校验逻辑，匹配原始未压缩的平面YUV格式：

func averageLuminance(pixelBuffer: CVPixelBuffer) -> Float? {
    let supportedFormats: [OSType] = [
        kCVPixelFormatType_420YpCbCr8BiPlanarFullRange,
        kCVPixelFormatType_420YpCbCr8BiPlanarVideoRange
    ]
    guard supportedFormats.contains(CVPixelBufferGetPixelFormatType(pixelBuffer)) else {
        print("Pixel buffer format mismatch.")
        return nil
    }
    
    CVPixelBufferLockBaseAddress(pixelBuffer, .readOnly)
    defer { CVPixelBufferUnlockBaseAddress(pixelBuffer, .readOnly) }
    
    guard let baseAddress = CVPixelBufferGetBaseAddressOfPlane(pixelBuffer, 0) else {
        print("Unable to get base address.")
        return nil
    }
    
    let lumaWidth = CVPixelBufferGetWidthOfPlane(pixelBuffer, 0)
    let lumaHeight = CVPixelBufferGetHeightOfPlane(pixelBuffer, 0)
    let lumaRowBytes = CVPixelBufferGetBytesPerRowOfPlane(pixelBuffer, 0)
    
    var totalLuminance: Float = 0.0
    
    for row in 0..<lumaHeight {
        let rowBaseAddress = baseAddress.advanced(by: row * lumaRowBytes)
        let rowPointer = rowBaseAddress.assumingMemoryBound(to: UInt8.self)
        
        var rowSum: Float = 0.0
        for col in 0..<lumaWidth {
            rowSum += Float(rowPointer[col])
        }
        
        totalLuminance += rowSum / Float(lumaWidth)
    }
    
    return totalLuminance / Float(lumaHeight)
}

方案二：解码Lossy格式为原始YUV（若必须处理该格式） #

如果输入的pixelBuffer确实是Lossy格式，需要先用VTDecompressionSession解码压缩数据，得到原始YUV buffer后再计算亮度。核心步骤：

1. 创建视频解码会话并配置参数
2. 将Lossy格式的pixelBuffer送入解码器
3. 获取解码后的原始YUV buffer，再调用上述亮度计算逻辑

性能优化建议 #

双重循环计算效率较低，可以用苹果Accelerate框架的vDSP库加速求和：

import Accelerate

func averageLuminance(pixelBuffer: CVPixelBuffer) -> Float? {
    let supportedFormats: [OSType] = [
        kCVPixelFormatType_420YpCbCr8BiPlanarFullRange,
        kCVPixelFormatType_420YpCbCr8BiPlanarVideoRange
    ]
    guard supportedFormats.contains(CVPixelBufferGetPixelFormatType(pixelBuffer)) else {
        print("Pixel buffer format mismatch.")
        return nil
    }
    
    CVPixelBufferLockBaseAddress(pixelBuffer, .readOnly)
    defer { CVPixelBufferUnlockBaseAddress(pixelBuffer, .readOnly) }
    
    guard let baseAddress = CVPixelBufferGetBaseAddressOfPlane(pixelBuffer, 0) else {
        print("Unable to get base address.")
        return nil
    }
    
    let lumaWidth = CVPixelBufferGetWidthOfPlane(pixelBuffer, 0)
    let lumaHeight = CVPixelBufferGetHeightOfPlane(pixelBuffer, 0)
    let lumaRowBytes = CVPixelBufferGetBytesPerRowOfPlane(pixelBuffer, 0)
    
    var totalSum: Float = 0.0
    let totalPixels = Float(lumaWidth * lumaHeight)
    
    for row in 0..<lumaHeight {
        let rowAddress = baseAddress.advanced(by: row * lumaRowBytes).assumingMemoryBound(to: UInt8.self)
        var rowSum: Float = 0.0
        // 用vDSP快速计算一行的总和
        vDSP_sve(rowAddress, 1, &rowSum, vDSP_Length(lumaWidth))
        totalSum += rowSum
    }
    
    return totalSum / totalPixels
}

内容的提问来源于stack exchange，提问作者houdini w