YUV4:2:0转RGB后图像偏绿的D3D11渲染问题排查

问题背景 #

解码得到YUV420平面帧（UV分量未打包），使用D3D11渲染时需转换为RGB格式（假设渲染目标视图不支持YUV）。创建3个DXGI_FORMAT_R8_UNORM类型的纹理和ShaderResourceView，将帧的每个平面数据复制到对应视图，依靠采样器处理Y与UV平面的尺寸差异。黑白画面显示正常，但彩色画面严重偏绿。已尝试交换UV平面、调整转换系数，遵循官方图像转换指南，问题仍未解决。

相关代码 #

像素着色器代码 #

min16float4 main(PixelShaderInput input) : SV_TARGET
{
    float y = YChannel.Sample(defaultSampler, input.texCoord).r;
    float u = UChannel.Sample(defaultSampler, input.texCoord).r - 0.5;
    float v = VChannel.Sample(defaultSampler, input.texCoord).r - 0.5;

    float r = y + 1.13983 * v;
    float g = y - 0.39465 * u - 0.58060 * v;
    float b = y + 2.03211 * u;

    return min16float4(r, g, b , 1.f);
}

创建ShaderResourceView代码 #

D3D11_TEXTURE2D_DESC texDesc;
ZeroMemory(&texDesc, sizeof(texDesc));
texDesc.Width = 1670;
texDesc.Height = 626;
texDesc.MipLevels = 1;
texDesc.ArraySize = 1;
texDesc.Format = DXGI_FORMAT_R8_UNORM;
texDesc.SampleDesc.Count = 1;
texDesc.SampleDesc.Quality = 0;
texDesc.Usage = D3D11_USAGE_DYNAMIC;
texDesc.BindFlags = D3D11_BIND_SHADER_RESOURCE;
texDesc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE;


dev->CreateTexture2D(&texDesc, NULL, &pYPictureTexture);
dev->CreateTexture2D(&texDesc, NULL, &pUPictureTexture);
dev->CreateTexture2D(&texDesc, NULL, &pVPictureTexture);

D3D11_SHADER_RESOURCE_VIEW_DESC shaderResourceViewDesc;
shaderResourceViewDesc.Format = DXGI_FORMAT_R8_UNORM;
shaderResourceViewDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2D;
shaderResourceViewDesc.Texture2D.MostDetailedMip = 0;
shaderResourceViewDesc.Texture2D.MipLevels = 1;

dev->CreateShaderResourceView(pYPictureTexture, &shaderResourceViewDesc, &pYPictureTextureResourceView);

dev->CreateShaderResourceView(pUPictureTexture, &shaderResourceViewDesc, &pUPictureTextureResourceView);

dev->CreateShaderResourceView(pVPictureTexture, &shaderResourceViewDesc, &pVPictureTextureResourceView);

复制FFmpeg解码后AVFrames代码 #

int height = 626;
int width = 1670;    

D3D11_MAPPED_SUBRESOURCE msY;
D3D11_MAPPED_SUBRESOURCE msU;
D3D11_MAPPED_SUBRESOURCE msV;
devcon->Map(pYPictureTexture, 0, D3D11_MAP_WRITE_DISCARD, 0, &msY);

memcpy(msY.pData, frame->data[0], height * width);
devcon->Unmap(pYPictureTexture, 0);

devcon->Map(pUPictureTexture, 0, D3D11_MAP_WRITE_DISCARD, 0, &msU);
memcpy(msU.pData, frame->data[1], (height*width) / 4);
devcon->Unmap(pUPictureTexture, 0);


devcon->Map(pVPictureTexture, 0, D3D11_MAP_WRITE_DISCARD, 0, &msV);
memcpy(msV.pData, frame->data[2], (height*width) / 4);
devcon->Unmap(pVPictureTexture, 0);

问题排查方向 #

• UV纹理尺寸错误：YUV420的UV平面宽高应为Y平面的1/2（即Width=835，Height=313），当前UV纹理和Y平面尺寸一致，采样时会重复读取数据导致色彩偏移。
• 行间距（Pitch）未正确处理：FFmpeg的AVFrame->linesize可能大于实际宽度，直接用height*width拷贝会导致数据错位，应使用linesize[0]、linesize[1]、linesize[2]计算拷贝字节数，逐行拷贝而非直接memcpy整个区域。
• 色彩空间系数不匹配：当前转换系数对应BT.601标准，若视频为BT.709格式，需调整系数：

  float r = y + 1.5748 * v;
  float g = y - 0.1873 * u - 0.4681 * v;
  float b = y + 1.8556 * u;
  

• 采样器设置问题：确保采样器使用D3D11_FILTER_MIN_MAG_MIP_LINEAR，地址模式设置合理，避免UV采样时拉伸错误。

内容的提问来源于stack exchange，提问作者luckybroma