Create CVPixelBuffer from YUV with IOSurface backed

Question

So I am getting raw YUV data in 3 separate arrays from a network callback (voip app). From what I understand you cannot create IOSurface backed pixel buffers with CVPixelBufferCreateWithPlanarBytes according to here

Important: You cannot use CVPixelBufferCreateWithBytes() or CVPixelBufferCreateWithPlanarBytes() with kCVPixelBufferIOSurfacePropertiesKey. Calling CVPixelBufferCreateWithBytes() or CVPixelBufferCreateWithPlanarBytes() will result in CVPixelBuffers that are not IOSurface-backed

So thus you have to create it with CVPixelBufferCreate, but how do you transfer the data from the call back to the CVPixelBufferRef that you create?

- (void)videoCallBack(uint8_t *yPlane, uint8_t *uPlane, uint8_t *vPlane, size_t width, size_t height, size_t stride yStride,
                      size_t uStride, size_t vStride)
    NSDictionary *pixelAttributes = @{(id)kCVPixelBufferIOSurfacePropertiesKey : @{}};
    CVPixelBufferRef pixelBuffer = NULL;
    CVReturn result = CVPixelBufferCreate(kCFAllocatorDefault,
                                          width,
                                          height,
                                          kCVPixelFormatType_420YpCbCr8BiPlanarVideoRange,
                                          (__bridge CFDictionaryRef)(pixelAttributes),
                                          &pixelBuffer);

I am unsure what to do afterwards here? Eventually I want to turn this into a CIImage which then I can use my GLKView to render the video. How do people "put" the data into the buffers from when you create it?

cvu · Accepted Answer · 2015-08-05T19:54:47.033

10

I figured it out and it was fairly trivial. Here is the full code below. Only issue is that I get a BSXPCMessage received error for message: Connection interrupted and it takes a while for the video to show.

NSDictionary *pixelAttributes = @{(id)kCVPixelBufferIOSurfacePropertiesKey : @{}};
CVPixelBufferRef pixelBuffer = NULL;
CVReturn result = CVPixelBufferCreate(kCFAllocatorDefault,
                                      width,
                                      height,
                                      kCVPixelFormatType_420YpCbCr8BiPlanarVideoRange,
                                      (__bridge CFDictionaryRef)(pixelAttributes),
                                      &pixelBuffer);

CVPixelBufferLockBaseAddress(pixelBuffer, 0);
uint8_t *yDestPlane = CVPixelBufferGetBaseAddressOfPlane(pixelBuffer, 0);
memcpy(yDestPlane, yPlane, width * height);
uint8_t *uvDestPlane = CVPixelBufferGetBaseAddressOfPlane(pixelBuffer, 1);
memcpy(uvDestPlane, uvPlane, numberOfElementsForChroma);
CVPixelBufferUnlockBaseAddress(pixelBuffer, 0);

if (result != kCVReturnSuccess) {
    DDLogWarn(@"Unable to create cvpixelbuffer %d", result);
}

CIImage *coreImage = [CIImage imageWithCVPixelBuffer:pixelBuffer]; //success!
CVPixelBufferRelease(pixelBuffer);

I forgot to add the code to interleave the two U and V planes, but that shouldn't be too bad.

edited Aug 05 '15 at 19:54

answered Aug 05 '15 at 18:15

cvu

482
2
6
20

1

Please provide code for U and V planes and numberOfElementsForChroma – JULIIncognito May 31 '16 at 23:11
@JULIIncognito The uvPlane is just an array of uint_8 I believe. This is so long ago sorry. numberOfElementsForChroma will just be whatever fixed size you need. – cvu Jun 01 '16 at 16:01
@JULIIncognito Is numberOfElementsForChroma same of width * height ? Because It's not working. – lilouch Jul 21 '16 at 08:48
@lilouch It was not working for me too. So, I`ve just changed color format to the kCVPixelFormatType_32BGRA and did another transformation – JULIIncognito Jul 27 '16 at 16:16
For Y420 or NV12, numberOfElementsForChroma is width*height/2. Because both U plane and V plane are quarter size of Y plane. – ooOlly Aug 31 '18 at 12:33

score 4 · Answer 2 · edited Sep 17 '21 at 14:24

Here is the full conversion in obj-c. Also for all those geniuses who say : "it's trivial" don't patronize anyone! if you are here to help, help, if you are here to show how "smart" you are, go do it somewhere else. Here is the link to detailed explanation on YUV processing : www.glebsoft.com

    /// method to convert YUV buffers to pixelBuffer in otder to feed it to face unity methods
-(CVPixelBufferRef*)pixelBufferFromYUV:(uint8_t *)yBuffer vBuffer:(uint8_t *)uBuffer uBuffer:(uint8_t *)vBuffer width:(int)width height:(int)height  {
    NSDictionary *pixelAttributes = @{(id)kCVPixelBufferIOSurfacePropertiesKey : @{}};
    CVPixelBufferRef pixelBuffer;
    /// NumberOfElementsForChroma is width*height/4 because both U plane and V plane are quarter size of Y plane
    CGFloat uPlaneSize =  width * height / 4;
    CGFloat vPlaneSize = width * height / 4;
    CGFloat numberOfElementsForChroma = uPlaneSize + vPlaneSize;

    CVReturn result = CVPixelBufferCreate(kCFAllocatorDefault,
                                          width,
                                          height,
                                          kCVPixelFormatType_420YpCbCr8BiPlanarVideoRange,
                                          (__bridge CFDictionaryRef)(pixelAttributes),
                                          &pixelBuffer);

    ///for simplicity and speed create a combined UV panel to hold the pixels
    uint8_t *uvPlane = calloc(numberOfElementsForChroma, sizeof(uint8_t));
    memcpy(uvPlane, uBuffer, uPlaneSize);
    memcpy(uvPlane += (uint8_t)(uPlaneSize), vBuffer, vPlaneSize);
    
    
    CVPixelBufferLockBaseAddress(pixelBuffer, 0);
    uint8_t *yDestPlane = CVPixelBufferGetBaseAddressOfPlane(pixelBuffer, 0);
    memcpy(yDestPlane, yBuffer, width * height);
    
    uint8_t *uvDestPlane = CVPixelBufferGetBaseAddressOfPlane(pixelBuffer, 1);
    memcpy(uvDestPlane, uvPlane, numberOfElementsForChroma);
    CVPixelBufferUnlockBaseAddress(pixelBuffer, 0);
    CVPixelBufferRelease(pixelBuffer);
    free(uvPlane);
    return pixelBuffer;
}

score 3 · Answer 3 · edited May 23 '17 at 11:54

I had a similar question and here is what I have in SWIFT 2.0 with informations that I got from answers to others questions or links.

func generatePixelBufferFromYUV2(inout yuvFrame: YUVFrame) -> CVPixelBufferRef?
{
    var uIndex: Int
    var vIndex: Int
    var uvDataIndex: Int
    var pixelBuffer: CVPixelBufferRef? = nil
    var err: CVReturn;

    if (m_pixelBuffer == nil)
    {
        err = CVPixelBufferCreate(kCFAllocatorDefault, yuvFrame.width, yuvFrame.height, kCVPixelFormatType_420YpCbCr8BiPlanarVideoRange, nil, &pixelBuffer)
        if (err != 0) {
            NSLog("Error at CVPixelBufferCreate %d", err)
            return nil
        }
    }

    if (pixelBuffer != nil)
    {
        CVPixelBufferLockBaseAddress(pixelBuffer!, 0)
        let yBaseAddress = CVPixelBufferGetBaseAddressOfPlane(pixelBuffer!, 0)
        if (yBaseAddress != nil)
        {
            let yData = UnsafeMutablePointer<UInt8>(yBaseAddress)
            let yDataPtr = UnsafePointer<UInt8>(yuvFrame.luma.bytes)

            // Y-plane data
            memcpy(yData, yDataPtr, yuvFrame.luma.length)
        }

        let uvBaseAddress = CVPixelBufferGetBaseAddressOfPlane(m_pixelBuffer!, 1)
        if (uvBaseAddress != nil)
        {
            let uvData = UnsafeMutablePointer<UInt8>(uvBaseAddress)
            let pUPointer = UnsafePointer<UInt8>(yuvFrame.chromaB.bytes)
            let pVPointer = UnsafePointer<UInt8>(yuvFrame.chromaR.bytes)

            // For the uv data, we need to interleave them as uvuvuvuv....
            let iuvRow = (yuvFrame.chromaB.length*2/yuvFrame.width)
            let iHalfWidth = yuvFrame.width/2

            for i in 0..<iuvRow
            {
                for j in 0..<(iHalfWidth)
                {
                    // UV data for original frame.  Just interleave them.
                    uvDataIndex = i*iHalfWidth+j
                    uIndex = (i*yuvFrame.width) + (j*2)
                    vIndex = uIndex + 1
                    uvData[uIndex] = pUPointer[uvDataIndex]
                    uvData[vIndex] = pVPointer[uvDataIndex]
                }
            }
        }
        CVPixelBufferUnlockBaseAddress(pixelBuffer!, 0)
    }

    return pixelBuffer
}

Note: yuvFrame is a structure with y, u, and v plan buffers and width and height. Also, I have the CFDictionary? parameter in the CVPixelBufferCreate(...) set to nil. If I give it IOSurface attribute, it will fail and complain that it's not IOSurface-backed or error -6683.

Visit these links for more information: This link is about UV interleave: How to convert from YUV to CIImage for iOS

and related question: CVOpenGLESTextureCacheCreateTextureFromImage returns error 6683

i have some similar task to perform can i have a gist of YUVFrame Struct and the conversion code.From here i was unable to get the results for converting Y-U-V pointers to pixel buffer. Thanks — Agent Smith, May 04 '17 at 07:51
My code is a mix of objective-c & SWIFT. The struct that I have is in objective-c. It consists of three NSData and two NSInteger: property (strong, nonatomic) NSData *luma; property (strong, nonatomic) NSData *chromaB; property (strong, nonatomic) NSData *chromaR; property NSInteger width; property NSInteger height; // The @'s before property are removed — mdang, May 05 '17 at 21:42

Create CVPixelBuffer from YUV with IOSurface backed

3 Answers3

Linked