void XnUncompressedYUVtoRGBImageProcessor::ProcessFramePacketChunk(const XnSensorProtocolResponseHeader* pHeader, const XnUChar* pData, XnUInt32 nDataOffset, XnUInt32 nDataSize)
{
XN_PROFILING_START_SECTION("XnUncompressedYUVtoRGBImageProcessor::ProcessFramePacketChunk")
XnBuffer* pWriteBuffer = GetWriteBuffer();
if (m_ContinuousBuffer.GetSize() != 0)
{
// fill in to a whole element
XnUInt32 nReadBytes = XN_MIN(nDataSize, XN_YUV_TO_RGB_INPUT_ELEMENT_SIZE - m_ContinuousBuffer.GetSize());
m_ContinuousBuffer.UnsafeWrite(pData, nReadBytes);
pData += nReadBytes;
nDataSize -= nReadBytes;
if (m_ContinuousBuffer.GetSize() == XN_YUV_TO_RGB_INPUT_ELEMENT_SIZE)
{
if (CheckWriteBufferForOverflow(XN_YUV_TO_RGB_OUTPUT_ELEMENT_SIZE))
{
// process it
XnUInt32 nActualRead = 0;
XnUInt32 nOutputSize = pWriteBuffer->GetFreeSpaceInBuffer();
YUV422ToRGB888(m_ContinuousBuffer.GetData(), pWriteBuffer->GetUnsafeWritePointer(), XN_YUV_TO_RGB_INPUT_ELEMENT_SIZE, &nActualRead, &nOutputSize);
pWriteBuffer->UnsafeUpdateSize(XN_YUV_TO_RGB_OUTPUT_ELEMENT_SIZE);
}
m_ContinuousBuffer.Reset();
}
}
if (CheckWriteBufferForOverflow(nDataSize / XN_YUV_TO_RGB_INPUT_ELEMENT_SIZE * XN_YUV_TO_RGB_OUTPUT_ELEMENT_SIZE))
{
XnUInt32 nActualRead = 0;
XnUInt32 nOutputSize = pWriteBuffer->GetFreeSpaceInBuffer();
YUV422ToRGB888(pData, pWriteBuffer->GetUnsafeWritePointer(), nDataSize, &nActualRead, &nOutputSize);
pWriteBuffer->UnsafeUpdateSize(nOutputSize);
pData += nActualRead;
nDataSize -= nActualRead;
// if we have any bytes left, store them for next packet.
if (nDataSize > 0)
{
// no need to check for overflow. there can not be a case in which more than XN_INPUT_ELEMENT_SIZE
// are left.
m_ContinuousBuffer.UnsafeWrite(pData, nDataSize);
}
}
XN_PROFILING_END_SECTION
}
void XnUncompressedDepthProcessor::ProcessFramePacketChunk(const XnSensorProtocolResponseHeader* pHeader, const XnUChar* pData, XnUInt32 nDataOffset, XnUInt32 nDataSize)
{
XN_PROFILING_START_SECTION("XnUncompressedDepthProcessor::ProcessFramePacketChunk")
// when depth is uncompressed, we can just copy it directly to write buffer
XnBuffer* pWriteBuffer = GetWriteBuffer();
// make sure we have enough room
if (CheckWriteBufferForOverflow(nDataSize))
{
// sometimes, when packets are lost, we get uneven number of bytes, so we need to complete
// one byte, in order to keep UINT16 alignment
if (nDataSize % 2 != 0)
{
nDataSize--;
pData++;
}
// copy values. Make sure we do not get corrupted shifts
XnUInt16* pRaw = (XnUInt16*)(pData);
XnUInt16* pRawEnd = (XnUInt16*)(pData + nDataSize);
XnDepthPixel* pWriteBuf = (XnDepthPixel*)pWriteBuffer->GetUnsafeWritePointer();
while (pRaw < pRawEnd)
{
*pWriteBuf = GetOutput(XN_MIN(*pRaw, XN_DEVICE_SENSOR_MAX_SHIFT_VALUE-1));
++pRaw;
++pWriteBuf;
}
pWriteBuffer->UnsafeUpdateSize(nDataSize);
}
XN_PROFILING_END_SECTION
}
void XnPassThroughImageProcessor::ProcessFramePacketChunk(const XnSensorProtocolResponseHeader* /*pHeader*/, const XnUChar* pData, XnUInt32 /*nDataOffset*/, XnUInt32 nDataSize)
{
XN_PROFILING_START_SECTION("XnUncompressedYUVImageProcessor::ProcessFramePacketChunk")
// when image is uncompressed, we can just copy it directly to write buffer
XnBuffer* pWriteBuffer = GetWriteBuffer();
// make sure we have enough room
if (CheckWriteBufferForOverflow(nDataSize))
{
pWriteBuffer->UnsafeWrite(pData, nDataSize);
}
XN_PROFILING_END_SECTION
}
void XnUncompressedBayerProcessor::ProcessFramePacketChunk(const XnSensorProtocolResponseHeader* pHeader, const XnUChar* pData, XnUInt32 nDataOffset, XnUInt32 nDataSize)
{
XN_PROFILING_START_SECTION("XnUncompressedBayerProcessor::ProcessFramePacketChunk")
// if output format is Gray8, we can write directly to output buffer. otherwise, we need
// to write to a temp buffer.
XnBuffer* pWriteBuffer = (GetStream()->GetOutputFormat() == XN_OUTPUT_FORMAT_GRAYSCALE8) ? GetWriteBuffer() : &m_UncompressedBayerBuffer;
// make sure we have enough room
if (CheckWriteBufferForOverflow(nDataSize))
{
pWriteBuffer->UnsafeWrite(pData, nDataSize);
}
XN_PROFILING_END_SECTION
}
XnStatus XnPacked11DepthProcessor::Unpack11to16(const XnUInt8* pcInput, const XnUInt32 nInputSize, XnUInt32* pnActualRead)
{
const XnUInt8* pOrigInput = pcInput;
XnUInt32 nElements = nInputSize / XN_INPUT_ELEMENT_SIZE; // floored
XnUInt32 nNeededOutput = nElements * XN_OUTPUT_ELEMENT_SIZE;
*pnActualRead = 0;
XnBuffer* pWriteBuffer = GetWriteBuffer();
if (!CheckWriteBufferForOverflow(nNeededOutput))
{
return XN_STATUS_OUTPUT_BUFFER_OVERFLOW;
}
XnUInt16* pnOutput = (XnUInt16*)pWriteBuffer->GetUnsafeWritePointer();
// Convert the 11bit packed data into 16bit shorts
for (XnUInt32 nElem = 0; nElem < nElements; ++nElem)
{
// input: 0, 1, 2,3, 4, 5, 6,7, 8, 9,10
// -,---,---,-,---,---,---,-,---,---,-
// bits: 8,3,5,6,2,8,1,7,4,4,7,1,8,2,6,5,3,8
// ---,---,-----,---,---,-----,---,---
// output: 0, 1, 2, 3, 4, 5, 6, 7
pnOutput[0] = GetOutput((XN_TAKE_BITS(pcInput[0],8,0) << 3) | XN_TAKE_BITS(pcInput[1],3,5));
pnOutput[1] = GetOutput((XN_TAKE_BITS(pcInput[1],5,0) << 6) | XN_TAKE_BITS(pcInput[2],6,2));
pnOutput[2] = GetOutput((XN_TAKE_BITS(pcInput[2],2,0) << 9) | (XN_TAKE_BITS(pcInput[3],8,0) << 1) | XN_TAKE_BITS(pcInput[4],1,7));
pnOutput[3] = GetOutput((XN_TAKE_BITS(pcInput[4],7,0) << 4) | XN_TAKE_BITS(pcInput[5],4,4));
pnOutput[4] = GetOutput((XN_TAKE_BITS(pcInput[5],4,0) << 7) | XN_TAKE_BITS(pcInput[6],7,1));
pnOutput[5] = GetOutput((XN_TAKE_BITS(pcInput[6],1,0) << 10) | (XN_TAKE_BITS(pcInput[7],8,0) << 2) | XN_TAKE_BITS(pcInput[8],2,6));
pnOutput[6] = GetOutput((XN_TAKE_BITS(pcInput[8],6,0) << 5) | XN_TAKE_BITS(pcInput[9],5,3));
pnOutput[7] = GetOutput((XN_TAKE_BITS(pcInput[9],3,0) << 8) | XN_TAKE_BITS(pcInput[10],8,0));
pcInput += XN_INPUT_ELEMENT_SIZE;
pnOutput += 8;
}
*pnActualRead = (XnUInt32)(pcInput - pOrigInput);
pWriteBuffer->UnsafeUpdateSize(nNeededOutput);
return XN_STATUS_OK;
}
XnStatus XnPacked11DepthProcessor::Unpack11to16(const XnUInt8* pcInput, const XnUInt32 nInputSize, XnUInt32* pnActualRead)
{
const XnUInt8* pOrigInput = pcInput;
XnUInt32 nElements = nInputSize / XN_INPUT_ELEMENT_SIZE; // floored
XnUInt32 nNeededOutput = nElements * XN_OUTPUT_ELEMENT_SIZE;
*pnActualRead = 0;
XnBuffer* pWriteBuffer = GetWriteBuffer();
// Check there is enough room for the depth pixels
if (!CheckWriteBufferForOverflow(nNeededOutput))
{
return XN_STATUS_OUTPUT_BUFFER_OVERFLOW;
}
XnUInt16* pnOutput = (XnUInt16*)pWriteBuffer->GetUnsafeWritePointer();
XnUInt16 a0,a1,a2,a3,a4,a5,a6,a7;
#ifdef XN_NEON
XnUInt16 depth[8];
uint16x8_t Q0;
#endif
// Convert the 11bit packed data into 16bit shorts
for (XnUInt32 nElem = 0; nElem < nElements; ++nElem)
{
if(m_nScaleFactor > 1)
{
XnUInt32 px = m_nOffsetInFrame%m_CurrentVideoMode.resolutionX;
XnUInt32 py = (m_nOffsetInFrame)/m_CurrentVideoMode.resolutionX;
if(py%m_nScaleFactor != 0)
{
// Skip as many pixels as possible
XnUInt32 nEltsToSkip =
XN_MIN(nElements - nElem,
(m_CurrentVideoMode.resolutionX - px)/8
+ (m_nScaleFactor-(py%m_nScaleFactor) - 1)*m_CurrentVideoMode.resolutionX/8);
// ::memset(pnOutput, 0, nEltsToSkip*8*sizeof(XnUInt16));
pcInput += nEltsToSkip*XN_INPUT_ELEMENT_SIZE;
pnOutput += nEltsToSkip*8;
m_nOffsetInFrame += nEltsToSkip*8;
nElem += (nEltsToSkip-1);
continue;
}
}
// input: 0, 1, 2,3, 4, 5, 6,7, 8, 9,10
// -,---,---,-,---,---,---,-,---,---,-
// bits: 8,3,5,6,2,8,1,7,4,4,7,1,8,2,6,5,3,8
// ---,---,-----,---,---,-----,---,---
// output: 0, 1, 2, 3, 4, 5, 6, 7
if(m_nScaleFactor == 2)
{
a0 = (XN_TAKE_BITS(pcInput[0],8,0) << 3) | XN_TAKE_BITS(pcInput[1],3,5);
a2 = (XN_TAKE_BITS(pcInput[2],2,0) << 9) | (XN_TAKE_BITS(pcInput[3],8,0) << 1) | XN_TAKE_BITS(pcInput[4],1,7);
a4 = (XN_TAKE_BITS(pcInput[5],4,0) << 7) | XN_TAKE_BITS(pcInput[6],7,1);
a6 = (XN_TAKE_BITS(pcInput[8],6,0) << 5) | XN_TAKE_BITS(pcInput[9],5,3);
}
else if(m_nScaleFactor == 4)
{
a0 = (XN_TAKE_BITS(pcInput[0],8,0) << 3) | XN_TAKE_BITS(pcInput[1],3,5);
a4 = (XN_TAKE_BITS(pcInput[5],4,0) << 7) | XN_TAKE_BITS(pcInput[6],7,1);
}
else
{
a0 = (XN_TAKE_BITS(pcInput[0],8,0) << 3) | XN_TAKE_BITS(pcInput[1],3,5);
a1 = (XN_TAKE_BITS(pcInput[1],5,0) << 6) | XN_TAKE_BITS(pcInput[2],6,2);
a2 = (XN_TAKE_BITS(pcInput[2],2,0) << 9) | (XN_TAKE_BITS(pcInput[3],8,0) << 1) | XN_TAKE_BITS(pcInput[4],1,7);
a3 = (XN_TAKE_BITS(pcInput[4],7,0) << 4) | XN_TAKE_BITS(pcInput[5],4,4);
a4 = (XN_TAKE_BITS(pcInput[5],4,0) << 7) | XN_TAKE_BITS(pcInput[6],7,1);
a5 = (XN_TAKE_BITS(pcInput[6],1,0) << 10) | (XN_TAKE_BITS(pcInput[7],8,0) << 2) | XN_TAKE_BITS(pcInput[8],2,6);
a6 = (XN_TAKE_BITS(pcInput[8],6,0) << 5) | XN_TAKE_BITS(pcInput[9],5,3);
a7 = (XN_TAKE_BITS(pcInput[9],3,0) << 8) | XN_TAKE_BITS(pcInput[10],8,0);
}
#ifdef XN_NEON
depth[0] = GetOutput(a0);
depth[1] = GetOutput(a1);
depth[2] = GetOutput(a2);
depth[3] = GetOutput(a3);
depth[4] = GetOutput(a4);
depth[5] = GetOutput(a5);
depth[6] = GetOutput(a6);
depth[7] = GetOutput(a7);
// Load
Q0 = vld1q_u16(depth);
// Store
vst1q_u16(pnOutput, Q0);
#else
if(m_nScaleFactor == 2)
{
*pnOutput++ = GetOutput(a0);
*pnOutput++ = 0;
*pnOutput++ = GetOutput(a2);
*pnOutput++ = 0;
*pnOutput++ = GetOutput(a4);
*pnOutput++ = 0;
*pnOutput++ = GetOutput(a6);
*pnOutput++ = 0;
}
else if(m_nScaleFactor == 4)
{
*pnOutput++ = GetOutput(a0);
*pnOutput++ = 0;
*pnOutput++ = 0;
*pnOutput++ = 0;
*pnOutput++ = GetOutput(a4);
*pnOutput++ = 0;
*pnOutput++ = 0;
*pnOutput++ = 0;
}
else
{
*pnOutput++ = GetOutput(a0);
*pnOutput++ = GetOutput(a1);
*pnOutput++ = GetOutput(a2);
*pnOutput++ = GetOutput(a3);
*pnOutput++ = GetOutput(a4);
*pnOutput++ = GetOutput(a5);
*pnOutput++ = GetOutput(a6);
*pnOutput++ = GetOutput(a7);
}
#endif
pcInput += XN_INPUT_ELEMENT_SIZE;
m_nOffsetInFrame+=8;
}
*pnActualRead = (XnUInt32)(pcInput - pOrigInput);
pWriteBuffer->UnsafeUpdateSize(nNeededOutput);
return XN_STATUS_OK;
}