LPDXVA_QMatrix_MJPEG - NtDoc
Native API online documentation, based on the System Informer
(formerly Process Hacker) phnt
headers
// dxva.h
typedef struct _DXVA_QMatrix_MJPEG {
UINT16 quantvals[4][64];
} DXVA_QMatrix_MJPEG, *LPDXVA_QMatrix_MJPEG;
View the official Windows Driver Kit DDI referenceNo description available.
The DXVA_QMatrix_MJPEG structure provides the quantization table of a compressed picture for MJPEG video decoding.
quantvals[4]Array for each quantization table slot i, containing an array of 64 entries j. Each entry i specifies 64 entries j, where each is a quantization step parameter for each coefficient. This array is filled directly from the DQT marker of the input MJPEG bitstream. The order of entries is the same as defined in the standard JPEG DQT marker: zig-zag scan order. All values are stored in 16-bit WORDs even if the table only specifies 8-bit values (Pq = 0). If a quantization table is not present all entries will be initialized to zero. The exact quantization table used for a particular component of a picture shall be read from the picture parameters structure field: quantizationTableSelector.
This structure is used for MJPEG video coding when the buffer type is D3D12_VIDEO_DECODE_ARGUMENT_TYPE_INVERSE_QUANTIZATION_MATRIX in D3D12 and D3D11_VIDEO_DECODER_BUFFER_INVERSE_QUANTIZATION_MATRIX in D3D11.
If DXVA_QMatrix_MJPEG is left unspecified by the host, the accelerator shall use the default quantization table as specified in Annex K of the JPEG specification.
D3D12_VIDEO_DECODE_ARGUMENT_TYPE
D3D11_VIDEO_DECODER_BUFFER_TYPE