Algorithm used to estimate the motion vectors from previous image to current image. More...

Functions
VPIStatus	vpiCreateOpticalFlowDense (uint32_t backends, int32_t width, int32_t height, VPIImageFormat inputFmt, VPIOpticalFlowQuality quality, VPIPayload *payload)
	Creates payload for vpiSubmitOpticalFlowDense. More...

VPIStatus	vpiSubmitOpticalFlowDense (VPIStream stream, uint32_t backend, VPIPayload payload, VPIImage prevImg, VPIImage curImg, VPIImage mvImg)
	Runs dense Optical Flow on two frames. More...

Detailed Description

Algorithm used to estimate the motion vectors from previous image to current image.

Function Documentation

VPIStatus vpiCreateOpticalFlowDense	(	uint32_t	backends,
		int32_t	width,
		int32_t	height,
		VPIImageFormat	inputFmt,
		VPIOpticalFlowQuality	quality,
		VPIPayload *	payload
	)

Parameters

[in]	backends	VPI backends that are eligible to execute the algorithm. Currently only VPI_BACKEND_NVENC is accepted.
[in]	width,height	Dimensions of current and previous image
[in]	inputFmt	Format of current and previous image.
[in]	quality	Quality of the dense optical flow algorithm.
[out]	payload	Pointer to memory where the created payload handle will be written to.

VPIStatus vpiSubmitOpticalFlowDense	(	VPIStream	stream,
		uint32_t	backend,
		VPIPayload	payload,
		VPIImage	prevImg,
		VPIImage	curImg,
		VPIImage	mvImg
	)

Runs dense Optical Flow on two frames.

Outputs motion vectors for every 4x4 pixel block in the input image.

Parameters

[in]	stream	The stream where the operation will be queued in.
[in]	backend	Backend that will execute the algorithm. Must be one of the backends specified during payload creation. If 0, VPI will select one of the eligible backends from the payload that accepts the given parameters, usually the fastest one.
[in]	payload	Payload created with vpiCreateOpticalFlowDense
[in]	prevImg	Previous frame
[in]	curImg	Current frame
[out]	mvImg	Motion vectors output. Only VPI_IMAGE_FORMAT_2S16_BL is supported. Motion vector is in fixed point S10.5 format. Note VPI_BACKEND_NVENC expects dimension of the motion vectors output to be 1/4 of the current and previous image.