Class cv::cuda::NvidiaOpticalFlow_1_0#

Class for computing the optical flow vectors between two images using NVIDIA Optical Flow hardware and Optical Flow SDK 1.0. View details

#include <opencv2/cudaoptflow.hpp>

Collaboration diagram for cv::cuda::NvidiaOpticalFlow_1_0:

Public Types#

enum NVIDIA_OF_PERF_LEVEL {
    NV_OF_PERF_LEVEL_UNDEFINED,
    NV_OF_PERF_LEVEL_SLOW = 5,
    NV_OF_PERF_LEVEL_MEDIUM = 10,
    NV_OF_PERF_LEVEL_FAST = 20,
    NV_OF_PERF_LEVEL_MAX
}

Public Member Functions#

Public Member Functions inherited from cv::cuda::NvidiaHWOpticalFlow

Return	Name	Description
`void`	`calc(` `InputArray inputImage,` `InputArray referenceImage,` `InputOutputArray flow,` `Stream & stream = Stream::Null(),` `InputArray hint = cv::noArray(),` `OutputArray cost = cv::noArray() )`	Calculates Optical Flow using NVIDIA Optical Flow SDK.
`void`	`collectGarbage()`	Releases all buffers, contexts and device pointers.
`int`	`getGridSize()`	Returns grid size of output buffer as per the hardware’s capability.

Public Member Functions inherited from cv::Algorithm

Return	Name	Description
	`Algorithm()`
	`~Algorithm()`
`void`	`clear()`	Clears the algorithm state.
`bool`	`empty()`	Returns true if the Algorithm is empty (e.g. in the very beginning or after unsuccessful read.
`String`	`getDefaultName()`
`void`	`read(const FileNode & fn)`	Reads algorithm parameters from a file storage.
`void`	`save(const String & filename)`
`void`	`write(` `const Ptr< FileStorage > & fs,` `const String & name = String() )`
`void`	`write(FileStorage & fs)`	Stores algorithm parameters in a file storage.
`void`	`write(` `FileStorage & fs,` `const String & name )`

Static Public Member Functions#

Static Public Member Functions inherited from cv::Algorithm

Return	Name	Description
`static Ptr< _Tp >`	`load(` `const String & filename,` `const String & objname = String() )`	Loads algorithm from the file.
`static Ptr< _Tp >`	`loadFromString(` `const String & strModel,` `const String & objname = String() )`	Loads algorithm from a String.
`static Ptr< _Tp >`	`read(const FileNode & fn)`	Reads algorithm from the file node.

Additional Inherited Members#

Detailed Description#

Class for computing the optical flow vectors between two images using NVIDIA Optical Flow hardware and Optical Flow SDK 1.0.

Note

A sample application demonstrating the use of NVIDIA Optical Flow can be found at opencv_contrib_source_code/modules/cudaoptflow/samples/nvidia_optical_flow.cpp
An example application comparing accuracy and performance of NVIDIA Optical Flow with other optical flow algorithms in OpenCV can be found at opencv_contrib_source_code/modules/cudaoptflow/samples/optical_flow.cpp

Member Enumeration Documentation#

enum NVIDIA_OF_PERF_LEVEL

`NV_OF_PERF_LEVEL_UNDEFINED`
`NV_OF_PERF_LEVEL_SLOW`	Slow perf level results in lowest performance and best quality
`NV_OF_PERF_LEVEL_MEDIUM`	Medium perf level results in low performance and medium quality
`NV_OF_PERF_LEVEL_FAST`	Fast perf level results in high performance and low quality
`NV_OF_PERF_LEVEL_MAX`

Member Function Documentation#

upSampler()#

void cv::cuda::NvidiaOpticalFlow_1_0::upSampler(
InputArray flow,
cv::Size imageSize,
int gridSize,
InputOutputArray upsampledFlow )

The NVIDIA optical flow hardware generates flow vectors at granularity gridSize, which can be queried via function getGridSize(). Upsampler() helper function converts the hardware-generated flow vectors to dense representation (1 flow vector for each pixel) using nearest neighbour upsampling method.

Parameters

flow — Buffer of type CV_16FC2 containing flow vectors generated by calc().
imageSize — Size of the input image in pixels for which these flow vectors were generated.
gridSize — Granularity of the optical flow vectors returned by calc() function. Can be queried using getGridSize().
upsampledFlow — Buffer of type CV_32FC2, containing upsampled flow vectors, each flow vector for 1 pixel, in the pitch-linear layout.

create()#

static Ptr< NvidiaOpticalFlow_1_0 > cv::cuda::NvidiaOpticalFlow_1_0::create(
cv::Size imageSize,
cv::cuda::NvidiaOpticalFlow_1_0::NVIDIA_OF_PERF_LEVEL perfPreset = cv::cuda::NvidiaOpticalFlow_1_0::NV_OF_PERF_LEVEL_SLOW,
bool enableTemporalHints = false,
bool enableExternalHints = false,
bool enableCostBuffer = false,
int gpuId = 0,
Stream & inputStream = Stream::Null(),
Stream & outputStream = Stream::Null() )

Instantiate NVIDIA Optical Flow.

Parameters

imageSize — Size of input image in pixels.
perfPreset — Optional parameter. Refer NV OF SDK documentation for details about presets. Defaults to NV_OF_PERF_LEVEL_SLOW.
enableTemporalHints — Optional parameter. Flag to enable temporal hints. When set to true, the hardware uses the flow vectors generated in previous call to calc() as internal hints for the current call to calc(). Useful when computing flow vectors between successive video frames. Defaults to false.
enableExternalHints — Optional Parameter. Flag to enable passing external hints buffer to calc(). Defaults to false.
enableCostBuffer — Optional Parameter. Flag to enable cost buffer output from calc(). Defaults to false.
gpuId — Optional parameter to select the GPU ID on which the optical flow should be computed. Useful in multi-GPU systems. Defaults to 0.
inputStream — Optical flow algorithm may optionally involve cuda preprocessing on the input buffers. The input cuda stream can be used to pipeline and synchronize the cuda preprocessing tasks with OF HW engine. If input stream is not set, the execute function will use default stream which is NULL stream;
outputStream — Optical flow algorithm may optionally involve cuda post processing on the output flow vectors. The output cuda stream can be used to pipeline and synchronize the cuda post processing tasks with OF HW engine. If output stream is not set, the execute function will use default stream which is NULL stream;

Source file#

The documentation for this class was generated from the following file:

opencv2/cudaoptflow.hpp