Documentation Center

  • Trial Software
  • Product Updates

Contents

vision.OpticalFlow System object

Package: vision

Estimate object velocities

Description

The OpticalFlow System object™ estimates object velocities from one image or video frame to another. It uses either the Horn-Schunck or the Lucas-Kanade method.

Construction

opticalFlow = vision.OpticalFlow returns an optical flow System object, opticalFlow. This object estimates the direction and speed of object motion from one image to another or from one video frame to another.

opticalFlow = vision.OpticalFlow(Name,Value) returns an optical flow System object, H, with each specified property set to the specified value. You can specify additional name-value pair arguments in any order as (Name1, Value1,...,NameN,ValueN).

To estimate velocity:

  1. Define and set up your text inserter using the constructor.

  2. Call the step method with the input image, I, the optical flow object, opticalFlow, and any optional properties. See the syntax below for using the step method.

VSQ = step(opticalFlow,I) computes the optical flow of input image, I, from one video frame to another, and returns VSQ, specified as a matrix of velocity magnitudes.

V = step(opticalFlow,I) computes the optical flow of input image. I, from one video frame to another, and returns V, specified as a complex matrix of horizontal and vertical components. This applies when you set the OutputValue property to 'Horizontal and vertical components in complex form'.

[...] = step(opticalFlow,I1,I2) computes the optical flow of the input image I1, using I2 as a reference frame. This applies when you set the ReferenceFrameSource property to 'Input port'.

[..., IMV] = step(opticalFlow,I) outputs the delayed input image, IMV. The delay is equal to the latency introduced by the computation of the motion vectors. This property applies when you set the Method property to 'Lucas-Kanade', the TemporalGradientFilter property to 'Derivative of Gaussian', and the MotionVectorImageOutputport property to true.

Properties

Method

Optical flow computation algorithm

Specify the algorithm to compute the optical flow as one of Horn-Schunck | Lucas-Kanade.

Default: Horn-Schunck

ReferenceFrameSource

Source of reference frame for optical flow calculation

Specify computing optical flow as one of Property | Input port. When you set this property to Property, you can use the ReferenceFrameDelay property to determine a previous frame with which to compare. When you set this property to Input port, supply an input image for comparison.

This property applies when you set the Method property to Horn-Schunck. This property also applies when you set the Method property to Lucas-Kanade and the TemporalGradientFilter property to Difference filter [-1 1].

Default: Property

ReferenceFrameDelay

Number of frames between reference frame and current frame

Specify the number of frames between the reference and current frame as a positive scalar integer. This property applies when you set the ReferenceFrameSource property to Current frame and N-th frame back.

Default: 1

Smoothness

Expected smoothness of optical flow

Specify the smoothness factor as a positive scalar number. If the relative motion between the two images or video frames is large, specify a large positive scalar value. If the relative motion is small, specify a small positive scalar value. This property applies when you set the Method property to Horn-Schunck. This property is tunable.

Default: 1

IterationTerminationCondition

Condition to stop iterative solution computation

Specify when the optical flow iterative solution stops. Specify as one of Maximum iteration count | Velocity difference threshold | Either . This property applies when you set the Method property to Horn-Schunck.

Default: Maximum iteration count

MaximumIterationCount

Maximum number of iterations to perform

Specify the maximum number of iterations to perform in the optical flow iterative solution computation as a positive scalar integer. This property applies when you set the Method property to Horn-Schunck and the IterationTerminationCondition property to either Maximum iteration count or Either. This property is tunable.

Default: 10

VelocityDifferenceThreshold

Velocity difference threshold

Specify the velocity difference threshold to stop the optical flow iterative solution computation as a positive scalar number. This property applies when you set the Method property to Horn-Schunck and the IterationTerminationCondition property to either Maximum iteration count or Either. This property is tunable.

Default: eps

OutputValue

Form of velocity output

Specify the velocity output as one of Magnitude-squared | Horizontal and vertical components in complex form.

Default: Magnitude-squared

TemporalGradientFilter

Temporal gradient filter used by Lucas-Kanade algorithm

Specify the temporal gradient filter used by the Lucas-Kanade algorithm as one of Difference filter [-1 1] | Derivative of Gaussian. This property applies when you set the Method property to Lucas-Kanade.

Default: Difference filter [-1 1]

BufferedFramesCount

Number of frames to buffer for temporal smoothing

Specify the number of frames to buffer for temporal smoothing as an odd integer from 3 to 31, both inclusive. This property determines characteristics such as the standard deviation and the number of filter coefficients of the Gaussian filter used to perform temporal filtering. This property applies when you set the Method property to Lucas-Kanade and the TemporalGradientFilter property to Derivative of Gaussian.

Default: 3

ImageSmoothingFilterStandardDeviation

Standard deviation for image smoothing filter

Specify the standard deviation for the Gaussian filter used to smooth the image using spatial filtering. Use a positive scalar number. This property applies when you set the Method property to Lucas-Kanade and the TemporalGradientFilter property to Derivative of Gaussian.

Default: 1.5

GradientSmoothingFilterStandardDeviation

Standard deviation for gradient smoothing filter

Specify the standard deviation for the filter used to smooth the spatiotemporal image gradient components. Use a positive scalar number. This property applies when you set the Method property to Lucas-Kanade and the TemporalGradientFilter property to Derivative of Gaussian.

Default: 1

DiscardIllConditionedEstimates

Discard normal flow estimates when constraint equation is ill-conditioned

When the optical flow constraint equation is ill conditioned, set this property to true so that the motion vector is set to 0. This property applies when you set the Method property to Lucas-Kanade and the TemporalGradientFilter property to Derivative of Gaussian. This property is tunable.

Default: false

MotionVectorImageOutputport

Return image corresponding to motion vectors

Set this property to true to output the image that corresponds to the motion vector being output by the System object. This property applies when you set the Method property to Lucas-Kanade and the TemporalGradientFilter property to Derivative of Gaussian.

Default: false

NoiseReductionThreshold

Threshold for noise reduction

Specify the motion threshold between each image or video frame as a positive scalar number. The higher the number, the less small movements impact the optical flow calculation. This property applies when you set the Method property to Lucas-Kanade. This property is tunable.

Default: 0.0039

 Fixed-Point Properties

Methods

cloneCreate optical flow object with same property values
getNumInputsNumber of expected inputs to step method
getNumOutputsNumber of outputs from step method
isLockedLocked status for input attributes and non-tunable properties
release Allow property value and input characteristics changes
stepEstimate direction and speed of object motion between video frames

Examples

expand all

Track Cars Using Optical Flow

Set up objects.

videoReader = vision.VideoFileReader('viptraffic.avi','ImageColorSpace','Intensity','VideoOutputDataType','uint8');
converter = vision.ImageDataTypeConverter; 
opticalFlow = vision.OpticalFlow('ReferenceFrameDelay', 1);
opticalFlow.OutputValue = 'Horizontal and vertical components in complex form';
shapeInserter = vision.ShapeInserter('Shape','Lines','BorderColor','Custom', 'CustomBorderColor', 255);
videoPlayer = vision.VideoPlayer('Name','Motion Vector');

Convert the image to single precision, then compute optical flow for the video. Generate coordinate points and draw lines to indicate flow. Display results.

while ~isDone(videoReader)
    frame = step(videoReader);
    im = step(converter, frame);
    of = step(opticalFlow, im);
    lines = videooptflowlines(of, 20);
    if ~isempty(lines)
      out =  step(shapeInserter, im, lines); 
      step(videoPlayer, out);
    end
end

Close the video reader and player

release(videoPlayer);
release(videoReader);

Algorithms

To compute the optical flow between two images, you must solve the following optical flow constraint equation:

In this equation, the following values are represented:

  • , and are the spatiotemporal image brightness derivatives

  • u is the horizontal optical flow

  • v is the vertical optical flow

Because this equation is underconstrained, there are several methods to solve for u and v:

  • Horn-Schunck Method

  • Lucas-Kanade Method

See the following two sections for descriptions of these methods

Horn-Schunck Method

By assuming that the optical flow is smooth over the entire image, the Horn-Schunck method computes an estimate of the velocity field, , that minimizes this equation:

In this equation, and are the spatial derivatives of the optical velocity component u, and scales the global smoothness term. The Horn-Schunck method minimizes the previous equation to obtain the velocity field, [u v], for each pixel in the image, which is given by the following equations:

In this equation, is the velocity estimate for the pixel at (x,y), and is the neighborhood average of . For k=0, the initial velocity is 0.

When you choose the Horn-Schunck method, u and v are solved as follows:

  1. Compute and using the Sobel convolution kernel: , and its transposed form for each pixel in the first image.

  2. Compute between images 1 and 2 using the kernel.

  3. Assume the previous velocity to be 0, and compute the average velocity for each pixel using as a convolution kernel.

  4. Iteratively solve for u and v.

Lucas-Kanade Method

To solve the optical flow constraint equation for u and v, the Lucas-Kanade method divides the original image into smaller sections and assumes a constant velocity in each section. Then, it performs a weighted least-square fit of the optical flow constraint equation to a constant model for in each section, , by minimizing the following equation:

Here, W is a window function that emphasizes the constraints at the center of each section. The solution to the minimization problem is given by the following equation:

When you choose the Lucas-Kanade method, is computed using a difference filter or a derivative of a Gaussian filter.

The two following sections explain how , , , and then u and v are computed.

Difference Filter

When you set the Temporal gradient filter to Difference filter [-1 1], u and v are solved as follows:

  1. Compute and using the kernel and its transposed form.

    If you are working with fixed-point data types, the kernel values are signed fixed-point values with word length equal to 16 and fraction length equal to 15.

  2. Compute between images 1 and 2 using the kernel.

  3. Smooth the gradient components, , , and , using a separable and isotropic 5-by-5 element kernel whose effective 1-D coefficients are . If you are working with fixed-point data types, the kernel values are unsigned fixed-point values with word length equal to 8 and fraction length equal to 7.

  4. Solve the 2-by-2 linear equations for each pixel using the following method:

    • If

      Then the eigenvalues of A are

      In the fixed-point diagrams,

    • The eigenvalues are compared to the threshold, , that corresponds to the value you enter for the threshold for noise reduction. The results fall into one of the following cases:

      Case 1: and

      A is nonsingular, the system of equations are solved using Cramer's rule.

      Case 2: and

      A is singular (noninvertible), the gradient flow is normalized to calculate u and v.

      Case 3: and

      The optical flow, u and v, is 0.

Derivative of Gaussian

If you set the temporal gradient filter to Derivative of Gaussian, u and v are solved using the following steps. You can see the flow chart for this process at the end of this section:

  1. Compute and using the following steps:

    1. Use a Gaussian filter to perform temporal filtering. Specify the temporal filter characteristics such as the standard deviation and number of filter coefficients using the BufferedFramesCount property.

    2. Use a Gaussian filter and the derivative of a Gaussian filter to smooth the image using spatial filtering. Specify the standard deviation and length of the image smoothing filter using the ImageSmoothingFilterStandardDeviationt property.

  2. Compute between images 1 and 2 using the following steps:

    1. Use the derivative of a Gaussian filter to perform temporal filtering. Specify the temporal filter characteristics such as the standard deviation and number of filter coefficients using the BufferedFramesCount parameter.

    2. Use the filter described in step 1b to perform spatial filtering on the output of the temporal filter.

  3. Smooth the gradient components, , , and , using a gradient smoothing filter. Use the GradientSmoothingFilterStandardDeviation property to specify the standard deviation and the number of filter coefficients for the gradient smoothing filter.

  4. Solve the 2-by-2 linear equations for each pixel using the following method:

    • If

      Then the eigenvalues of A are

    • When the object finds the eigenvalues, it compares them to the threshold, , that corresponds to the value you enter for the NoiseReductionThreshold parameter. The results fall into one of the following cases:

      Case 1: and

      A is nonsingular, so the object solves the system of equations using Cramer's rule.

      Case 2: and

      A is singular (noninvertible), so the object normalizes the gradient flow to calculate u and v.

      Case 3: and

      The optical flow, u and v, is 0.

References

[1] Barron, J.L., D.J. Fleet, S.S. Beauchemin, and T.A. Burkitt. Performance of optical flow techniques. CVPR, 1992.

See Also

Was this topic helpful?