Optical flow or Optic flow is the
sequence of visible motion of object in a displayed scene which is the
resultant of relative motion in between observer and scene. Optical flow
reflects the changes occurred in an image when objects of the given image
moves, obviously it is inconvenient to say the objects in an image are moving,
it will be the video with a constant environmental background and we are going
to consider it as a picture. Optical flow is the building block for some
robotics where we need navigation as well as motion detection. There are
variety of motion which can be study with the help of Optical Flow e.g. moving
observer and static objects, static observer and moving objects, both moving.
Optical flow is the representation
of motion of an image or specifically the objects of the image in term of
motion vector. To draw the motion vector, we will analyses our image at
different time suppose at t1 and t2 we can then draw the estimation for the
next sequence of motion in term of Motion vectors.
There are number of methods to
estimate the optical flow based on partial derivatives of the images i.e. lower
or/and higher-order partial derivatives, such as:
a. Lucas–Kanade method –
regarding image patches and an affine model for the flow field.
b. Horn–Schunck method – optimizing a
functional based on residuals from the brightness constancy constraint, and a regularization
term expressing the expected smoothness of the flow field.
c. Buxton–Buxton method –
based on a model of the motion of edges in image sequences.
d. Black–Jepson method –
coarse optical flow via correlation.
e. General variational methods – a range of
modifications/extensions of Horn–Schunck, using other data terms and other
There is another term related to
Optical flow which is optical flow field, it is the velocity field that represents
the 3-D motion of object points across a 2-D image.
There are some restrictions for the
optical flow as well which are:
It shouldn’t be sensitive to illumination changes and
motion of unimportant objects like shadow.
non-zero optical flow is detected if a fixed sphere is
illuminated by a moving source.
smooth sphere rotating under constant illumination
provides no optical flow.
There are some assumptions to made
while estimating the optical flow
brightness of the object is constant over time.
2) We know
that every image has some coordinates so we assume that the nearby plane of the
image moves the same way as the image does, this is called velocity smoothness
Optical Flow analysis in term of Pixel motion:
Optical flow detects the moving
object from a 2D converted frame of a 3D image. It performs two basic operations
i.e Evaluating and Resetting. Evaluation area of Optical Flow covers the motion
properties of object in an Optical flow field by providing information about
its spatial segmentation. Resetting area combine this information and create a
framework about the guided segmentation of the moving object.
It works on the motion of pixels in
an image by providing point to point pixel correspondence. For example, at time
t it locates the pixel position then after time t+?t it again locates the new position of pixels.
To conclude the optical flow, it
provides a movement detection for a given pixel. We specify pixels coordinate
in the first frame and some target parameters. After this any change occurs in
pixels coordinates is detected.