Parlevision is a software utility for rapid development of
computer vision applications. The process from input to output is
up into filters, each of which has a distict function. The large
of available filters and the possibility of designing a new filter,
makes Parlevision a flexible and fast tool.
Computer vision, rapid developmentThe theme in this work is: Computer Vision
Parlevision is written in C++ and uses Intel's Open-CV image
library. This library contains many optimized image processing
functions. Parlevision is a graphical interface to this library and
contains many higher-order functions.
|Example pipeline that locates hands and faces in
on skin color.
||Pipeline with edge detector process showing
A processor is simply a function that transforms input to output.
Therefore, processors have input and output pins. These pins can be
connected to pins of other processors. Input and output types can be
images, bounding boxes, numbers or any user-defined type. Since the
of each pin is known, only pins with similar types can be connected.
Each processor can have several parameters, each of which can be
specified by a user.
A series of processors is called a pipeline. A pipeline typically
starts with a producer and has one or more processors connected.
pipeline can be stored, with all processor parameters included.
This allows for a fast change of settings.
Within Parlevision, many processors are included. These include
producers for camera, video and image input. Processors that
on images include edge detection, background subtraction, color
detection, cropping, adding and subtracting and masking. Also,
level processors are available. Searching for the largest blobs,
shadow and point tracking is available in the system.
Parlevision contains a development environment. It is possible to
new processors using the processor development kit (PDK).
processor can be done fast and consistency with the system is
This eliminates the need for exhaustive type-checking.
The Parlevision system is used in many applications including
of hands, estimating human poses, corridor surveillance,
and analysis of facial expressions.