Advanced Imaging


Advanced Imaging Magazine

Updated: July 8th, 2008 05:26 PM CDT

Real-Time, High-Speed Image Processing

How to Choose the Right Hardware and Organize it Properly
Multiplexed Processor
Parallel Processors
Cascaded Processors
Cascaded Processors With Memory Buffers

By Philip Colet

When it comes to image processing, fast is usually better and virtually all image processing systems are expected to provide results quickly. To achieve the real-time and high-speed performance that most manufacturers desire, the correct image processing hardware must be chosen and the processors need to be organized appropriately.

Before selecting image-processing hardware, the effective image data rate for the purpose of image processing must be known. To determine the effective image data rate, the following four features must be known:

  • The image resolution (RI): the number of rows and columns of pixels in each image.
  • The time between each image acquisition (TI).
  • The allowable image processing latency from trigger to output (TP)
  • The image acquisition time (TA): the time starting with the trigger and including the exposure and the transmission of the image to the computer.

The steps in determining the effective image data rate for image processing (PR) are:

  1. Determine the number of pixels in an image (NP) from the image resolution (RI).
  2. Determine the effective image data rate (PR) in pixels/second using the minimum of TI or TP-TA: PR=NP/min[TI,TP-TA]

For example, suppose a system uses a synchronous (free-running) camera with an image resolution (RI) of 640x480 pixels (NP=307,200 pixels), needs to acquire an image every 0.5 second, takes between 0.05 and 0.088 second to expose and transfer the image and must have the results available within 0.2 second after image acquisition is triggered. Then the effective image data rate is:

PR=307,200 pixels/min(0.5 sec, 0.2-0.088 sec)=307,200/0.112 = 2,743,000 pixels/sec.

1 2 3 4 next

Subscribe to our RSS Feeds