Design and Implementation of Artificial Vision Systems on Reconfigurable Platforms

The aim of this project is the design, development and hardware implementation of new architectures that incorporate all the processing stages of a complete vision system (out of the acquisition of the images) on the same silicon substrate, so that the execution is efficient in time performance and accuracy. With these objectives in mind, we realizes firstly an analysis of the different types of algorithms and mathematical operations mostly used in computer vision, together with a study of the capabilities of the hardware platforms and computing paradigms often used for image processing. From these studies we design benchmarks to evaluate the novel architectures and to compare them with other already existing solutions.

The prototypes are implemented on FPGA. These are reconfigurable hardware architectures capable to fit different applications. In this sense, FPGA are very suitable to implement different computing paradigms and then to fit the different processing stages of the computer vision systems.


The purpose of this project is to advance the current state of the art in the design and implementation of hardware architectures for embedded vision systems. This will perform a thorough analysis of transactions involving common applications in computer vision both low and high level. From this analysis we propose hardware solutions with specific modules that optimize the processing of each stage.

The project development would take shape in the following points:
  • Carrying out a rigorous and effective test for assessment platforms that host computer vision systems.
  • Design new architectures that integrate an efficient way of processing all levels of a complete machine vision system.
  • Hardware implementation of proposed architectures on reconfigurable platforms (FPGA), and expected return on ASICs.