Hokkaido University Research Profiles

Information and Communication

Real-time Video Processing Technology

Algorithm development and its hardware implementation

In this laboratory, we are promoting research and development of various image processing algorithms and their real-time implementation, focusing mainly on image smoothing and brightness correction of video images, which are recently increasing in capacity (high resolution and high frame rate).

Content of research

Since the amount of data handled during image processing is generally huge, it is essential to optimize the system as a whole by combining hardware and software. In this laboratory, we are investigating the configuration of image processing systems by studying image processing algorithms and their implementations complementarily. One of the results of our research is real-time adaptive brightness correction of video images based on the Retinex theory (Fig. 1), which can adaptively correct the brightness of video images taken under conditions of large changes in illumination, such as backlighting, in real time. We are also working on high-quality image smoothing (Fig. 2) based on cost optimization, which is expected to be applied to image processing such as photo illustration, pre-processing of various image processes, brightness correction, and detail enhancement.

  • (a) Input image (b) Estimated illumination (c) Brightness correction result
    Fig. 1 Adaptive brightness correction of moving images based on the Retinex theory

  • (a) Input image (b) Image smoothing result (c) Example of FPGA board used for implementation
    Fig. 2 Image smoothing based on cost optimization

Potential for social implementation

  • ・Cameras and other imaging devices
  • ・Medical image processing equipment
  • ・Security cameras
  • ・Various displays
  • ・Computer vision

Appealing points to industry and local governments

In this laboratory, we are conducting research and development of practical applications, from the algorithm level to its LSI/FPGA implementation (Fig. 3). We believe that good algorithm and LSI architecture design can dramatically improve the performance of information processing systems and have a great impact on the world.