About Us

Click here for an introduction PPT about us.

Super Resolution

  • What is Super Resolution?

    Super Resolution Reconstruction is to enhance resolution of low resolution images by using software technique to overcome the blurring and noising effect in imaging process so as to present more detailed information of images in higher resolution. Comparing with methods of hardware-based resolution enhancement, software-based Super Resolution Reconstruction costs less. Furthermore, most existing low resolution imaging systems can be still utilized, which offers an economical transition scheme.

  • What are applications of Super Resolution?

     

    • Satellite imaging and medical imaging

      To reconstruct one higher resolution image from low resolution image sequence.

      lr2hr
    • Synthetic zooming of region of interest[1]

       

      mshd
    • Face analysis and recognition

      You happened to find a precious group[2] photo with your buddies long ago from the drawer, but due to long time storage the photo is obscured. Unfortunately, you cannot remember each face clearly which you were once quite familiar with. How to deal with it? Through Face Hallucination technique, these familiar smiling faces come back once again! You will never worry about the passage of time.

      face

Frame Rate Up-Conversion

We often see the slow motion playback of the predation moment in “Animal World”, as shown in the following image[3]. In fact these predation actions only stay for a few tenths of seconds. It achieves by using high speed camera which makes a video at speed of nearly one thousand frames per second and then plays the video at the normal speed of 30 frames per second so that action in one second is made into fifteen minutes. By slowing the speed of actions of animals by one thousand times, it manages to record those invisible details.

Considering the limitation of photography equipment and network transmission, the frame rate of most existing digital video source is relatively low, which cannot provide audience with fluent visual experience, especially in the drastic sports contest (e.g. soccer etc.). How to enhance the frame rate of low frame rate video effectively is a problem which is studied constantly in video processing field. It involves motion analysis, image segmentation, human vision features and many other fields. Focusing on enhancing the frame rate to improve the subjective visual effect of video, Frame Rate Up-Conversion is based on consistency and continuity of video sequences, and combine techniques of image Super Resolution Reconstruction to further improve the play quality of video.

cb

  • FRUC application example

    Dynamic images produced from static images (click to watch the demo[4]).

    firstFrame loli

References

[1] http://research.microsoft.com/en-us/um/redmond/groups/ivm/HDView/

[2] http://people.csail.mit.edu/celiu/FaceHallucination/fh.html

[3] http://www.hitachi.com/

[4] D. Mahajan, F.Huang, W.Matusik, R.Ramamoorthi, P.Belhumeur, Moving Gradients: A Path-based Method for Plausible Image Interpolation, Siggraph, 2009.

top