Super-resolution method
for blurred image captured by handheld camera
Super-resolution method using continuous image taken with handheld camera.
Estimating plane and blur, using tracked features,and generate super-resolution image.
We can succeed super-resolution for multiple independetly moving 3D objects and multiple direction blurred object.

Plane Estimation
  • Detecting planes represented by feature points,and selecting plane for each pixsel.
    A feature point may belong to multiple groups in our technique.
    It is an advantage for handling curved surface objects and non-rigid objects.
    Input image
    		 SR result Plane Estimation result

    Blur Estimation
  • Pixel based blur estimation using homography which come from selected plane.
    Input image
    		 SR result Deblur+SR result
    Input image
    		 SR result
    		 Deblur+SR result

    Publications
      [1]Yuki Matushita,Shinya Fukumoto,Hiroshi Kawasaki,Takuma Yamaguchi,Ryo Furuakwa,Hisato Fukuda "Sanjigen keijyou wo fukumu scene niokeru tebure eizou no cyoukaizou shori" 14th Meeting on Image Recognition and Understanding(MIRU2011), pp.951-958, 7.2011

      [2]Takuma Yamaguchi,Hisato Fukuda,Ryo Furukawa,Hiroshi Kawasaki,Peter Sturm "Video deblurring and super-resolution technique for multiple moving objects" The Tenth Asian Conference on Computer Vision (ACCV2010), Vol.4, pp.2227-2240, 11.2010
    • Kawasaki Laboratory