Scribd Logo
Upload

Welcome to Scribd!

  • 27 views

    Radiometry of Image Forma0on: Jitendra Malik

    Uploaded by

    Elisée Ndjabu DH
    This document discusses the physics of image formation, including how light interacts with surfaces and is captured by cameras to form pixel brightness values in an image. Specifically, it describes how a scene point is projected through a pinhole camera onto the image, and how pixel brightness is proportional to the number of photons captured. It also discusses radiometry concepts like radiance and irradiance, and models of reflectance including the Lambertian and specular models. The challenges of inferring 3D scene properties from 2D images are also summarized.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Radiometry of Image Forma0on: Jitendra Malik

    Uploaded by

    Elisée Ndjabu DH
    27 views14 pages
    This document discusses the physics of image formation, including how light interacts with surfaces and is captured by cameras to form pixel brightness values in an image. Specifically, it describes how a scene point is projected through a pinhole camera onto the image, and how pixel brightness is proportional to the number of photons captured. It also discusses radiometry concepts like radiance and irradiance, and models of reflectance including the Lambertian and specular models. The challenges of inferring 3D scene properties from 2D images are also summarized.

    Original Title

    5

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document discusses the physics of image formation, including how light interacts with surfaces and is captured by cameras to form pixel brightness values in an image. Specifically, it describes how a scene point is projected through a pinhole camera onto the image, and how pixel brightness is proportional to the number of photons captured. It also discusses radiometry concepts like radiance and irradiance, and models of reflectance including the Lambertian and specular models. The challenges of inferring 3D scene properties from 2D images are also summarized.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    27 views14 pages

    Radiometry of Image Forma0on: Jitendra Malik

    Uploaded by

    Elisée Ndjabu DH
    This document discusses the physics of image formation, including how light interacts with surfaces and is captured by cameras to form pixel brightness values in an image. Specifically, it describes how a scene point is projected through a pinhole camera onto the image, and how pixel brightness is proportional to the number of photons captured. It also discusses radiometry concepts like radiance and irradiance, and models of reflectance including the Lambertian and specular models. The challenges of inferring 3D scene properties from 2D images are also summarized.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    You are on page 1of 14

    Radiometry

     of  Image  Forma0on  

     
    Jitendra  Malik  
    What  is  in  an  image?  

    The  image  is  an  array  of  brightness  


    values  (three  arrays  for  RGB  images)  
    A  camera  creates  an  image  …  

    The  image  I(x,y)  measures  how  much  light  is  captured  at  pixel  (x,y)  

    We  want  to  know  


    •  Where  does  a  point  (X,Y,Z)  in  the  world  get  imaged?  
    •  What  is  the  brightness  at  the  resul0ng  point  (x,y)?  
    The  pinhole  camera  models  where  a  scene  
    point  is  projected  

    y  
    x  
    Now  let  us  try  to  understand  brightness  at  
    a  pixel  (x,y)  …  

    The  image  I(x,y)  measures  how  much  light  is  captured  at  pixel  (x,y).  
    Propor0onal  to  the  number  of  photons  captured  at  the  sensor  
    element  (CCD/CMOS/Rod/cone/..)  in  a  0me  interval.  
    Radiance  is  a  direc0onal  quan0ty  
    Radiant  power  travelling  in  a  given  direc0on  per  unit  
    area  (measured  perpendicular  to  the  direc0on  of  
    travel)  per  unit  solid  angle  
    Image  irradiance  is  propor0onal  to  scene  
    radiance  in  the  direc0on  of  the  camera  
    What  causes  the  outgoing  radiance  at  a  scene  patch?  
    What  causes  the  outgoing  radiance  at  a  scene  patch?  

    Two  special  cases:  


    •  Specular  surfaces    -­‐  Outgoing  radiance  direc0on  obeys  angle  of  
    incidence=angle  of  reflec0on,  and  co-­‐planarity  of  incident  &  
    reflected  rays  &  the  surface  normal.  
    •  Lamber0an  surfaces  -­‐  Outgoing  radiance  same  in  all  direc0ons  
    The  Lamber0an  model  

    We  o\en  model  reflectance  by  a  combina0on  of  a  Lamber0an  term  and  a  specular  term.  If  we  
    want  to  be  precise,  we  use  a  BRDF  (Bidirec0onal  Reflectance  Distribu0on  func0on)  which  is  a  4D  
    func0on  corresponding  to  the  ra0o  of  outgoing  radiance  in  a  par0cular  direc0on  to  the  
    incoming  irradiance  in  some  other  direc0on.  This  can  be  measured  empirically.  
    Edges  are  important  
    •  Edges  are  curves  in  the  image,  across  which  the  brightness  
    changes  “a  lot”.  These  arise  because  of  discon0nui0es  in  
    reflectance,  illumina0on  or  surface  geometry.  
    Real  world  scenes  have  addi0onal  complexity…  

    •  Objects  are  illuminated  not  just  by  light  sources,  but  also  by  
    reflected  light  from  other  surfaces.  In  computer  graphics,  ray  
    tracing  and  radiosity  are  techniques  that  address  this  issue.  
    •  Shadows  
    Inver0ng  the  physics  of  image  forma0on  is  hard  

    •  Shape-­‐from-­‐shading  (SFS)  seeks  to  go  from  the  measured  


    irradiance  values  in  the  image  to  the  scene  geometry,  
    reflectances  and  illumina0on  that  caused  it.  
    •  This  is  the  inverse  of  the  computer  graphics  rendering  
    problem  where  the  goal  is  to  produce  the  image,  given  the  
    scene.  
    •  The  inverse  problem  is  much  harder  than  the  forward  
    problem;  tradi0onal  SFS  only  works  under  gross  simplifying  
    assump0ons  on  the  physics.  
    •  Computer  vision  has  been  much  more  successful  in  exploi0ng  
    the  geometry  of  image  forma0on  with  mul0ple  views.  

    You might also like