This preview shows pages 1–3. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: #9 ECE 253a Digital Image Processing 10/26/11 Sampling in 2 dimensions Sampling refers to making the image discrete in its spatial coordinates. To discuss this, we need to introduce notation and define some functions: The 2D discrete delta function is defined by: ( n 1 , n 2 ) = braceleftBigg 1 ( n 1 , n 2 ) = (0 , 0) else The 1D continuous delta function can be defined by: ( x ) = 0 for x negationslash = 0 and lim integraldisplay ( x ) dx = 1 and the 2D continuous delta function can be defined in terms of this 1D function by: ( x, y ) = ( x ) ( y ) in which case it is separable by definition. The bedofnails function, also called an impulsive sheet, is comb ( x, y ; x, y ) = summationdisplay j = summationdisplay k = ( x j x, y k y ) = S ( x, y ) This is composed of an infinite array of Dirac delta functions arranged in a grid of spacing ( x, y ). The 2D Fourier Transform pair that we will use in this handout is: F ( u, v ) = integraldisplay integraldisplay f ( x, y ) e i 2 ( ux + vy ) dxdy f ( x, y ) = integraldisplay integraldisplay F ( u, v ) e + i 2 ( ux + vy ) dudv where x and y are the spatial coordinates of the original image, and u and v are the spatial frequency coordinates of the Fourier transform of the image. The extension to 3D is obvious. This can all be written in vector notation: 1 F ( s ) = integraldisplay f ( x ) e i 2 s T x d x f ( x ) = integraldisplay F ( s ) e + i 2 s T x d s where s = ( s 1 , s 2 , . . ., s N ) and the units of the coordinate s i are the inverse of the units of the corresponding spatial coordinate x i . Important property: if a function is separable, then its Fourier transform is separable. Using to denote a F.T. pair, we have if f ( x ) = producttext i f i ( x i ) and f ( x ) F ( s ) then F ( s ) = producttext i F i ( s i ) , where f i ( x i ) F i ( s i ) Let f I ( x, y ) denote a continuous, infinite extent ideal image field representing the luminance, photographic density, or some desired parameter of a physical image. In a perfect sampling system, spatial samples of the ideal image would be obtained by multiplying by the spatial sampling function S ( x, y ): f p ( x, y ) = f I ( x, y ) S ( x, y ) = f I ( x, y ) summationdisplay j = summationdisplay k = ( x j x, y k y ) f p ( x, y ) = summationdisplay j = summationdisplay k = f I ( j x, k y ) ( x j x, y k y ) where it is observed that f I ( x, y ) may be brought inside the summation and evaluated only at the sample points ( j x, k y )....
View Full
Document
 Fall '11
 Cosman
 Image processing

Click to edit the document details