Image Digitization

1
Image Digitization
2
Outlines

• This Lecture discusses
• Image sampling and quantization,
• Spatial resolution,
• Quality of image,
• Basics about pixels

3
An Overview of a Complete Process
4
Image Model

• An image f(x,y) perceived is the light reflected
  from object.
• Let i(x,y) be the amount of light incident on the
  scene, r(x,y) be the reflectance of an object,
  then f(x, y) i(x, y) r(x, y)

The gray level (l) of the image at point (x, y)
is the intensity of a monochrome image f(x,
y). Lmin ? l ? Lmax and Lmin, Lmax is the
gray scale
5
Image Sampling and Quantization

• The output of sensor is a continuous voltage
• To convert it to digital form, we need sampling
  and quantization
• Sampling is digitizing the coordinate values
• To create grid
• Quantization is digitizing the amplitude values
• The continuous grey levels are quantized simply
  by assigning one of eight discrete gray levels to
  each sample

6
Image Sampling and Quantization
7
Image Sampling

• Sampling process is partitioning the x y plane
  into grid
• A pair of integers (x, y) is assigned to each grid

8
Image Quantization

• The size of an image is normally M rows by N
  columns, with L2k gray levels.
• To store this image we need N x M x k bits.

9
Spatial Resolution

• the smallest discernible detail in an image
• determined by the sampling ratio
• Example the subsampled images are brought up to
  1024 by 1024

10
Spatial Resolution

• the sub-images are brought up to 1024 by 1024

11
Gray-Level Resolution

• the smallest discernible change in gray level
• false contouring is caused by the use of an
  insufficient number of gray levels in smooth
  areas of an image

12
Quality of Image

• Quality of an image depends on both spatial
  resolution and of bits used to represent
  intensities.

13
Isopreference Curve

• Quality of an image vs. image resolution
• Experiment
• Observers were provided with a set of images of
  spatial and Gray-level resolution
• Observers ranked the images based on their
  quality
• Points lying on an isopreference curve correspond
  to images of equal subjective quality

14
Isopreference Curve

• For images with a large amount of detail, only a
  few gray levels may be needed !

15
Aliasing

• A function can be represented in terms of sines
  and cosines of various frequencies
• It is possible to recover the original function,
  if the sampling rate 2fmax
• Sampling frequent enough
• Aliasing may occur if function is under-sampled
• The corruption is introduced by the sampling
  function

16
Pixels

• Picture element
• Neighbors of a Pixel
• N4(p) denotes the 4-neighbors of p (x1,y),
  (x-1,y), (x,y1), (x,y-1)
• ND(p) is the diagonal neighbors of p (x1,y1),
  (x1,y-1), (x-1,y1), (x-1,y-1)
• N8(p) is the 8-neighbors of p, it consists of
  both N4(p) and ND(p).

17
Pixels

• Let V be the set of gray-level values used to
  define connectivity
• 4-adjacency
• 8-adjacency
• m-adjacency

18
Pixels

• A path
• Pixels p and q are connected
• connected component
• boundary of a region

19
Distance Measures

• Let (x,y), (s,t) be coordinates for pixels p and
  q.
• The Euclidean distance
• The D4 distance

20
Image Mask

• The idea of mask is to let the value assigned to
  a pixel be a function of its gray level and the
  gray level of its neighbor pixels.

mask
image