An Introduction to Digital Imaging

1
An Introduction to Digital Imaging

• H. Colijn
• OSU Campus Electron Optics Facility

2
Digital Imaging

• Image Acquisition
• Image Storage
• Image Processing
• Image Printing

3
What is an image?

• A 2-dimensional representation of the function, I
  f(x,y) where
• ? f(x,y) is a continuous function for analog
  images and
• ? f(x,y) is a discrete function for digital
  images.

4
Acquisition Considerations

• The image magnification and intensity should be
  calibrated.
• The image should not be geometrically distorted.
• The entire range of the original image function
  should be recorded with a known response function
  (preferably linear).

5
How do you acquire a digital image?

• Film / scanner
• Video camera / framestore
• Digital camera
• Digital scan generator

6
Film

• Film, while an old technology, has many benefits
  as a recording medium.
• High spatial resolution
• Parallel image recording
• Long storage life
• No special reader needed
• Cheap

7
Film
Film is characterized by

• Film grain/speed -- resolution is determined by
  the grain size.
• Contrast and Optical Density (OD) range

8
Polaroid Type 55

• Intrinsic Film Resolution
• Positive - 20-25 lp/mm (1000 dpi)
• Negative - 150 lp/mm (7500 dpi)
• SEMs use, at most, a 2000 line photo CRT (500
  dpi).

9
Film Characteristic Curve
10
Scanner

• Optical resolution vs. interpolated resolution
• Input Density range
• Digitizing precision ( of bits)
• Beware of gamma (non-linear) corrections during
  scanning

11
Video

• Advantages
• Commonly available and cheap
• Good Temporal resolution
• Disadvantages
• Limited Spatial resolution
• Limited Dynamic range
• Interlaced acquisition

12
Video Formats

• NTSC - Standard US format
• 525 lines, interlaced, 1/30 sec refresh
• 43 aspect ratio
• PAL - Standard European format
• 625 lines, 1/25 sec refresh
• French use Secam (700 lines)
• HDTV
• 1000 lines with a 169 aspect ratio

13
Framestores

• Image processing functions in hardware.
• Averaging and integration capabilities.
• Background subtraction can improve dynamic range

14
Digital Cameras

• CCD vs. linear array
• Consumer vs. scientific grade CCDs
• High quality CCDs (linear, few defects, etc.) are
  expensive (e.g. 1k ? 1k TEM system 100k)
• The entire camera system is important, not just
  the CCD.

15
Digital Scan generator

• Easily capable of 4k x4k resolution (12 bit DAC)
• 14 and 16 bit DACs now available
• Requires a raster or scanning type instrument

16
Digital image considerations

• The ultimate questions are
• ? What does the human eye see?
• ? Does the image convey the necessary
  information?
• Consider the final image requirements before
  acquiring the initial image.

17
Spatial Resolution

• The human eye has a resolution of 100-200mm (125
  - 250 per inch)
• People will position themselves so that the
  smallest feature of interest subtends roughly 1
  minute of arc i.e. 1/60?.

18
Intensity Response

• The eye can respond to a light intensity range of
  1010, although the instantaneous dynamic range is
  about 100 gray levels
• However, at any one point in a complex image the
  eye can detect only about 15 - 25 gray levels.

19
Visual Response
Power law
Logarithmic
Perceived Brightness
Luminance
20
The bottom line is

• You need a sufficiently high pixel frequency for
  the image to appear smooth.
• For most complex images, you need at least 100
  gray levels for the eye to see a reasonably
  smooth display.

21
Digital Image Storage

• Format
• Lossless (LZW, e.g. .zip)
• Lossy (JPEG)
• Media
• Zip disks
• CD-ROM
• DVD

22
Storage Formats

• TIFF
• Usually lossless format
• Mostly standardized
• JPEG
• lossy format
• variable compression ratio

23
TIFF vs. JPEG
TIFF
JPEG
24
TIFF vs. JPEG
TIFF
JPEG
25
TIFF vs. JPEG
TIFF
JPEG
26
Storage Formats

• BMP
• Windows, Lossless
• GIF
• One byte deep
• PICT
• Mac standard format

27
Storage Formats

• Proprietary Formats
• Gatan
• EDAX
• IMG (Philips SEM)

28
Dealing with digital images

• Enlarging
• Correcting aberrations
• Image Processing and Enhancing
• Feature Analysis
• Stereology and Reconstruction

29
Enlarging

• Overenlarging analog images gradually leads to
  noise or fuzziness.
• Overenlarging digital images allows us to see the
  pixels in the image.

30
Correcting Image Defects

• Non-square pixels
• Image distortion
• Uneven illumination

31
Non-Square Pixels
Original XL-30 image
Corrected XL-30 image
32
Image Distortion
Original distorted image
Corrected image
33
Image Processing

• Contrast enhancement
• Gamma corrections, Histogram equalization,
  Unsharp masking, etc.
• Image math (addition, subtraction, division)
• Fourier filtering
• Kernel (convolution) operators
• 3-D reconstruction

34
Feature Analysis

• Usually necessary to create a binary image
• Enhance image first, then threshold.
• Use erosion and dilation operators to segment the
  image.

35
Printing the image

• Gray scale printers
• Dye-sub
• Video printers
• Binary printers
• Laser
• Ink-jet

36
Grayscale images on binary printers

• Use a half-toning process
• half-toning and screens
• screen angles
• half-tone response

37
Half-toning and screening
38
Binary half-toning
39
What is Gamma?
Output (Input)1/?
Output Intensity
? gt 1
? lt 1
Input Intensity
40
Dithering
Original
Random noise added
41
Publication quality images?

• Images in publications are half-toned
• Newspapers use 100 lpi
• e.g. Dispatch, Lantern
• Magazines use 130 lpi
• e.g. Time, Phil. Mag.
• Some Journals use 170 lpi
• e.g. Microscopy Microanalysis

42
Image Quality
Newspaper - 100 lpi
Magazine - 130 lpi
43
Image Quality
Book - 140 lpi
Journal - 170 lpi
44
Best Image Quality
45
What resources do we have?

• Books
• John Russ - The Image Processing Handbook
• Software
• PhotoShop/IP Toolkit (PC/Mac)
• NIH Image (Mac, free)
• Gatan Digital Micrograph (Mac, expensive)
• Matlab Image Toolkit (tools only)
• GraphicConverter (Mac)
• MS Imager (PC, included with MS Office)