LECTURE 2

Chapter 3: Enhancement in the spatial domain

3.1 Image enhancement: Background

• Subjective: perceptions differ from person to person

• OVERVIEW

Spatial domain Chapter 3

methods Direct manipulation of pixels
g(x, y) = T [f (x, y)]
• Point processing (1 × 1 masks)
s = T (r)
• Mask processing (Spatial filtering)
3 × 3 masks, 5 × 5 masks, etc.
• g(x, y) = T [set of input images]

f (x, y) −→ Filter −→ g(x, y)

Frequency domain Chapter 4

methods Modify Fourier transform

f (x, y) −→ Transform −→ Filter

−→ Inverse transform −→ g(x, y)

3.1 Background Lecture 2 (page 2)

Connection between above methods: Convolution theorem

Combined methods: Spatial/frequency domain

Mask processing: g(x, y) = T [f (x, y)]

Point processing: s = T (r)

3.2 Gray scale transformations Lecture 2 (page 3)

Point processing: s = T (r)

• Linear: identity, negative

• Logarithmic: log, inverse-log
• Power-law: nth power, nth root
3.2 Gray scale transformations Lecture 2 (page 4)

3.2.1 Image Negatives

s=L−1−r

3.2.2 Log Transformations

s = c log(1 + r)
3.2 Gray scale transformations Lecture 2 (page 5)

3.2.3 Power-Law Transformations: s = crγ

Example: Gamma correction

3.2 Gray scale transformations Lecture 2 (page 6)

Example 3.1: Contrast enhancement

3.2 Gray scale transformations Lecture 2 (page 7)

Example 3.2
3.2 Gray scale transformations Lecture 2 (page 8)

3.2.4. Piecewise-Linear Transformation Functions

Advantage: arbitrarily complex

Disadvantage: more user input

Contrast stretching
3.2 Gray scale transformations Lecture 2 (page 9)

Gray-level slicing
3.2 Gray scale transformations Lecture 2 (page 10)

Bit-plane slicing
3.2 Gray scale transformations Lecture 2 (page 11)

Bit-plane slicing
3.4 Arithmetic/Logic operations Lecture 2 (page 12)

• Pixel-by-pixel between 2 or more images (NOT: 1 image)

• Logic operations: AND, OR, NOT

• Operate on strings of binary numbers

• NOT: performs negative transformation
• AND, OR: masking, region of interest (ROI) processing

• Arithmetic operations: −, +, ×, /
• −, +: more important
• ×: gray-level masking
3.4 Arithmetic/Logic operations Lecture 2 (page 13)

3.4.1 Image Subtraction

g(x, y) = f (x, y) − h(x, y)

• Enhancement of differences between images

3.4 Arithmetic/Logic operations Lecture 2 (page 14)

3.4.1 Image Subtraction...

Example 3.7: Mask mode radiography

g(x, y) = f (x, y) − h(x, y)

• f (x, y): Dynamic TV image with dye injected

• h(x, y): Mask: Still TV image without dye
• g(x, y): Dynamic TV image with mask subtracted out

• Scaling of difference image

• Application to image segmentation (Chapter 10)
3.4 Arithmetic/Logic operations Lecture 2 (page 15)

3.4.2 Image Averaging

g(x, y) = f (x, y) + η(x, y)

• f (x, y): Original image

• η(x, y): Noise (uncorrelated, zero average)

• g(x, y): Noisy image

1 XK
g(x, y) = gi(x, y)
K i=1

⇒ E{g(x, y)} = f (x, y),

2 1 2
σg(x,y) = ση(x,y),
K

1
σg(x,y) = √ ση(x,y).
K

• In practice gi(x, y) must be registered (aligned)

3.4 Arithmetic/Logic operations Lecture 2 (page 16)

Example 3.8: Noise reduction by image averaging