Image Enhancements

Question 1

What is the ‘aim’ (or goal) for image enhancement?

Answer 1

The aim of image enhancement is to alter the distribution of DNs in each image channel in the image to increase the contrast, making the image interpretation easier.

Question 2

Define Contrast Stretching

2 Different types?

Answer 2

Contrast stretching is designed to expand the original brightness values in an image to increase the contrast

Linear and Non-Linear

Question 3

Linear Stretch

Answer 3

A Linear stretch increases the contrast while preserving the original radiance relationship between the pixels.

Question 4

Non-Linear Stretch

Answer 4

Increases the contrast without preserving the original radiance relationship between pixels.
- different portions of the image may be stretched/compressed.

Question 5

Spatial Filtering

Answer 5

the process of extracting the information from only the spatial domain (not spectral).

Question 6

Convolution Filters

Example?

Answer 6

convolution filters multiply the kernel by a matrix of filtering coefficients
- example low pass filter

Question 7

Non-convolution Filters

Example

Answer 7

Only work with numbers within the kernel.

- Example median filter

Question 8

What is a Kernel

Answer 8

Kernels are smaller square subsets of the image on which the filtering operation is based.

Question 9

Spatial Filtering - Spatial Frequency

Answer 9

The number of times a certain pixel occurs within an image.

Question 10

Low frequency distribution

Answer 10

values do not change significantly over the range of the image.

Question 11

High frequency distribution

Answer 11

values change drastically over the range of the image.

Question 12

Low Pass Filter

Answer 12

filter is designed to remove or de-emphasize the imagery with a high spatial frequency
-smoothes the image.

Question 13

How can you remove the effects of blurring from a low pass filter?

Answer 13

can use unequal weightings for the kernel.

Question 14

High Pass Filter

Answer 14

designed to enhance abrupt changes in data within the image.

• fine detail reduced or removed in extreme cases.
• This tends to highlight edges and other sharp boundaries

Question 15

Fast Fourier Transform (FFT)

Answer 15

A Fast Fourier Transform breaks an image down into two components: frequency and direction
- We can use this information to remove specific types of information depending on frequency
(to use low pass or high pass filter)

Question 16

FFT: Pass procedure

Answer 16

only the information that occurs within the mask is retained.

Question 17

FFT: Cut Procedure

Answer 17

only the information that occurs outside the mask is retained.

Question 18

Edge Detection Filters: Non-directional filters.

Answer 18

Non-directional filters do not focus on a particular direction within the image but rather deal with linear features in all directions.
Examples: LaPlacian Filter

Question 19

Edge Detection Filters: Directional Filters

Answer 19

Directional filters are only interested in accenting the features in one direction, or a limited number of directions.
Examples: Sobel Filters

Question 20

Compass Gradient Filter

Answer 20

•A compass gradient filter enhances linear features according to compass directions

Question 21

Sobel Filter

Answer 21

This is a nonlinear filter in that it combines the pixel values in a nonlinear fashion.

Question 22

Intensity-Hue-Saturation

Answer 22

Only makes appearance look better = eye candy

Question 23

Spectral Ratioing: Band-ratios

Answer 23

Band-ratios are commonly used to enhance or amplify the signal.
- enhance differences in brightness values from identical surface material

Question 24

Principle Component Analysis - 3 functions

Answer 24

1) The first is to reduce the dimensionality of the data set.
2) The second reason for applying a PCA is to decorrelate the data.
3) for choosing the most appropriate data set(s) for use in a specific application.

Question 25

Tasselled Cap Transformation

Answer 25

The result of the transformation is a series of new data sets, where each represents a unique reflectance component.