Image Enhancement

BASIC CHARACTER OF DIGITAL IMAGE Continuous tone photograph.Orginal 200*200 digital image Enlargement showing 20*20 of pixels 10*10 enlargement Digital numbers corresponding to radiance of each pixels

INTRODUCTION Some mathematical operations that are to be applied to digital remote sensing input data to improve the visual appearance of an image for better interpretability or subsequent digital analysis

REASONS Low contrast is due to  Low sensitivity of detectors  Weak signal of object present on earth  Similar reflectance of different object  Environmental conditions at the time of recording

In remote sensing many digital enhancement algorithms are available.They are  Contrast stretching enhancement  Ratioing Linear combinations  Principle component analysis  Spatial filtering

The digital enhancement increases the seperability between the interested classes or features. Broadly enhancement technique are categorized as Point operations Local operations

POINT OPERATIONS Modify the values of each pixel in image data set independently Local operations modify the values of each pixel in the context of the pixels values surrounding it.

CONTRAST ENHANCEMENT • It increasing the contrast between targets and their background.

• CONTRAST ENHANCEMENT In this pixels values clustered in narrow range of grey values. I t can altered to full range of grey values contrast between dark and light areas of image would be improved

Objective of contrast is expand the narrow dynamic range of grey values present in a input image over a wide range of grey value for the desired out put.

CONTRAST ENHANCEMENT Two types  Linear contrast stretching  Non linear contrast stretching

LINEAR CONTRAST ENHANCEMENT • To expand the original brightness values to make use of full range of radiometric scale of sensor • The lower value of original histogram is assigned a zero brightness and upper value assigned 255

LINEAR CONTRAST STRETCH

Enhancement

Contrast Stretch Maximum/Minimum

Contrast Stretch Percentage Linear

Enhances a specific region indicated by the analysis

LINEAR CONTRAST STRETCH Digital number values DN in the lower end of original histogram is assigned to zero that is extremely black and value at higher end is assigned to white DN =127 .The intermediate values are interpolated between 0 and 127

Y=a+Bx X and y are input grey values of any pixels and output grey values of same pixels. A and b are intercept and slope

NON LINEAR ENHANCEMENT • Histogram of input image does not show a uniform distribution

LOGARITHMIC CONTRAST Logarithmic contrast is very much useful for non linear contrast enhancement.Here out put pixel grey values will be generated from input pixel grey values Yij-a log(Xij)+b

HISTOGRAM • It is graphical representation of the brightness value that comprises an image. • The brightness value displayed along the x axis of the graph and the frequency of occurrence of each value in the image on y axis

HISTOGRAM EQUALISATION In this enhancement the original histogram has been readjusted to produce uniform population density of pixels along the horizontal grey value axis.

METHODS • The basic idea is to redistribute the original histogram so that each brightness level has approximately equal number of pixels. • It increase the contrast in the heavily populated range of histogram while reducing the contrast at sparsely

HISTOGRAM EQUALISATION This technique first introduced by Gonzalez and Wintz in 1977. Histogram equalisation consist of two process.  Compute the histogram of orginal image and cumulative frequency density percentage.  Computaion of transformation function based on contrast manipulation in out put.

DIGITAL FILTERS

SPATIAL FILTERING Spatial filters are designed to highlight or suppress specific features in an image based on their spatial frequency.. Filtering is performed by using convolution windows.

Continue………. Used to enhance the appearance of an image It is based on concept of image texture It highlight or suppress specific features in an image based on their spatial frequency

PROCEDURE • The window is moved over the input image from extreme top left hand corner of the scene • The discrete mathematical function transforming the original input image digital number to a new digital value.

SPATIAL FREQUENCY • Spatial frequency is related to the concept of image texture • It refers to the frequency of the variations in tone that appear in an image. –

"Rough" texture: abrupt tonal change, high spatial frequencies

–

"smooth" texture: little tonal variation, low spatial frequencies.

Digital Filters SPATIAL INFORMATION FREQUENCY • Low frequency spatial information • High frequency spatial information • Median frequency spatial information

DIGITAL FILTERS • Low Pass (Averaging) • High Pass – Edge Detection – Edge Enhancement • Directional

LOW PASS FILTERS • Designed to emphasize larger, homogeneous areas of similar tone and reduce the smaller detail in an image. • Thus, low-pass filters generally serve to smooth the appearance of an image.

CONVOLUTION FILTERING • Involves moving a 'window' of a few pixels in dimension (e.g. 3x3, 5x5, etc.) • Central pixel of window is ‘replaced’ by a new value calculated from window neighborhood

LOW PASS FILTERING • The window is moved along and the calculation is repeated until the entire image has been filtered and a "new" image has been generated. • By varying the calculation performed and the weightings of the individual pixels in the filter window, filters can be designed to enhance or suppress different types of features.

Spatial Filtering (Masking) Portion of a digital image

Mask w 1 w2 w3

z1 z2 z3

w4 w5 w6

z4 z5 z6

w7 w8 w9

z7 z8 z9 Replace with

R

= w1z1 + w2z2 + ….. +w9z9

Image Texture Coarse (rough) texture Smooth Texture

Moving Window

Moving Window Concept Projection of 3x3 Kernel The Moving Window (kernel) scans the 3x3 neighborhood of every pixel in the classified image.

Classified Image

Moving Window Concept

A value is computed, depending on the type of kernel, from the 9 values in the input file and placed in the corresponding cell of the output file. Output File

Moving Window Concept

Output File

Moving Window Concept

Output File

Moving Window Concept

Output File

Moving Window Concept

Output File

Moving Window Concept

Output File

Moving Window Concept

Output File

Moving Window Concept

Output File

Moving Window Concept

Output File

Example: Mean Kernel Calculation

8 8 2 8 2 6

9 3 8 6 2 9

9 11 11 11 9 3 9 11 3 5 5 9 8 6 8 11 4 4 6 9 11 4 4 6

1/9

1/9

1/9

1/9

1/9

1/9

1/9

1/9

1/9

(3 + 5 + 5 + 8 + 6 + 8 + 4 + 4 + 6) / 9 =

5.44

or

Round(3 + 5 + 5 + 8 + 6 + 8 + 4 + 4 + 6) / 9 =

5

Example: Mode Kernel Calculation

8 8 2 8 2 6

9 3 8 6 2 9

9 11 11 11 9 3 9 11 3 5 5 9 8 6 8 11 4 4 6 9 11 4 4 6

No Formula

In this example: The mode is initially a 3. Then two 5s set the mode to 5. Several other pairs of numbers are then found but they can’t beat the pair of 5s. Output =

5

Low Pass • Used in removing noise • Used in agggregating areas for classification • Used to map long wavelength, low amplitude trends.

Low-pass Filters Moving Average Filter (1/9)* 1 1

1

1

1

1

1

1

1

Median Filter z1 z2 z3 z4 z5 z6 z7 z8 z9

Replace with R

= median(z1, z2 , ….. , z9)

Digital Filters HIGH PASS FILTER • Used to “sharpen” blurred images • Used to map high spatial frequency features • Edge detection • Edge enhanced

It sharpen the appearance of fine detail in an image First apply low pass filter and then to subs tract the result from the original

A simple high pass filter may be implemented by subtracting a low pass filtered image from the original unprocessed image. I t is by image subtracting method.

• Thomas et al developed a model I” =I-FI’+C I”= filtered pixel value I=Original pixel value I’=Average of window F=proportion vary from 0 to 1 C=constant.

High-pass Filters Basic HP Filter (1/9)* -1 -1 -1 -1 8 -1 Gradient Filter z1 z2 z3 z4 z5 z6 z7 z8 z9

-1 -1 -1

1

0

-1 0

1 -1 0

0

demos/demo2spatial_filtering/highpassdemo.m

Low vs. High Pass Filter

Original

Low pass

High pass

Digital Filters Directional • Used to map high spatial frequency features at a particular orientation. • Provide “illumination” across an image (interpretation aid) • Enhances structures orthogonal to the specified direction.

Detection of Discontinuities Point Detection -1 -1 -1 -1 8 -1 -1 -1 -1 Line Detection (Prewitt’s Gradient) -1 -1 -1

-1 0

1

0

0

0

-1 0

1

1

1

1

-1 0

1

demos/demo2spatial_filt

Directional or Edge Filters • Designed to highlight linear features, such as roads or field boundaries. • These filters can also be designed to enhance features which are oriented in specific directions.

Edge Detection Sobel Masks

-1 -2 -1

-1 0

1

0

0

0

-2 0

2

1

2

1

-1 0

1

demos/demo2spatial_filtering/edgegradientdemo.m

>>edgedemo >>edge

Edge Filters

IMAGE TRANSFORMATION  Arithmetic operations done to combine and transform the original bands into "new" images which better display or highlight certain features in the scene. e.g Normalized Difference Vegetation Index (NDVI), PCA, HSI transforms  These use multiple bands

• Generate new images from two or more sources which tend to highlight particular features or properties of better than orginal image • Addition ,substraction ,Division and multiplication

IMGE DIVISION • Data from two different spectral bands yield useful information regarding the object.

NDVI IMAGE

PRINCIPAL COMPONENTS ANALYSIS

• It is to reduce the dimensionality in the data and compress as much of information in original bands into fewer bands

PRINCIPAL COMPONENTS ANALYSIS • Plotted in spectral space, all image data tend to be highly correlated.

PRINCIPAL COMPONENTS ANALYSIS •Principle components analysis plots the directions of maximum variability in a data set

PRINCIPAL COMPONENTS ANALYSIS •These directions are then used as new image axis.

PRINCIPAL COMPONENTS ANALYSIS

• De-correlates the image data. • Produces an image of maximum variability. • Good for detecting subtle spectral changes.