Lec 8 - Image Enhancement Flashcards
Applied to more effectively displayed or record the image data for subsequent visual interpretation.
Image Enhancement
It is improving the perception of information in images for human viewers and providing better input for the further image processing techniques.
Image Enhancement
Commonly applied to remotely sensed data to improve the appearance of an image and a new enhanced image is produced. The enhanced image is generally easier to interpret than the original image.
Image Enhancement
Image Enhancement Process Flow (3)
(1) Input Image
(2) Application of Enhancement Techniques
(3) Better Image
Examples of Enhancement Techniques (2)
(1) Noise Removal,
(2) Contrast Adjustment
Enhancement Methods (3)
(1) Linear Contrast Stretching,
(2) Histogram Equalization
(3) Spatial Filtering
A common image processing technique used to improve the visual quality of digital images. It does not alter the shape of the original histogram but focuses on expanding the intensity range to enhance contrast within the existing data.
Linear Contrast Stretching
How Linear Contrast Stretching Works (3)
(1) Analyze the histogram of image.
(2) Identifying Minimum and Maximum Intensity Values
(3) Stretching the Intensity Range.
Linear Contrast Stretching Formula
NewPixelValue
= (OriginalPixelValue - Min) * (NewMax NewMin) / (Max - Min) + NewMin
Wherein :
Original Pixel Value - The original intensity value of a pixel.
Min - The minimum intensity value in the image.
Max - The maximum intensity value in the image.
NewMin - The desired minimum intensity value (usually 0).
NewMax - The desired maximum intensity value (usually 255 for
8-bit images).
A uniform distribution of the input range of values across the full range may not be always be an appropriate enhancement, particularly if the input range is not uniformly distributed.
Histogram Equalization
It assigns more display values (range) to the frequently occurring portion of the histogram. It stretches the histogram of an image so that it covers the entire available intensity range, making the image more visually appealing.
Histogram Equalization
A technique used in image enhancement to modify the pixel values of an image based on the values of neighboring pixels. It is designed to highlight or suppress specific features on an image base on their Spatial Frequency.
Spatial Filtering
Spatial Filtering Purposes (3)
(1) Noise Reduction
(2) Edge Detection
(3) Image Sharpening
Spatial Filtering can be classified into two:
(1) High Pass Filter
(2) Low Pass Filter
Designed to emphasize a larger, homogenous areas of similar tone and reduce the smaller details in an image. This generally serve to smooth the appearance of an image. Average and median filters, often for radar imagery.
Low Pass Filter
Serve to sharpen the appearance of fine detail in an image. One implementation of this filter first applies a low pass filter to an image and then subtracts the result from the original, leaving behind only the high spatial frequency information.
High Pass Filter
Spectral information of the object recorded in multiple bands.
Satellite Images
These images may be separate spectral bands from a single multi spectral data set, or they may be individual bands from data sets that have been recorded at different dates or using different sensors.
Satellite Images
Multi-band Operations (6)
(1) The use of ratio images to reduce topographic effects.
(2) Vegetation indexes, some of which are more complex than ratios only
(3) Multi-band statistics.
(4) Principal components analysis.
(5) Image algebra
(6) Image fusion
When a satellite passes over an area with relief, it records both…
shaded and sunlit areas
In the individual Landsat-TM bands 3 and 4, the DNS of the silt stone are… However, the ratio values are…
lower in the shaded than in the sunlit areas.
nearly identical, irrespective of illumination conditions.
The process of visually analyzing and extracting information from images, particularly in the context of remote sensing, geospatial analysis, and satellite or aerial imagery.
Image Interpretation
The ratio of the near-infrared band to the red band.
Ratio Vegetation Index (RVI)
Ratio Vegetation Index (RVI) Formula
RVI = NIR/Red
Normalized Difference Vegetation Index (NDVI) Formula
NDVI = (NIR - Red)/(NIR + Red)
Various mathematical combinations of satellite bands that have been found to be sensitive indicators of the presence and condition of green vegetation.
Vegetation Indices
Types of Vegetation Indices (2)
(1) Simple Vegetation Index (VI)
(2) Normalized Difference Vegetation Index (NDVI).
These have a relatively high reflection in the near-infrared and a low reflection in the visible range of the spectrum.
Vegetated Areas
These have larger visual than near-infrared reflectance.
Clouds, water and snow
These have similar reflectance in both spectral regions.
Rock and bare soil
These are produced by generating a normalized difference vegetation index from an infrared image and then doing a vegetation classification.
Vegetation maps
This reflects very strongly in the near infrared light range and therefore infrared images can detect stress in many crops before it is visible with the naked eye.
Green Vegetation
This is used to separate green vegetation from the background soil brightness.
Normalized Difference Vegetation Index (NDVI)
It is the difference between the near infrared and red reflectance normalized over the sum of these bands.
Normalized Difference Vegetation Index (NDVI)
Normalized Difference Vegetation Index (NDVI) Formula
NDVI = (IR-Red)/(IR+Red)
These are used to classify NDVI maps and display as vegetation maps with different colors representing different levels of vegetation.
Vegetation Categories
It is used to reduce layers of images by removing redundant data and create smaller data sets. This process results in data sets that are easier to visualize and analyze.
Principal Component Analysis
The output of Principal Component Analysis (PCA) contains… (5)
(1) Variance/Covariance Matrix
(2) Correlation Matrix
(3) Eigenvalues: amount of bands explained by each of the principal components
(4) Eigenvectors: parameters of the linear combination of the principal components for an inverse transformation back to the original bands
(5) Loadings (Pattern Matrix)
Principal Component Analysis Process
(1) correlated hi-d data
(2) Center the points
(3) Compute the variance matrix
(4) Eigenvectors + eigenvalues
(5) Pick m<d eigenvalues w. highest eigenvalues
(6) Project data points to those eigenvectors
(7) Uncorrelated low-d data
The multiplication and addition carried out for each picture element and pixel. It depends on the number of values present.
Linear Combination
Image Processing and Analysis (4)
(1) Spectral Signatures
(2) Transformation/Clustering
(3) Maximum Likelihood Classifier
(4) Classified Image (Map)
Incorporating data from remote sensing images into GIS layers is vital in creating maps that display the Earth’s surface.
Modeling/GIS
How is Multispectral Band Visualized? (2)
Computer monitors display colours as a combination of RED, GREEN, and BLUE. Combination of various intensities of each give a full spectrum of colors.
(1) Colour composites information from 3 bands is merged into one.
(2) Colour composites are described by the sequence of bands assigned to R, G, and B display.
These are digital numbers and these can be added, divided, etc., to produce new images enhancing some features.
Remotely Sensed Data
In remotely sensed images, various bands are related. Thus, the images do not have much color saturation.
Principal Component Analysis
Pixels in the RS image have different DNs, but similar places in the feature or data space.
Image Classification
Types of Image Classification (4)
(1) Supervised Classification
(2) Unsupervised Classification
(3) Box Classifier
(4) Maximum Likelihood Classifier
DNs of all pixels are plotted in the feature space and an objective statistical procedure is applied to automatically divide the feature space into pre-defined number of statistically meaningful classes.
Unsupervised Classification
Space Occupied by a given class can be determined on the basis of “sampling”.
Supervised Classification
On the basis of visual interpretation, we define classes and indicate these belonging to each class.
Pixels
On the basis of spectral characteristics of these samples, this part of the image is classified.
Reminder
Define a range of DNs for each class. Assign pixels on the basis of belonging to given volume. Pixels outside the boxes are not classified.
Box Classifier
Defines a typical pixel for each class and calculates the probability that each pixel in the image belongs to that class.
Maximum Likelihood Classifier
Usually the best classifier, because the shape of the class’ feature space is taken into account.
Maximum Likelihood Classifier