Presentation is loading. Please wait.

Presentation is loading. Please wait.

Lecture 6: Image Processing and Interpretation

Similar presentations


Presentation on theme: "Lecture 6: Image Processing and Interpretation"— Presentation transcript:

1 Lecture 6: Image Processing and Interpretation

2 two prime approaches in the use of remote sensing
1) standard photo-interpretation of scene content 2) use of digital image processing and classification techniques that are generally the mainstay of practical applications of information extracted from sensor data sets To accomplish this, we will utilize just one Landsat TM subscene that covers the Morro Bay area on the south-central coast of California

3 Image below, brighter portions relate to higher energy levels

4 Image interpretation relies on one or both of these approaches:
Photointerpretation:the interpreter uses his/her knowledge and experience of the real world to recognize scene objects (features, classes, materials) in photolike renditions of the images acquired by aerial or satellite surveys of the targets (land; sea; atmospheric; planetary) that depict the targets as visual scenes with variations of gray-scale tonal or color patterns (more generally, spatial or spectral variability that mirror the differences from place to place on the ground) machine-processing manipulations (usually computer-based) that analyze and reprocess the raw data into new visual or numerical products, which then are interpreted either by approach 1 or are subjected to appropriate decision-making algorithms that identify and classify the scene objects into sets of information

5 Image Processing Computer-Assisted Scene Interpretation (CASI); also called Image Processing The techniques fall into three broad categories: Image Restoration and Rectification Image Enhancement Image Classification There is a variety of CASI methods: contrast stretching, band ratioing, band transformation, Principal Component Analysis, Edge Enhancement, Pattern Recognition, and Unsupervised and Supervised Classification

6 Classification Classification is probably the most informative means of interpreting remote sensing data The output from these methods can be combined with other computer-based programs. The output can itself become input for organizing and deriving information utilizing what is known as Geographic Information Systems (GIS)

7 For the Morro Bay subscene the various images shown in this Section
were created using the IDRISI software processing program (it's worthwhile to check their ClarkLabs website ( [Clark University in Worchester, Mass.]). The IDRISI program is especially user-friendly to students wishing to gain experience in image processing

8 Image Classification In classifying features in an image we use the elements of visual interpretation to identify homogeneous groups of pixels which represent various features or land cover classes of interest. In digital images it is possible to model this process, to some extent, by using two methods: Unsupervised Classifications and Supervised Classifications.

9 Unsupervised Classifications
this is a computerized method without direction from the analyst in which pixels with similar digital numbers are grouped together into spectral classes using statistical procedures such as nearest neighbour and cluster analysis. The resulting image may then be interpreted by comparing the clusters produced with maps, airphotos, and other materials related to the image site.

10 Supervised Classification:

11 Limitations to Image Classification:
have to be approached with caution because it is a complex process with many assumptions. In supervised classifications, training areas may not have unique spectral characteristics resulting in incorrect classification. Unsupervised classifications may require field checking in order to identify spectral classes if they cannot be verified by other means (i.e. maps and airphotos).

12 Online work

13 Image Processing Digital Images:remote sensed images can also be represented in a computer as arrays of pixels (picture elements), with each pixel corresponding to a digital number, representing the brightness level of that pixel in the image. In this case, the data are in a digital format. These types of digital images are referred to as raster images in which the pixels are arranged in rows and columns

14 Data visualization The images that we view are visual representations
of the digital output from the sensor 8-bit gray shade image is the case when the sensor output is converted to one of 256 gray shades (0 to 255) 24-bit color does the same except in shades or red, green, and blue

15 Image Processing: Pixel Values
Pixel Values: The magnitude of the electromagnetic energy captured in a digital image is represented by positive digital numbers. The digital numbers are in the form of binary digits (or 'bits') which vary from 0 to a selected power of 2 Image Type Pixel Value Color Levels 8-bit image = 16-bit image 216 = 24-bit image 224 =   

16 Image Processing: Image Resolution
Image Resolution: the resolution of a digital image is dependant on the range in magnitude (i.e. range in brightness) of the pixel value. With a 2-bit image the maximum range in brightness is 22 = 4 values ranging from 0 to 3, resulting in a low resolution image. In an 8-bit image the maximum range in brightness is 28 = 256 values ranging from 0 to 255, which is a higher resolution image

17 2-bit Image (4 grey levels)

18 Image Processing Procedures
Image Restoration: most recorded images are subject to distortion due to noise which degrades the image. Two of the more common errors that occur in multi-spectral imagery are striping (or banding) and line dropouts

19 Image Processing Procedures
Dropped Lines are errors that occur in the sensor response and/or data recording and transmission which loses a row of pixels in the image.

20 Image Enhancement One of the strengths of image processing is that it gives us the ability to enhance the view of an area by manipulating the pixel values, thus making it easier for visual interpretation. There are several techniques which we can use to enhance an image, such as Contrast Stretching and Spatial Filtering.

21 Image Enhancement Image Histogram: For every digital image the pixel value represents the magnitude of an observed characteristic such as brightness level. An image histogram is a graphical representation of the brightness values that comprise an image. The brightness values (i.e ) are displayed along the x-axis of the graph. The frequency of occurrence of each of these values in the image is shown on the y-axis. 8-bit image ( brightness levels) Image Histogram x-axis = 0 to 255 y-axis = number of pixels

22 Image Enhancement Contrast Stretching: Quite often the useful data in a digital image populates only a small portion of the available range of digital values (commonly 8 bits or 256 levels). Contrast enhancement involves changing the original values so that more of the available range is used, this then increases the contrast between features and their backgrounds. There are several types of contrast enhancements which can be subdivided into Linear and Non-Linear procedures.

23 Image Enhancement Linear Contrast Stretch: This involves identifying lower and upper bounds from the histogram (usually the minimum and maximum brightness values in the image) and applying a transformation to stretch this range to fill the full range. Equalized Contrast Stretch: This stretch assigns more display values (range) to the frequently occurring portions of the histogram. In this way, the detail in these areas will be better enhanced relative to those areas of the original histogram where values occur less frequently.

24 Linear Stretch Example:
Before Linear Stretch After Linear Stretch The linear contrast stretch enhances the contrast in the image with light toned areas appearing lighter and dark areas appearing darker, making visual interpretation much easier. This example illustrates the increase in contrast in an image before (left) and after (right) a linear contrast stretch.

25 Spatial Filtering Spatial filters are designed to highlight or suppress features in an image based on their spatial frequency. The spatial frequency is related to the textural characteristics of an image. Rapid variations in brightness levels ('roughness') reflect a high spatial frequency; 'smooth' areas with little variation in brightness level or tone are characterized by a low spatial frequency. Spatial filters are used to suppress 'noise' in an image, or to highlight specific image characteristics. Low-pass Filters High-pass Filters Directional Filters etc

26 Spatial Filtering Low-pass Filters: These are used to emphasize large homogenous areas of similar tone and reduce the smaller detail. Low frequency areas are retained in the image resulting in a smoother appearance to the image. Linear Stretched Image Low-pass Filter Image

27 Spatial Filtering High-pass Filters: allow high frequency areas to pass with the resulting image having greater detail resulting in a sharpened image Linear Contrast Stretch Hi-pass Filter

28 Spatial Filtering Directional Filters:are designed to enhance linear features such as roads, streams, faults, etc.The filters can be designed to enhance features which are oriented in specific directions, making these useful for radar imagery and for geological applications. Directional filters are also known as edge detection filters. Edge Detection Lakes & Streams Edge Detection Fractures & Shoreline

29 Image Ratios It is possible to divide the digital numbers of one image band by those of another image band to create a third image. Ratio images may be used to remove the influence of light and shadow on a ridge due to the sun angle. It is also possible to calculate certain indices which can enhance vegetation or geology

30 Sensor Image Ratio EM Spectrum Application Landsat TM Bands 3/2 red/green Soils Bands 4/3 PhotoIR/red Biomass Bands 7/5 SWIR/NIR Clay Minerals/Rock Alteration For example: Normalized Difference Vegetation Index (NDVI): a commonly use vegetation index which uses the red and infrared bands of the EM spectrum.

31 Image Ratio example: NDVI
NDVI image of Canada. Green/Yellow/Brown represent  decreasing magnitude of the vegetation index.

32 Principle Components Analysis
Different bands in multispectral images like those from Landsat TM have similar visual appearances since reflectances for the same surface cover types are almost equal. Principle Components Analysis is a statistical procedure designed to reduce the data redundancy and put as much information from the image bands into fewest number of components. The intent of the procedure is to produce an image which is easier to interpret than the original.

33 Data Visualization Contrast enhancement or stretch reassigns the DN range that corresponds to the 256 gray shades Top row of images are ETM+ data with no enhancement and bottom row consists of linear contrast stretches of the image DNs to the full gray shades

34 Data Visualization Ability to quickly discern features is improved by using 3-band color mixes Image below assigns blue to band 2, green to band 4, and red to band 7 Vegetation is green Surface water is blue Playa is gray and white (Playas are dry lakebeds)

35 Color display Rely on display hardware to convert between DN and gray levels Digital Numbers (DNs) are image data Grey Levels (GLs) are numerical display values Look-Up Tables (LUTs) map DNs —> GLs and change image brightness, contrast and colors Actual displayed colors depend on the color response characteristics of the display system

36 Data Visualization Changing the color assignment to red, green, and blue does not alter the surface material only appearance of the image All images below show only combinations of bands 2, 4, and 7 of ETM+

37 Data Visualization Other band combinations of the same data set bring out different features (or in some cases lack there of) All images below show only combinations of bands 2, 4, and 7 of ETM+

38 Multispectral display - CIR
• Visualize spectral content with 3- band color composites • Example: color infrared (CIR) – red channel assigned to near IR sensor band – green channel assigned to red – blue channel assigned to green • vegetation appears red, soil appears yellow - grey, water appears blue - black

39 Image formats

40 File formats File formats play an important role in that many are automatically recognized in image processing packages Makes life very easy Raw data typically have no header information GeoTIFF is a variant of TIFF that includes geolocation information in header ( HDF or Hierarchical Data Format ( is a self- documenting format All metadata needed to read image file contained within the image file First developed for web sites in the 1980s Allows for variable length subfiles EOS-HDF is NASA version ( NITF National Imagery Transmission Format ( Department of Defense

41 Data processing levels
Recently, operational processing of remote sensing data has led to multiple processing levels “Standard” types of preprocessing Radiometric calibration Geometric calibration Noise removal Formatting Generic description Level 0: raw, unprocessed sensor data Level 1: radiometric (1R or 1B) or geometric processing (1G) Level 2: derived product, e.g. vegetation index


Download ppt "Lecture 6: Image Processing and Interpretation"

Similar presentations


Ads by Google