Digital Image Processing Lecture Notes PDF

 PIXELS smallest individual element in an image, holding quantized values that represent the brightness of a given color at any specific point.  DIGITAL DATA: Introduction Digital form – Brightness expressed in array of numbers Brightness can be mathematically...

 PIXELS smallest individual element in an image, holding quantized values that represent the brightness of a given color at any specific point.  DIGITAL DATA: Introduction Digital form – Brightness expressed in array of numbers Brightness can be mathematically manipulated Increases ability to display, examine, and analyze remotely sensed data Each pixel represents brightness value  Band Interleaved by Pixel (BIP) Line 1, pixel 1, band 1 Line 1, pixel 1, band 2 Line 1, pixel 1, band 3 Line 1, pixel 2, band 1 Values for all bands written before values for next pixel Band Interleaved by Pixel (BIP)  Band Interleaved by Line (BIL) Line 1 for band 1 Line 1 for band 2 Line 1 for band 3 Line 1 for band 4 Line 2 for band 1 Line 2 for band 2 Band Interleaved by Line (BIL)  Band Sequential (BSQ) Each band treated separately Band 1 Band 2 Band 3 For many applications Most practical structure  RESOLUTION Ability of imaging system to record fine detail in a distinguishable manner Practical limit to level of detail possible from aerial/satellite image Affected by character of scene, atmospheric conditions, illumination, experience and ability of interpreter  RESOLUTION SPATIAL RADIOMETRIC Fineness of spatial Ability of sensor to detail visible in record many levels image of brightness Fine detail – small 8 bit → 28 BVs (0- objects can be 255) identified  SPATIAL RESOLUTION  MIXED PIXELS Pixels not completely occupied by single, homogeneous category Contain energy from more than one feature RADIOMETRIC RESOLUTION  RESOLUTION SPECTRAL TEMPORAL Ability of sensor to Ability to record define fine sequence of images wavelength over time; how intervals frequent SPECTRAL RESOLUTION  High Resolution: < 24H – 3D TEMPORAL Medium Resolution: 4D – 16D RESOLUTION Low Resolution: > 16D  LANDSAT 7/8: 16 Days Sentinel 2: 10 Days SPOT 5: 2 – 3 Days TEMPORAL IKONOS: ~3 Days RESOLUTION QuickBird: 1 – 3 Days NOAA-AVHRR: 12 Hours Himawari: 10 minutes DIGITAL IMAGE PROCESSING Manipulation and Interpretation of digital images Data are inserted in an equation or series of equations Computation results typically form a new image/s  Types of Image Processing IMAGE RECTIFICATION AND RESTORATION Correct distorted or degraded image data Geometric distortion correction, radiometric calibration, noise removal Preprocessing operations  Types of Image Processing IMAGE ENHANCEMENT More effectively display or record for interpretation Derive images to increase the amount of information visually interpretable from the data  Types of Image Processing IMAGE CLASSIFICATION Quantitative techniques for automating the identification of features in a scene Spectral pattern recognition; Spatial pattern recognition (optional) Additional Steps DATA MERGING AND GIS INTEGRATION HYPERSPECTRAL IMAGE ANALYSIS  Image Rectification Typical Operations Geometric Radiometric Noise Correction Correction Removal **Preprocessing Bring image into Adjust DNs for operations change registration with effect of hazy data and may introduce maps/images artifacts not atmosphere immediately observable  PROCESS WORKFLOW GEOMETRIC CORRECTION  Geometric Distortion No remote sensing images are free of geometric SYSTEMATIC distortions Inherent in remote sensing images fall into two NONSYSTEMATIC categories  Geometric Distortion Systematic Distortion Factors Panoramic Platform Earth’s Distortion Velocity Curvature Earth’s Scan Mirror Scan Rotation Skew Velocity  Geometric Distortion Distortions can be rectified using knowledge of internal sensor distortion. Panoramic Distortion Due to spacing of detectors and regular sampling Produces along-scan distortion Increases with swath  Earth’s Rotation Due to the time taken to build an image as the sensor scans the earth surface features Each line offset to the west from the previous one Earth’s Rotation  Scan Skew Due to the forward motion of the platform during the time required for each mirror sweep.  Nonsystematic Distortions Sensor system’s attitude and altitude Corrected using of Ground Control Points (GCPs) Relief displacement due to terrain variation is the most serious of the displacement types, especially in mountainous terrain. Sensor Attitude and Position Sensor Attitude and Position Sensor Attitude and Position Sensor Attitude and Position  Terrain Related Distortions Due to small changes in aspect Corrected by Orthorectification Requires a digital elevation model (DEM) Can be corrected using a “rubbersheet” rectification based on ground control points. IMAGE REGISTRATION  1. Deterministic Approach Establishes models for the nature and magnitude of the sources of distortion and uses these to establish correction formulate Relies on data of the flight parameters and the terrain information  2. Statistical Approach Uses GCP data set Establishes mathematical relationship between image coordinates and their corresponding map coordinates using standard statistical procedures. Relationships is used to correct the image geometry irrespective of the source and type of distortion.  2. Statistical Approach Polynomial Trend Mapping (PTM) Employs polynomial regression equations to relate image coordinates and their corresponding map coordinates. Georeferencing Assigning map coordinates to image data May be projected but not referenced to a CRS Rectification Involves Georeferencing All map projection are associated with map coordinates. Image to Image Registration Involves georeferencing only if the reference image is already georeferenced  ORTHORECTIFICATION Corrects for terrain displacements using a DEM of the study area Based on collinearity equations, derived by using 3D GCPs Orthorectification is recommended in mountainous areas Orthorectification  When to Rectify? Necessary in cases where the pixel grid of the image must be changed to fit a map projection system or a reference image.  When to Rectify? Comparing pixels scenes - change detection Identifying training samples - classification Creating accurate scaled photomaps Overlaying an image with vector data Comparing different scaled images Mosaicking images  Rectification via GCPs GCP Rectification (Star & Estes 1990) Localize visible points in the images to same points in reality (maps) Establish relationship – Image and Map Coordinate System Form polynoms and their coefficients calculated by regression  Rectification via GCPs GCP Rectification (Star & Estes 1990) Error is given as RMS error (Root Mean Square) Difference between output for a GCP and real coordinates for the same point Interpolation to determine values for the new grid point.  Ground Control Points Specific pixels in an image where output map coordinates are known. Consist of two X,Y pairs of coordinates: 1. Source coordinates: Coordinates in the image being rectified 2. Reference coordinates: coordinates of reference image to which the source image is being registered  2D Transformation Translation Rotation Scaling  Residuals Distances between the source and retransformed coordinates in one direction. If the GCPs are consistently off in either the X or the Y direction, more points should be added in that direction. RMS Error  RMS Error Tolerance Advantageous to tolerate a certain amount of error rather use a more complex transformation. RMS error tolerated can be thought of as a window around each source coordinate, inside which a retransformed coordinate is considered to be correct **If the RMS error tolerance is 2, then the retransformed pixel can be 2 pixels away from the source pixel and still be considered accurate.  RMS Error Tolerance Error is determined by end use, data type, GCP accuracy and ancillary data used RMS error is in PIXELS If using LANDSAT and want the rectification to be accurate to within 30 meters, the RMS error should not exceed 1.00.  Options in Rectification 1. Remove GCP with highest RMS Error 2. Tolerate a higher amount of RMS error 3. Increase the complexity of transformation 4. Select only the points which you are most confident  Rubber – Sheet Image is stretched to fit most of the coordinates. Important to have many GCPs with good coverage Process builds a numeric coordinate transformation between the original and the rectified map. Nonlinear Rubber – Sheet RESAMPLING METHODS  Resampling The process of extrapolating data values to a new grid. Calculates pixel values for the rectified grid from the original data grid. Images treated as simple array to be manipulated to create another array  Nearest Neighbor Uses the value of the closest input pixel for the output pixel value Output values are the original input values Recommended before classification  Bilinear Interpolation Weighted average of four nearest neighbors Advantage: Image has more “natural” look Brightness values of original scene lost Decreased resolution by averaging over areas – gives kind of smearing effect Computationally more expensive than nearest neighbor  Cubic Convolution Uses weighted average of values of 16 pixels Images generally more attractive than other procedures Data altered more Computations more intensive and number of GCPs is higher