# Difference between revisions of "Image processing"

## Introduction

For a general overview, see "Introduction: image processing in GRASS GIS" at imageryintro.

### General introduction

Digital numbers and physical values (reflection/radiance-at-sensor):

Satellite imagery is commonly stored in Digital Numbers (DN) for minimizing the storage volume, i.e. the originally sampled analog physical value (color, temperature, etc) is stored a discrete representation in 8-16 bits. For example, Landsat data are stored in 8bit values (i.e., ranging from 0 to 255); other satellite data may be stored in 10 or 16 bits. Having data stored in DN, it implies that these data are not yet the observed ground reality. Such data are called "at-satellite", for example the amount of energy sensed by the sensor of the satellite platform is encoded in 8 or more bits. This energy is called radiance-at-sensor. To obtain physical values from DNs, satellite image providers use a linear transform equation (y = a * x + b) to encode the radiance-at-sensor in 8 to 16 bits. DNs can be turned back into physical values by applying the reverse formula (x = (y - b) / a).

The GRASS GIS module i.landsat.toar easily transforms Landsat DN to radiance-at-sensor. The equivalent module for ASTER data is i.aster.toar. For other satellites, r.mapcalc can be employed.

When radiance-at-sensor has been obtained, still the atmosphere influences the signal as recorded at the sensor. This atmospheric interaction with the sun energy reflected back into space by ground/vegetation/soil needs to be corrected. There are two ways to apply atmospheric correction for satellite imagery. The simple way for Landsat is with i.landsat.toar, using the DOS correction method. The more accurate way is using i.atcorr (which works for many satellite sensors). The atmospherically corrected sensor data represent surface reflectance, which ranges theoretically from 0 % to 100 %. Note that this level of data correction is the proper level of correction to calculate vegetation indices.

### Image processing in GRASS GIS

Satellite imagery and orthophotos (aerial photographs) are handled in GRASS as raster maps and specialized tasks are performed using the imagery (i.*) modules. All general operations are handled by the raster modules.

• Data import is generally handled by the r.in.gdal module

## Importing

The wxGUI offers a convenient tool for single map and bulk import:

### Satellite Data

#### High Resolution Data

Commercial satellite imagery

### Orthophotos

The wxGUI offers a convenient tool for single map and bulk import:

### Imagery groups

A multi-band image may be grouped with i.group (some commands actually require the input being defined as a imagery group rather than a list of map names).

## Preprocessing

### Geometric preprocessing/Georectification

A multi-band image may be grouped and georectified with a single set of ground control points (i.group, i.target, i.rectify).

### Correction for atmospheric effects

Visit the dedicated page on Atmospheric correction

### Correction for topographic/terrain effects

In rugged terrain, such correction might be useful to minimize negative effects.

• simple "cosine correction" using r.sunmask, r.mapcalc (tends to overshoot when slopes are high)
• In i.topo.corr the following correction methods are implemented: cosine, minnaert, percent, c-factor.
• Note, that for the sun's zenith (in degrees) parameter, the equation "Sun's Zenith = 90 - Sun's Elevation" is generally valid

Examples:

## Image classification

See the dedicated Image classification page.

## Image segmentation

• i.smap: Performs contextual image classification using sequential maximum a posteriori (SMAP) estimation
• i.segment: Image Segmentation
• r.smooth.seg: Performs image segmentation and discontinuity detection (based on the Mumford-Shah variational model) (addon)
• i.superpixels.slic: : Perform image segmentation using the SLIC segmentation method (addon)

## Edge detection

• i.zc: Zero-crossing edge detector
• i.edge: Canny edge detector (addon)

## Texture

A series of commonly used texture measures (derived from the Grey Level Co-occurrence Matrix, GLCM), also called Haralick's texture features are available:

• r.texture: In case of panchromatic maps or limited amount of channels, it is often recommended to generate synthetic channels through texture analysis

See here and here for the formulas to calculate texture. See also canopy texture mapping.

## Tasseled cap

• i.tasscap - performs Tasseled Cap (Kauth Thomas) transformation, resulting in 'Brightness' Tasseled Cap component 1, 'Greenness' Tasseled Cap component 2, 'Greenness' Tasseled Cap component 2, 'Wetness' Tasseled Cap component 3, and 'Atmospheric haze' Tasseled Cap component 4.

## Spectral unmixing

• i.spec.unmix is used to perform "Spectral Unmixing"
• i.spectral - displays spectral response at user specified locations in group or images.

### Spectral unmixing ideas for processing hyperspectral image data

• Make use of the Spectral Python (SPy) which is a pure Python module for processing hyperspectral image data

### Spectral angle mapping

• i.spec.sam is used to perform "Spectral Angle Mapping"

## Thermal remote sensing

• r.mapcalc can be used to convert from DN (digital number) of arbitrary sensors to Kelvin/Celsius/...
• i.landsat.toar - Calculates top-of-atmosphere radiance or reflectance and temperature for Landsat MSS/TM/ETM+/OLI
• i.landsat8.swlst (src) Practical split-window algorithm estimating Land Surface Temperature from Landsat 8 OLI/TIRS imagery
• i.aster.toar - Calculates top-of-atmosphere radiance or reflectance and temperature for ASTER
• MODIS

## Enhancements

### Geometric Enhancements - Image Fusion - Pansharpening - Image Segmentation

Image fusion and Pansharpening:

• i.fusion.hpf is fusing high resolution panchromatic and low resolution multi-spectral data based on the High-Pass Filter Addition technique (Gangkofner, 2008).
• i.pansharpen: Image fusion algorithms to sharpen multispectral with high-res panchromatic channels
• i.rgb.his and i.his.rgb: can be used for image fusion

Segmentation:

• i.segment: Identifies segments (objects) from imagery data
• i.superpixels.slic performs image segmentation using the SLIC segmentation method. (Addons)
• r.smooth.seg which performs image segmentation and discontinuity detection (based on the Mumford-Shah variational model). The module generates a piece-wise smooth approximation of the input raster map and a raster map of the discontinuities of the output approximation. The discontinuities of the output approximation are preserved from being smoothed. (Addons)

### Optimal channel selection for color composites

• i.oif: Calculates Optimum-Index-Factor table for spectral bands

## Vegetation indices

• i.vi: various vegetation indices
• r.mapcalc can be used to calculate uncommon vegetation indices

## Water indices

• i.wi: Calculates different types of water indices (addon)

## Evapotranspiration

Please look at Image_processing/Evapotranspiration for some background information.

## Ideas collection for improving GRASS' Image processing capabilities

Below modules need some tuning before being added to GRASS 6. Volunteers welcome.

### libCTL - Library for affine, Helmert and projective transformations in 2D

To be evaluated: plain C translation https://svn.code.sf.net/p/gvsigce/code/trunk/libraries/libCTL/

• requires GNU Scientific Library for the matrix algebra
• It is a small library that provides a handful of transformation methods from source to target (2D) coordinates. Currently, this includes affine, Helmert and projective transformations in 2D.
• The main library is written in plain C and the transformation functions are a plain C conversion of the methods found in the QGIS (www.qgis.org) Georeferencer Plugin (projective and Helmert transformations) and Olivier Dalang's "worldfile transform" (https://gist.github.com/olivierdalang/ba97fc986ade4545068d).
• author: B Ducke

### Image matching ideas

• i.points.auto: automated search of GCPs based on FFT correlation (as improved i.points)
Reference: M. Neteler, D. Grasso, I. Michelazzi, L. Miori, S. Merler, and C. Furlanello, 2005: An integrated toolbox for image registration, fusion and classification. International Journal of Geoinformatics, 1(1), pp. 51-61 PDF

### Stereo ideas

This is stand-alone stereo modeling software (DEM extraction etc). Waits for integration into GRASS.

Needed to orthorectify a series aerial images taken sequentially with overlap. "Historical" method which is nowadays interesting for UAV flights with octocopters and such.

Automatec GPC search could be done by "auto-sift".

Available: Octave code which prepares input to an i.ortho.photo batch job (contact Markus Neteler).

(see LIDAR)

### Improving the existing code

It might be sensible to merge the various image libraries:

• GRASS 6 standard libs:
• lib/imagery/: standard lib, in use (i.* except for i.points3, i.rectify3, see below)
• imagery/i.ortho.photo/libes/: standard lib, in use (i.ortho.photo, photo.*)
• GRASS 5 (! only) image3 lib:
• libes/image3/: never finished improvement which integrated the standard lib and the ortho lib. Seems to provide also ortho rectification for satellite data (i.points3, i.rectify3)
• GRASS 5/6 image proc commands:
• merge of i.points, i.vpoints, i.points3 (see above)
• merge of i.rectify and i.rectify3 (see above)
• addition of new resampling algorithms such as bilinear, cubic convolution (take from r.proj or r.resamp.aggreg) (done 10/2010)
• add other warping methods (maybe lanczos or thin splines from GDAL?): Addons#i.warp
• implement/finish linewise ortho-rectification of satellite data