E N D
1. Washington DC: Jan 20, 2009 Geo-Eye (50cm)http://www.geoeye.com/CorpSite/gallery/detail.aspx?iid=220&gid=1
2. Mapping surface cover types using ASTER data, Dr Abdullah MahEarth Resource Mapping Pty. Ltd., Australia
http://www.gisdevelopment.net/technology/rs/techrs0023.htm
3. Band histograms before and after decorrelation stretch
4. Band correlation before and after decorrelation stretch
5. Decorrelation stretch: Paterson’s curse, Australia
6. Paterson’s curse, Australia
7. Martin-Taylor enhancementCoastal discrimination from visible bands
8. TextureWhile colour and brightness are associated with single pixels, texture ~ graininess, is computed from a set of connected pixels
9. TextureIkonos 1m image: UNBC Radarsat forest image
10. PG- Ikonos data
11. Tone versus texture An area may have the same average DN, but quite different texture, e.g.: (simple example)
5 5 5 1 5 3
5 5 5 5 9 7
5 5 5 3 7 5
Av= 5, Tex=0 Av=5, Tex =8 (?)
e.g. replace central pixel with 3 x 3 kernel DN difference, but if this simple, it would be too much affected by a single ‘outlier’ value
12. Digital analysis of texture Based on the ..
Gray Level Co-occurrence Matrix - GLCM
Haralick, R. M., Shanmugam, K., and Dinstein, I. (1973) Texture parameters for image classification, IEEE Trans SMC 3, 610-621).
Texture Analyser: http://rsb.info.nih.gov/ij/plugins/texture.html
.. Also applied in PCI
13. Gray Level Co-occurrence Matrix - GLCM
14. Gray level Difference Vector: GLDV
15. Texture in image analysis Each pixel at the centre in a ‘n x n’ window is replaced by the texture measure for that window
Window must be odd number- 3, 5, 11 etc
# gray levels may be 16, 32, 64, 128, 256
16. Texture measures (used in PCI) Homogeneity, Contrast, Dissimilarity, Mean, Standard Deviation, Entropy, Angular Second Moment, Correlation, Inverse Difference (same as contrast?)
GLDV
Angular second moment, entropy, mean, contrast
17. What do they mean? Homogeneity: high when GLCM high along the diagonal.
Contrast: the opposite of Homogeneity - is a measure of the amount of local variation in the image.
Dissimilarity : Similar to Contrast. High when the local region has a high contrast.
Mean: Average grey level in the local window.
18. Variables (contd) Standard Deviation :High when there is a large grey level standard deviation in the local region.
Entropy : high when the elements of GLCM have relatively equal values. Low when the elements are close to either 0 or 1 (i.e. when the image is uniform in the window).
Angular Second Moment : This is the opposite of Entropy. It is high when the GLCM has few entries of large magnitude, low when all entries are almost equal. This is a measure of local homogeneity.
Correlation: Measures the linear dependency of grey levels of neighbouring pixels. When the scale of local texture is much larger than the studied distance, correlation is typically high.
Inverse Difference : Measures image homogeneity. It is high when most of the occurrences in the GLCM are concentrated near the main diagonal.
19. GLDV variables GLDV Angular Second Moment : Similar to Angular Second Moment, it measures the local homogeneity.
GLDV Entropy : This is the opposite of GLDV Angular Second Moment.
GLDV Mean : is mathematically equivalent to the Dissimilarity measure above.
GLDV Contrast: is mathematically equivalent to the Contrast measure above.
20. So which ones do we use ? Use best logic
Which do researchers use ?
Compute all and see which works best …
21. Texture analysis aids interpretationhttp://www.epmag.com/archives/techWatch/4139.htm R.M. Haralick and his colleagues demonstrated in 1973 the derivation of 14 different measures of textural features from GLCMs.
However, due to redundancy in these statistics, the following four measures generate the desired discrimination without any redundancy:
• Energy (Angular second moment) is a measure of textural uniformity. Energy is low when all elements in the GLCM are equal and shows geometry
• Entropy is a measure of complexity. It is large for surfaces that are texturally not uniform. In such cases, many GLCM elements have low values.
• Contrast is a measure of the amount of local variation in an image.
• Homogeneity is a measure of the overall smoothness of an image. It is high for GLCMs with elements localized near the diagonal.
22. Texture Tutorial http://www.fp.ucalgary.ca/mhallbey/tutorial.htm
23. Principal Components ofGLCM Texture Measures: http://www.fp.ucalgary.ca/mhallbey/longer%20Hall-Beyer_Srivastava_IGARSS_0608.pdf
24. Urban Feature Characterization using High-Resolution Satellite Imagery: Texture Analysis Approachincluding image fusion http://www.gisdevelopment.net/technology/ip/ma04228pf.htm
25. Texture Metric Comparison of Manual Forest Stand Delineation and Image Segmentation www.fs.fed.us/eng/rsac/RS2006/presentations/warnick.ppt
26. Filters http://eoedu.belspo.be/en/guide/filtr.asp?section=3.5
Low pass filters v High pass filters
NASA Tutorial: http://rst.gsfc.nasa.gov/Sect1/Sect1_13.html
27. Spatial frequency
Spatial frequency refers to the number of changes in brightness value, per unit distance, for any area of within a scene. An area with low spatial frequency will have gradual transitions in digital values (e.g. a smooth water surface). An area with high spatial frequency will have rapid change in digital values (e.g. urban areas and street networks).
28. Low pass filters
Low pass filters are used to emphasize low spatial frequency data and to smooth out an image with high spatial frequency. This may cause the image to be blurred.
29. Low pass filters Mean
Median
Mode
Gamma (radar)
consider this 3 x 3 array where the central pixel (50) will be replaced in a filter by the resulting value:
10 10 mean = 21.66, median = 15, mode = 30
50 15
30 30 15
30. High pass filters
High bandpass filters will enhance differences, while texture filters will measure the degree of heterogeneity
High pass filters, are used to emphasize high spatial frequency data. Often they are used to enhance and sharpen features such as roads, land water boundaries and slowly varying components of an image.
Edge detection filters emphasize these boundaries and make them easier to analyze. In addition edge detection filters can be manipulated to draw out direction and sun angle characteristics.
31. www.sli.unimelb.edu.au/people/jhl_stuff/RS-Filters.ppt
32. Low pass (above) and high pass (below)
33. High pass filters (PCI) Laplacian edge detector
0 1 0 -1 -1 -1
1 -4 1 -1 8 -1
0 1 0 -1 -1 -1
Sobel Edge detector
-1 0 1 1 2 1
-2 0 2 0 0 0
-1 0 1 -1 -2 -1
Prewitt edge detector
-1 0 1 1 1 1
-1 0 1 0 0 0
-1 0 1 -1 -1 -1
Edge sharpening: average filter, result subtracted from the original, then the difference added back in
34. Sample ‘custom’ high pass filters