Download presentation
Presentation is loading. Please wait.
Published byMelvyn Porter Modified over 6 years ago
1
Image fusion Goal: Combine higher spatial information in one band with higher spectral information in another dataset to create ‘synthetic’ higher resolution multispectral datasets and images
2
Motivation: why fuse ? Sharper image resolution (display)
Improved classification (and others) More data types and bands available now .. because we can
3
Sensors Panchromatic higher spatial resolution, visible wavelengths
some include NearIR since 1999 Why are they higher resolution ? Multispectral: Lower spatial resolution visible, NIR, … MIR, Thermal May include MIR depending on date (more recent) and with a lower resolution (why ?)
4
SPOT technical band specs
SPOT 1-3 HRV m m Vis,NIR SPOT m m MIR SPOT /5m 10/20m Pan band is in the visible spectrum (.5-.7) … ( ) Geobase orthoimage program Coverage of Canada using a combination of SPOT 4 and 5 giving Pan and MS 10 /20m
5
‘Medium-High’ resolution sensors: Landsat
Platform/Sensor date PAN MS (m) Thermal Landsat TM 4/ / Landsat ETM (-03) Pan is Visible + NIR ( microns) [Landsat MSS – 80m Vis/NIR ]
6
Cranbrook Hill, SPOT image 10m PAN and 20m MS
7
FUSE and IMGFUSE
8
IMGFUSE 432 and 234 432 RGB RGB
9
Cranbrook Hill – famous spots
1. Dreamcatcher movie 2003 (cabin) Cranbrook Hill airport ? Maps/GIS
10
SPOT PAN
11
Pan – Landsat 7
12
SPOT 20m MS
13
FUSE
14
NIR/Red ratio – SPOT 20m MS data
15
NIR / Red SPOT fused 10m data
17
RGB and IHS colour models
IHS Interactive colour wheel: RGB model
18
Fusing Method “IHS transformation” RGB image -> HSI Hue,Saturation, Intensity The intensity channel is replaced by the high res (PAN) channel and the transformation is reversed: HIS -> RGB
19
Problems with fuse / IHS
Colour shift Spectral fidelity Partly solved by ?? HIS v FUSE PANSHARP v FUSE IMGFUSE v FUSE
20
NWT Geomatics imagery http://www. gnwtgeomatics. nt
21
Wunderle et al, IJRS 2007: Regenerating forest structure estimation using SPOT-5 pan-sharpened imagery 10m MS bands b. 2.5m fused image Examined texture, and NDMI= SPOT `(3-4)/(3+4) versus forest structure Needed high resolution AND mid-IR ; used PCI PANSHARP
22
Combine from same or different sensors
MODIS (Vis/NIR) MIR 1km TIR SPOT – Landsat Landsat – MODIS MODIS – RADARSAT Orthophotos – Satellite Imagery Max 5: 1 maximum ratio recommended (?)
23
FUSION OF MODIS AND RADARSAT DATA FOR CROP TYPE CLASSIFICATION G
FUSION OF MODIS AND RADARSAT DATA FOR CROP TYPE CLASSIFICATION G. Hong , Y. Zhang , A. Zhangb, F. Zhou b J. Li MODIS: 250 / 500 m res. Radarsat: 12.5 m -> 25m
24
Fusion Technique to Extract Detail Information from Moderate Resolution Data for Global Scale Image Map Production Francis X.J. Canisius, Hugh Turral
25
The Principal Components Analysis method
Run PCA on the Multispectral data Create PC1, PC2, PC3 Replace PC1 (brightness) with the higher res PAN band Retransform PCA (decorrelate) back to RGB
27
Very high resolution sensors
Sensor date Pan MS (res: m) Ikonos Quickbird Orbview RIP 2003 ALOS Worldview Worldview Formosat GeoEye SPOT / Cartosat (no MS) EROS A, B 2000/ / (no MS) Rapideye x
28
(very) High resolution imagery = < 10 metres
Ikonos Quickbird GeoEye Worldview I Worldview II Cartosat 1 (India) EROS A, B (Israel) ALOS-PRISM SPOT 5 (France) Monitor-E (Russia) Korona – keyhole RapidEye 1-5 (Germany) TerraSAR (Germany) Task: research and describe one of these in a 4 slide powerpoint
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.