Image Analysis in ArcGIS 10

[larryzhang]

Hi, all,


In current Image Analysis in ArcGIS 10, we can do modaicing, clipping, differencing, etc. It is very efficient and adds good values!

Also, I am trying to find out a solution inside, which can derive or combine a new RGB image from two time-serial images under same AOI (Area Of Interest). Saying, the new RGB image is called T3; other two temporal images called T2 and T1. This new T3 RGB image should be derived through the following way (as an option for users to pick up any certain bands from other two images):

1. T3 band1 (R) from T2 band1 (R),
2. T3 band2 (G) from T1 band2 (G), and
3. T3 band3 (B) from T1 band3 (B), etc.

Do we have any way (instruction) to get this kind of operation done from the existing release?

Or, is it feasible for us to have this function sooner?

[RobertBerger]

Hi Larry,

You should be able to composite several datasets easily if you clip them first. So clip your bands based on the AOI, and then composite the three clipped layers.
Is that what you're looking for?

Robert

[larryzhang]

thx, rob,

let me clarify it first!

The 'composite' operation should be equal to 'layer stack' from two images (Erdas Imagine uses this concept, I guess), not 'merging' or 'mosaicing'. With this release, when 'composite', it gives me 'mosaic' image.

For testing, you have two images (T1, T2); and each image has 3 or more bands. After applying for this kind of operation ('composite'), what you should get from those two images is:

1. At least, 6-band new image (T3) is derived; and then you can configure which band for RGB display;
2. Ideally, the interface (option) should allow you to pick certain band from either T2 or T1 image, in order to get this new T3 RGB  image immediately.

[RobertBerger]

Hi Larry,

Please try the following: add the individual bands that you want to work with to ArcMap (you can load the individual bands as layer rather than the whole dataset by drilling into the dataset to see the individual bands). Then manipulate these bands with clip (aoi) however you need to. Lastly, when you have them the way you'd like, composite the layers together into a RGB image through selecting your layers in the Image Analysis Window layer list. Based on how I understand your description of what you are trying to accomplish, this should work.
Let me know if it does or does not work for you.

Robert

[RobertBerger]

Hi Larry,

I came across this which might help you: Some quick video's about how to perform clip with the Image Analysis window (including composite and clip). Hope this helps.

Robert

[larryzhang]

Thx, Robert, really helpful!

It works with raster datasets (cliped ones) for 'layer stack', and then manually choose bands for RGB display, even though it is not very ideal:

         1. It would be nice to have a built-in/hidden 'Clip' operation. In other word, if those two are raster images or image services, it will trigger the hidden 'Clip' running first;

         2. It is also better to allow user firstly to choose which one for RGB display from original two raster datasets, and then directly 'Composite' for RGB display as an OPTION, in order to avoid confusion for some users.

++++++++++++++++

On image services (and mosaic datasets), I tested some areas and then tried to serve the 'Composited ones' onto remote server again as new image service, which do not show very promising. The case is as follows:

From two very large time-serial image services or mosaic datasets under same AOI, Clip, ->Composite, -> Pick bands from Composited image for RGB display, and finally publish it onto remote ArcGIS Server 10 with image extension as a new image service. I assume that they can be managed and accessed by remote ArcGIS Server 10.


        All above (Clip, Composite, the image with choosed RGB display) should be temporary at local geodatabase. If I am right, we may have some efficient/effective ways to make the 'Composited' permernent, and then serve it as new image service. The reason is that all orginal sources for those two image services or mosaic datasets are also located at the same server (SAN storages).

Any advices?

[larryzhang]

Really want to thank SP1 for ArCGIS 10. Those really help us to incease productivity.

With just released SP1 for ArcGIS 10, it is amazing for some good improvements, especially, efficiency and stability, within image analysis and building pyramids on plenty of raw imagery (workspace). In addition, some functions such as 'composite' and 'difference' also can be directly working on larger image service(s) without 'clip' first. Certainly, very fast and productive!

On Mosaic Dataset, however, it is still lacking of effectiveness to do statistics, so that it is really hard to quickly adjust color in the mosaic image, especially, when using the color of reference image (manually-adjusted image) for re-tuning.  From initial results that we got so far, it looks not very promising. Anyhow, we are still testing ...

Spatially, it looks that GCPs (ground control points) of the references are still not considered to re-adjust Mosaic Dataset, while or after 'Geometric Function' is used to remove distortions. In fact, based on our tests, with IKONOS, QuickBird, and GeoEye images, the current function/chain still produce 5-15 pixels shifts, even though high-resolution DEM is used, in particular, densely-populated places or overlapping images acquired on different dates. For GeoEye, 3-10 m shifts are very common, if only RPC & DEM used (without combining surveying and historical spatial references).

Really wish that collecting GCP from references can be available somewhere on some of images and then those GCPs can be used to spatially re-tune mosaic dataset.

[FrancisBoscoe]

I have been trying to use Image Analysis, but all of the sliders, checkboxes and buttons are grayed out. I have tried a number of different rasters from a number of different sources. Same with Image Classification. Any ideas?

[RobertBerger]

Did you select the target layer(s) in the Image Analysis window layer list?

Robert

[FrancisBoscoe]

Did you select the target layer(s) in the Image Analysis window layer list?

Robert

OK, that was easy. What about Image Classification?