Lat long values are assigned to the underlying grid of the image. So it varies along image as well as within a single pixel.(also varies with chosen data frame)

So for assigning single lat-long value to pixel might require to run a algorithm, were you can choose to assign the lat long value at center of pixel on to entire pixel or likewise.

its a complicated and interesting problem!

Dera Gaurav,

if you need feature classes, you have to convert your image to point or polygon features (conversion toolbox -> From Raster). Use from ArcGIS toolbox to add for each point its coordinates (add an empty field and use calculate geometry); for polygons you may use to get the center point coordinates (tool Feature to Point).

Dietrich

Still I do not got it. The Value Attribute Table associated with the image gives you the number of pixels (count) with the same pixel value (value). It gives you no hint WHERE these pixels are located. That means, in your next time series image, the pixels with the same value as in the previous image may have a totally different location.

Thus from my point of view (as I have no idea about your application) it only make sense

- to tag the whole image with time. For that maybe another raster format like NetCDF as more appropriate

- or you have to classify your images, identify areas, vectorize them and tag them with time (or in the extreme tag each single pixel with time)

Regards

Dietrich

Gud day  Gaurav,

You can do raster to point conversion in ArcGIS conversion tools, which leaves you with a point vectors on the center of pixel with band pixel values. Later try adding lat-lon attribute to the vector created using add x-y coordinates tool (Data management toolbox). 

If you can convert this obtained point vector to raster, there is a chance for getting desired layer which you can stack with actual image.. (try feature to raster tool in ArcGIS)

all best 

Akhil is right in his first paragraph, that is exactly what I meant in my first answer: convert each pixel to a point, add the coordinates using the toolbox or using the option in the attribute table for the new point feature class.

Be aware, that you will get some thousands of points, which will be difficult to handle!

Still there is the problem to "track" the same pixel over time! For that I am still not sure if a pixel based analysis make sense or to do a classification, vectorize the resulting areas, and track them over time.

regards

Gaurav Tripathi

How many images (over how many dates) are you trying to track the change? Are the images from the same sensor?  You will need to do the prior procedures for each image.  If you have many, many dates, this will take time.

Have you considered using the change detection button in the Image Analysis window? The images should be from the same sensor suite (i.e. all Landsat, etc) and the same band.  Doing this, it will give you a new image with values that represent the change in pixel value between two images.  If you do this first, you will cut the number of images for which you need to extract the x,y and z values in half. 

Once you have these values, instead of converting each raster to a point, consider creating a fishnet with the image as the boundaries.  When you create a fishnet, you have the option of creating a point file at the same time.  This point file is the center of each grid (i.e. a pixel).   Then you could use this fishnet point file to extract the raster values.  This is less intensive than doing for each image because in Spatial Analyst tools, you have an extract multi values to point tool and can extract all the values from the images at one time.

Are you trying to track change in NDVI, perhaps?   If you have access to Matlab and TIMESAT, this might be better software to use.

Check the Python samples of

http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#//001200000007000000

and

http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#//001700000032000000

These are the tools you need to implement in a loop for all the rasters you have. I think that the Space Time Cubes in ArcGIS are suppose to be used for point events "crime or fire events, disease incidents, customer sales data, or traffic accidents"...trying to implement the Cubes for raster data (cleverly importing them as points) will probably not provide optimum results.

I would follow Tammy's suggestion to use Matlab (or IDL) for this kind of thing.

Georectify the TIFF image with the help of GCPs with lat-long values

I am not sure whether this would work for you but you could consider calculating longitude and latitude in seperate (raster) layers. It is not possible to have this information in the attribute table, but in this case you would end up with all required information in seperate raster files. Now all your gray scale pixel values map to their coordinates in seperate rasters which also makes it easier to use this information in subsequent processing.

Here are several ways to go achieve this:

http://gis.stackexchange.com/questions/138754/getting-latitude-and-longitude-raster-from-dem-using-arcgis-for-desktop

Cheers

If you georeference your image in ArcGIS, each pixel will include geocoordinates. Then, you can add fields of  X and Y with extract tool.   

"After conversion from raster to point, do you think that calculating latitude longitude would work?"

It will be almost instant once you convert to point.

 convert tiff to grid stacks, grid2xyz.