| 21 Oct 2013 10:04 AM |
|
|
I've finally decided to try out ENVI 5 after having SP3 installed (I've been using ENVI Classic as I really have issues with the ENVI 5 interface and (in my opinion lack of) user friendliness, and the fact that ENVI 5 is missing several tools from Classic that I use on a regular basis. My current project is using some Landsat 8 imagery so I was really looking forward to being able to use SP3 updates to do reflectance conversions, TCT, etc. easily on these Landsat 8 images. However, I'm running into several problems/questions I'm hoping the community can assist me with.
1. I was assuming (based on the Help documents), that the SP3 would update the tools to work with Landsat 8. For example, that the NDIV conversion dialog box would have an option for Landsat 8 and automatically update the Red/NIR band selection for Landsat 8. However, the only option I see is Landsat TM and the bands default to the Landsat 7 3/4 combo. Am I missing/misunderstanding how the SP3 update was supposed to work? Is there not an option for Landsat 8 OLI and the band selection to auto-update or are we supposed to manually update the bands? If that is the case, I'll stick with Classic if there are no real benefits to fighting with the ENVI 5 interface and/or be reduced to manually doing the calculations on any of my Landsat 8 images?. Nothing seemed to go wrong in the SP3 update (no errors, hang-ups, etc.). The TCT dialog box parameters also only have Landsat 5 TM, Landsat MSS, and Landsat 7 ETM options, no option for Landsat 8 like I was expecting. The Gram-Schmidt Pan Sharpening also does not appear to have a Landsat 8 option.
2. Is there any way to take polygon ROIs that I created in ENVI Classic and apply them to (the same) image in ENVI 5? While I was waiting to get the SP3 update I created several cloud masks and such on the images in ENVI Classic, I can open the mask bands in ENVI 5 but I cannot get them to show up as a band option when I try to "apply mask" nor can I get the ROIs that I created in Classic to open in ENVI 5 so that I could at least recreate the mask. I can get a mask band that I created in ENVI Classic to show up on the image in ENVI 5, but it will not show that band as an option when I select "Apply Mask" and need to designate the mask band. The Help documents seem to indicate I should just be able to Open a .roi file, and that does place a layer in the view, but nothing is visible on the image (and yes, I have them at the front and checked as visible). I tried doing a build mask by "Import ROI" and selecting the .roi file, but I get an error message stating "There are no available Regions to use" I'm hoping I'm just missing a step that the help documents don't address.
Sorry for the long post (and obvious frustration). Any help/suggestions would be appreciated.
|
|
|
|
MariM
Veteran Member
Posts:2396  
| 21 Oct 2013 10:35 AM |
|
|
Hi there,
I can try to help answer some of these questions:
- You are going to need to use the new ENVI 5 interface to open and calibration Landsat-8 data at the least. There are still many tools in ENVI Classic that are not yet present or functioning as well as they do in Classic but they are adding to the interface all the time. ENVI 5.1 will have a much improved ROI tool that will have the same functionality as ENVI Classic. So hopefully it will be less frustrating as it evolves.
- the NDVI tool was not updated for Landsat 8. I've always thought it was a bit of a poor idea to assume a sensor and thus pick bands from that. It always seemed a better idea to pick bands based on wavelength. This is what the Vegetation Index calculator does so you could try that for NDVI for Landsat 8. Otherwise, you will need to manually select the appropriate bands for your dataset.
- the Tassled Cap Transform was not updated but I don't believe new TC coefficients have been published for Landsat-8 yet. These values are empirically derived and specific to the sensors.
- when I use the Gram-Schmidt Pan Sharpening tool from the toolbox, it does show 'ldcm_oli' as an option.
- classic ROIs can be opened using the standard File->Open. I don't know why your masks and ROIs don't show up and you might want to contact technical support directly about that. Typically it means that the mask and ROIs were built on a an image that had different dimensions than the file you are trying to overlay them on. I would double-check if this is the case.
|
|
|
|
Deleted User
New Member
Posts: 
| 21 Oct 2013 12:38 PM |
|
|
MariM,
Thank you for the quick response. If I'm understanding your answers below correctly, there is almost no benefit to me using ENVI 5 at this point, even though I am using some Landsat 8 imagery, as most of the tools I was anticipating being updated have not been updated yet. Especially will the lack of ROI functionality it may not be worth the time/hassle/headache, as I would have to keep swapping between ENVI Classic and ENVI 5 in order to have most of the functions I need. Other than the GS Sharpening having Landsat 8 (which I did find that lcdm-oli option, I overlooked it originally, sorry) there doesn't appear to be any tools that are updated that I would use.
And in regards to the GS Sharpening, as far as I can tell there is no option in the set-up of that process to allow me to designate a mask layer during the process, so the sharpening keeps running on the full image and I have to remask it in the end anyway. I'll admit I'm not really comfortable with all of the mathematics that are used in the GS process, so I'm not sure what impact it using the whole image vs. masking out the clouds and shadow has on the pan sharpen. As this would be the only ENVI 5 tool I would be using at this point, is there any reason I couldn't just use the ENVI Classic option and select one of the other options in the GS Spectral Sharpen Parameters dialog box other than "Create by Sensor Type" and do the GS sharpen on the Landsat 8 images there?
|
|
|
|
MariM
Veteran Member
Posts:2396
| 21 Oct 2013 12:54 PM |
|
|
I don't believe you can open Landsat-8 datasets using the metadata file in ENVI Classic and you can't calibrate it in ENVI Classic. So if you need to perform these tasks (and most specifically calibration), then you would need ENVI 5.
I don't know why there is no ability to use a mask in the Gram-Schmidt sharpening but I have never used a mask and had not noticed that it provided poor results. I do see that ENVI Classic allows for a mask. I would probably test to see if the results are significantly different between using a mask and not.
|
|
|
|
Deleted User
New Member
Posts:
| 22 Oct 2013 10:23 AM |
|
|
MariM,
Thanks again for your prompt reply. It does appear that there is a slight difference in the pixel values output from a masked vs. unmasked image (so I've got to figure out which is more appropriate and/or if the difference is significant, but that's a different issue). I did want to confirm something that I noticed in the GS Sharpening process where the ENVI Classic and ENVI 5 Help documents are the same but the options/process appears different in the two versions. I'm hoping you can clarify for me.
The Help documents in both versions, the Exelis website documents, (and ENVI Classic's process) refer to selecting the method that is used to create the simulated pan band for the GS sharpening. But the ENVI 5 GS process never gives these options. Can you confirm which method ENVI 5 GS sharpening uses (simulating the pan band based on the lower-resolution spectral bands or by subsampling the designated high-resolution pan band)? Is it simulating it based on the lower-resolution spectral image (as the summary below seems to state) and this selection process was removed but the steps in the Help documents haven't been updated?
For reference, the Help documents state:
ENVI performs Gram-Schmidt spectral sharpening by:
Simulating a panchromatic band from the lower spatial resolution spectral bands.
Performing a Gram-Schmidt transformation on the simulated panchromatic band and the spectral bands, using the simulated panchromatic band as the first band.
Swapping the high spatial resolution panchromatic band with the first Gram-Schmidt band.
Applying the inverse Gram-Schmidt transform to form the pan-sharpened spectral bands.
Step 4: Choose one of the following methods for the low resolution pan:
Average of Low Resolution Multispectral File: Simulates a low resolution panchromatic image by using an average of the multispectral bands. ENVI simulates a panchromatic image for the corresponding sensor.
Select Input File: Select this option, then click Select Low Resolution Pan File to choose an existing one-band image with the spatial dimensions of the multispectral image.
Create By Sensor Type: Simulates a panchromatic image for the selected sensor. Choose the sensor type from the Select Sensor drop-down list. This option requires radiometrically calibrated data.
User Defined Filter Function: Simulates a panchromatic image for the selected filter function. Select this option, then click Select Input Filter Function File to choose a specific filter function to use. This option requires radiometrically calibrated data.
Thanks.
|
|
|
|
MariM
Veteran Member
Posts:2396
| 22 Oct 2013 11:50 AM |
|
|
In ENVI 5, you must select a high resolution pan band so it looks to me to be the only option. I am not sure why all of the options are not available but I can say that sharpening with a high resolution pan band is by far most common method of sharpening. I don't think I ever used the other methods.
|
|
|
|
Deleted User
New Member
Posts:
| 22 Oct 2013 01:54 PM |
|
|
Thanks again MariM,
So you would say that the ENVI 5 process uses the method of subsampling down the high-res pan band (which the user designates) for the simulated pan band step of the GS Sharpening? That's what I assumed/hoped but with the missing/inaccurate Help documents I wasn't sure.
|
|
|
|
MariM
Veteran Member
Posts:2396 |