scatter plot of whole image - R^2 calculation?

Prev Go to previous topic

Go to next topic

Last Post 02 Apr 2012 06:50 PM by anon

6 Replies

Sort:


You are not authorized to post a reply.

Author

Messages

anon

New Member

Posts:

02 Apr 2012 06:50 PM
Hi, help is needed. Keywords: Scatter plot, R^2 and NDVI An example to understand the problem: I have a Band 3 SLC-on and for the same scene 2 different gap-filled files (done by different methods) i want to compare SLC-on with gapfill1 and SLC-on with gapfill2 I want to make a scatter plot to compare them (see which gap-fill procedure is best). When i do so in ENVI, i only get the scatter plot of a small region - how can i get the scatter plot of the whole image?? (here is what envi help says about the scatter plot: "The dancing pixels in the display are the highlighted 2-band pixel values found in a 10-pixel by 10- pixel region centered on the cursor") next point, how can i calculate the R^2 in ENVI? I want to do the same procedure (page: 6 - chapter: 3.Results - Figure: 3) as described in the pdf which u can find at the following link Sorry for the enormous link, but this is what google gives me :) http://www.google.com/url...jaWARkVYJcDA&cad=rja I will then compare the original SLC-on NDVI with the resulting NDVI of gap-filled files. I think using the scatter plot and R^2 calculation might show which gap filled process gives the "close to reality" results (as described in the pdf, chapter 3). I will be thankful If anyone can find some time and ambition to solve this issue. Kind regards for this great and helpful forum Kchien

Deleted User

New Member

Posts:

03 Apr 2012 06:18 AM
ok, so far the only information about scater plotting the whole iamge window is: http://www.exelisvis.com/...855/scope/posts.aspx that envi doesnt provide this function. Well... still, cananyone suggestion an solution for calculating the R^2 for two images? Kind regards, kchien

Deleted User

New Member

Posts:

03 Apr 2012 11:07 AM
It's true that ENVI's scatter plots can only show you the plot for the part of the data that is in the main Image window, not the whole dataset. Look for that to change in the new GUI with ENVI 5. ENVI doesn't have a tool to calculate R squared. But in IDL you can use the CORRELATE routine to get Pearson's correlation coefficient, which I think is R. Then square that to get R squared. Regards, Peg.

Deleted User

New Member

Posts:

03 Apr 2012 02:08 PM
thanks for your reply peg, i was told that IDRISI is able to calculate the regression of two images. Now i got the software IDRISI Selva, but cant figure out how i can import my .hdr files from ENVI to IDRISI. Any suggestions about that? I will try the IDL routine, thanks. regards, ben

MariM

Veteran Member

Posts:2396
Veteran Member

03 Apr 2012 02:19 PM
In ENVI, you can use the Basic Tools ->Statistics to calculate covariance which includes a correlation matrix (R). These values can be squared to get the coefficient of determination (R^2). I also created a user function a long time ago that will scatterplot two bands using every 10th pixel and provides correlation (R), coefficient of determination (R^2) and the regression coefficients. You can access it here: http://www.exelisvis.com/...tween-two-bands.aspx

Deleted User

New Member

Posts:

03 Apr 2012 05:29 PM
Hi MariM, Thats exactly what i was looking for, thank you. Your scatter plot works fine and gives good results. Only thing what i am missing is, that it is not possible to select bands from different band groups. To do so, i need to save the bands i need into a file group, right? Thanks you two for the support.

MariM

Veteran Member

Posts:2396
Veteran Member

04 Apr 2012 07:06 AM
Yes the bands you want to scatterplot must be in the same file just like ENVI's regular scatterplot tool.

You are not authorized to post a reply.