1

Topic: Correlation Result

Dear Sir,
Correlation  Images have been derived from SPOT-1 (Pre-earthquake) and SPOT-4 (Post earthquake Image) with different Window-Step Size Combination (32-4,32-8,32-16, 64-8 & 64-16). As per my understanding after discarding the SNR value less than 0.9 the best SNR band should be considered as the best result (the respective E-W and N-S band).
1. Actually I have correlated SPOT Panchromatic 10m resolution Images. But I could not figure out which combination (window-step) is providing  the most optimal result as I don't have any field observation data yet.

2.  Is radiometric correction is necessary before analysis as I am working with images from different sensors!

3. Which one is suitable for profile stacking: correlation image or correlation image derived after discarding and destriping. My correlation images have stripes and I have removed it with the aid of ancillary data of SPOT-1 Image.

Please let me inform whether I am in a right direction or not.

Kind Regards
Dewan

2

Re: Correlation Result

Sir,

1. The optimal correlation window sizes will depend on the imagery content and will vary from image to image. Several factors come into play: temporal changes, radiometric noise, illumination difference, texture content... You have to test and see. That said, you shouldn't set up window size smaller than 32 or 16 pixels. Below that, the correlation noise and bias will be large.
2. No, it does not matter. The correlator works on the image frequencies, not on the absolute pixel values. Same thing with the statistic correlation which normalize the data values during correlation.
3. If you are pretty confident that you did not removed any signal of interest during the destriping, it is best to stack profiles on the cleaned correlation maps.

Best,
Francois