i have landsat8 and High Resolution Imagery (1m) in the same study area, is possible to merge them to get one imagery so that i can take advantage of the accuracy satellite imagery.

some_text http://rghost.net/8WGsT7Kgy/image.png

some_text http://rghost.net/6N9clcVYQ/image.png

  • 2
    Sounds like pan sharpening landsat.usgs.gov/panchromatic_image_sharpening.php. – user30184 Oct 4 '15 at 21:11
  • Sorry, but pan sharpening is not what you want to do here. This is typically used with a single band of a satellite sensor set with finer resolution than the visible spectrum sensors. You are taking the colour from the coarser resolution image and resampling to the finer resolution cell size and in areas of overlap you are using the cell values of a single band of your good imagery to recolour the coarser imageries single value. No way to create a raster with multiple cell size. Resample your Landsat to your good imageries cell size and then mosaic with the good image as the overwriting value. – Mike Oct 5 '15 at 7:07

It is technically possible to use the pansharpening algorithm with different sensors, and all your tagged software have pansharpening tools (sometime . However, the quality of the outputs will depend:

1) on the pixel number ratio. In your case, it will be very large (15*15 = 225). IMHO this will be too large, in the litterature you hardly find successful results above 6*6=36 (e.g. Bayesian data fusion (available in OTB) of ASTER TIR band)

2) on the correlation between the band. If the bands are not correlated, pushing structure information inside multispectral image will add spatial details but not reliably.

3) the synchronicity: obviously, if the dates are too much different, you will combine values that changed between the two acquisitions.

Therefore, I would not suggest you to try a pansharpening between 1m and 15/30m images. I would rather downsample the 1 m image to, e.g., 5 m as a compromise, or (better) create image-objects of a size close to your pixels based on the 1 m image, then use the landsat reflectance to enrich your objects characteristics.

As a final remark, you can also "visually" pansharpen your image using some transparancy with your layers. This can help photointerpretation, and your brain will do a better job than any existing algorithm.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.