4

I need to project some ASTER data from a geographic coordinate system to a projected one. I have read online that the cubic resampling technique is the most appropriate for this purpose. However, using this technique results in the output having negative values, which I cannot have. ArcGIS help instructs the user not to use the nearest, or majority techniques for the projection of continuous data. This leaves only the bilinear resampling technique, which does not have the negative value problem associated with the cubic technique. So, is the Bilinear resampling technique appropriate in this case?

EDIT: I have said that I am using this ASTER data to map a lithic formation which is composed of chert. I am also attempting to use it to discern between dolomitic formations which alternate betweeen chert rich and chert poor.

  • I think a better approach would be to investigate why you are negative values. Where are they coming from? – If you do not know- just GIS Sep 12 '14 at 16:38
  • @user The negative values arise because cubic convolution actually extrapolates the data a little bit in certain regions in order to maintain the full variation in the dataset. In regions with near-zero values and sharp variation from one cell to the next, cubic convolution will interpolate slightly negative values. There are some diagrams of this phenomenon at quantdec.com/SYSEN597/GTKAV/section9/map_algebra.htm. – whuber Sep 13 '14 at 19:33
  • Thanks whuber, When I view the ESRI help webhelp.esri.com/arcgisdesktop/9.3/… it appears negative would not be possible unless present in the original data (and the person asking is using Arc). Can you reconcile the differences in the two links? – If you do not know- just GIS Sep 13 '14 at 20:07
4

Oh, I so disagree with @Aaron's assertion that nearest neighbor is the best method. The common ASTER products are in radiance values and as such, are 32 bit floating point. Nearest Neighbor applied to float values will produce bias and artifacts. This is the common bias effect that results in the blocky appearance of DEM's that were reprojected using nearest neighbor resampling.

The negative values, mentioned in the post, are a result of cubic convolution fitting a smooth curve to a local window with a high deviance. Since the algorithm can extrapolate outside of the range of the data then values outside the observed data range can occur, including negative values. This same phenomena can occur with unconstrained splines applied to high deviance data (i.e., extremely different values across small distances). This is not the case with bilinear resampling, which will honor the original data range via a weighted mean of 4 neighboring cells.

The assertion that nearest neighbor will not effect brightness values is just not supported, even with 8 bit integer data which with digital numbers, commonly exhibits a continuous distribution. The nearest neighbor algorithm is really intended for discrete integer data with truncated ranges. Here is ESRI's brief description of the three resampling methods. It sounds appealing that "the values stay exactly the same" but this is exactly what introduces bias.

I believe that bilinear is an appropriate algorithm in this case. However, keep in mind that some degree of smoothing is not an undesirable characteristic and when applied to resampling a continuous distribution is quite inevitable. It is more about the acceptable degree of smoothing. Often cubic convolution obtains a better "fit" to the data but is much more sensitive to outliers than bilinear. There is a trade off and if your preference is cubic but are concerned over negative values you could always truncate the data back to the expected range.

  • 1
    Jensen makes this as an assertion as well. There is no discussion of balancing bias against precision given the algorithm. The term "smearing" has been used in remote sensing for decades and is apt in referring to smoothing and loss of signal. There is just not much out there on the degree that smoothing effects separability, just opinion. Additionally, it is quite inappropriate to apply NN to float data and we both know that, in general, Jensen was thinking DN's. As for ASTER, I am referring to L1B products which should be converted to floating point with a scale factor of 10. – Jeffrey Evans Sep 12 '14 at 22:25
  • +1 For the explanation of how CC may produce negative values. Thanks for your clarification. – Aaron Sep 12 '14 at 22:33
  • 1
    I stand corrected, I found a paper on the effects of resampling on classification. Please note, the authors did not consider nearest neighbor. However, read their justification on dismissing NN on pp.1086. And, oh my, the lead author has an unfortunate name. asprs.org/a/publications/pers/96journal/september/… – Jeffrey Evans Sep 12 '14 at 22:33
  • A good read... The paper you reference (Dikshit and Roy, 1996) states:" it is recommended that nearest-neighbor resampling is not used for applications where the textural properties of the image are important." – Aaron Sep 12 '14 at 22:47
  • Yes, keep in mind that "textural properties" in 1996 often referred to image contrast which is exactly what you are referring to in "small changes in brightness". – Jeffrey Evans Sep 12 '14 at 22:55

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.