Timeline for Transforming Non-Uniformly Sampled Image onto Rectangular Grid
Current License: CC BY-SA 4.0
11 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jul 19, 2018 at 20:08 | comment | added | Stephen Hartzell | Yes, I want to downsample. Thanks for your help and all the detail you gave it! | |
| Jul 19, 2018 at 20:07 | vote | accept | Stephen Hartzell | ||
| Jul 19, 2018 at 14:43 | comment | added | Louis Lac | Your problem is closely related to this (quite) old SE question as the algorithm you implemented is similar and targets the same goal (except the fact that you have a non-linear sampling). It contains many pieces of answers like this link that explains Digital Image Interpolation. Last but not least here is the Wikipedia page dealing with Image Scaling. | |
| Jul 19, 2018 at 14:01 | comment | added | Louis Lac | Well, sorry if I'm wrong but your question may be unclear. Based on your comment I think what you actually want is to downsample your original image to a linearly sampled (that is something like 14x14 according to your question). I'm right? Otherwise please edit your question to be more specific. | |
| Jul 19, 2018 at 13:45 | comment | added | Stephen Hartzell | Louis, I understand that interpolation doesn't keep the source point. My concern is that when interpolated to such a lower res grid, that using only neighborhood points from the source image is very unrepresentative. | |
| Jul 19, 2018 at 8:28 | comment | added | Louis Lac | I edited my answer to briefly explain this. | |
| Jul 19, 2018 at 8:28 | history | edited | Louis Lac | CC BY-SA 4.0 |
Illustration and example of resampling.
|
| Jul 19, 2018 at 7:19 | comment | added | Louis Lac | First there is a missconception in your last sentence as in the general case interpolation doesn't "keep" source point that intersect with the grid, it creates new ones bases on the neiborhood and a method (neareast neigbor, linear, cubic...). Secondly if you want to keep as much detail as possible you have to choose a grid with a sampling rate smaller than the highest source image sampling rate. | |
| Jul 18, 2018 at 16:46 | comment | added | Stephen Hartzell | Lous, thank you for the help. I considered interpolation, but I am concerned that it would be a poor sampling. Interpolation is only going to use the points very close to the target grid. Given that the target grid is a much more course resolution. How do I use interpolation to use more of the points in the source image that intersect with the target grid? | |
| Jul 18, 2018 at 12:20 | history | edited | Louis Lac | CC BY-SA 4.0 |
Add details about other function available for scattered data.
|
| Jul 18, 2018 at 12:06 | history | answered | Louis Lac | CC BY-SA 4.0 |