Pathfinder Sol 80 Options

[tedstryk]

This is a little pan I put together from the last "end-of-day" image Pathfinder returned of Sojourner on Sol 80. Sol 81 would be the last day for which Sojourner would return images. Sol 76 saw the last data return form the forward looking cameras. The color data was stolen from other sols. Sol 83 would be the last day any data would be returned, and Sol 92 would be the last day any contact would be made. The remaining images on Sols 81-83 were of the windsocks.

Attached thumbnail(s)

 

[Guest_alex_k_*]

Something like this?

[4th rock from th...]

Alex_k, most of what I see on that image is just small scale noise generated by processing.

You are processing the image at a higher resolution than the original.
So any "details" that become apparent but are below the original pixel size are just noise.
I'd recommend using a low pass filter on the end result to remove them.

Anyway, you can more or less trace the petal outlines by small changes on the noise pattern.
Makes sense because even if covered in dust, they are still smother that the surrounding soil.

[Guest_alex_k_*]

Alex_k, most of what I see on that image is just small scale noise generated by processing.

Details of the lander become distinguishable, don't they?

You are processing the image at a higher resolution than the original.
So any "details" that become apparent but are below the original pixel size are just noise.

Actually not a correct conclusion. Because neighbour pixels can also contain information, especially if a camera had a little disfocus. So it's possible to extract information from them - to acceptably approximate the "pixeled" image.

I'd recommend using a low pass filter on the end result to remove them.

If to use a low-pass filter before processing we'll lose details.
You're right that there're some artifacts in the resulted image. They can be avoided by using a low-pass filter after processing.

[4th rock from th...]

Yes, low pass at the end, after all filters.

On this case, the best option would be contour filters and rainbow / spectrum palettes instead of grayscale.
Those bring out structure and small contrast variations without changing the original values.

Exaggerated filtering will give you false details due to pixel value rounding errors that then get amplified over and over.

This is small example. The word "detail" printed on a small font size.
Right top is 3x resample. The "t" starts to show a false curved appearance, but in general, resolution seem to improve.
But heavy filtering (right bottom) gets you ringing artefacts and just false details (for example, a thin line connecting the "a" to the dot on the "i" as appeared.).



Just my 2cents, just trying to help,

[Guest_alex_k_*]

Exaggerated filtering will give you false details due to pixel value rounding errors that then get amplified over and over.

Of couse in general you are right. But it depends on the source image and filters to be appllied. "Over and over" without limitations will get pure noise.
Another question is about information in subpixels - can it be extracted and how.

upd: for testing the method I found appropriate pair of of images, low-res for processing and hi-res for comparing: http://www.esa.int/spaceinimages/Images/20.../Boulder_Cheops

You can estimate whether the details false or not.