IPB

Welcome Guest ( Log In | Register )

9 Pages V  « < 7 8 9  
Reply to this topicStart new topic
Juno perijove 6, May 19, 2017
Gerald
post Jun 17 2017, 11:06 PM
Post #121


Senior Member
****

Group: Members
Posts: 1816
Joined: 7-December 12
Member No.: 6780



Thanks! That's currently on rank 4 of my planned Juno activities. I'll probably just invalidate these patches, and reconstruct the missing color channel by the color of the immediate environment. Fixing this is on my plan since more than a year. I hope, that early next week, I'll get it implemented.
Go to the top of the page
 
+Quote Post
Sean
post Jun 21 2017, 09:22 AM
Post #122


Member
***

Group: Members
Posts: 366
Joined: 10-November 15
Member No.: 7837



Some image processing & retiming tests on Gerald's animation for perijove 06...

Click thru each image for a video









--------------------
Go to the top of the page
 
+Quote Post
Bjorn Jonsson
post Jun 22 2017, 11:49 PM
Post #123


IMG to PNG GOD
****

Group: Moderator
Posts: 1881
Joined: 19-February 04
From: Near fire and ice
Member No.: 38



I have been running into an interesting issue when processing some of the images. Here is an example, an enhanced crop from image PJ06_119:

Attached Image


The arrows point in the direction of a color discontinuity. The color change at left is because of a drop in the green intensity to the right of the discontinuity and the color change at right (which is more subtle) is because of a drop in the red intensity. These color changes are not a real feature (I verified this by checking image PJ06_118 which also shows this area).

As a sanity check I checked Gerald's images and this also appears there. It occurred to me that this might be a consequence of how the images are decompanded. In the original, raw image the intensity near the color change is ~210 in both cases (green and red). Keeping this in mind the companding table is rather interesting. Below are a few lines from the table. The columns are 8-bit_SQROOT_Value, 12-bit_Linear_Value and the change from the previous value in the 12 bit column.

207 1375 16
208 1391 16
209 1407 16 <---
210 1439 32 <---
211 1471 32
212 1503 32

Note that the difference between adjacent values jumps from 16 to 32. Of course this is normal but I strongly suspect that this creates visual artifacts (and maybe the human visual system also exaggerates the difference because of a Mach band effect or something similar).

I plan on exploring this further to see if I'm right (I'm not yet completely sure). I may even experiment with a slightly modified companding table. Strictly speaking this would result in slightly less accurate images where the intensity is ~210 but aesthetically they would look better if I'm successful.

I'm interesting in knowing if someone else (Gerald?) has also explored this issue.
Go to the top of the page
 
+Quote Post
Gerald
post Jun 23 2017, 08:46 AM
Post #124


Senior Member
****

Group: Members
Posts: 1816
Joined: 7-December 12
Member No.: 6780



I've noticed this before, but I didn't yet narrow it down to a candidate cause as good as yours. Before the decompanding table became available, I've run an empirical / statistical analysis on the red/green and blue/green quotients as a function of the green value, and found smoothed curves where the decompanding function has kinks between linear fragments. I attributed this to a smoothing effect caused by my analysis method. But with your observation, it might be worth to run a new analysis with formally decompanded values. I did so when I used the decompanding table for the first time, and it looked so much better than before, that I was happy, and didn't persue this any further.
When preparing for my EPSC talk about Jupiter's reflectance function (finally accepted!), and haze, I'll take another very close look at the effect of solar incidence angle and emission angle on brightness and color. I'd think, that any errors in the decompanding function should show up as anomalies.

Btw, Seán, your processing tests are really very! promising.
Go to the top of the page
 
+Quote Post
mcaplinger
post Jun 23 2017, 04:59 PM
Post #125


Senior Member
****

Group: Members
Posts: 1573
Joined: 13-September 05
Member No.: 497



QUOTE (Bjorn Jonsson @ Jun 22 2017, 03:49 PM) *
I'm interesting in knowing if someone else has also explored this issue.

Since the companding maps all input values from 1392 to 1407 to 209 (for example) there's no way to uniquely recover the input, obviously. This is more obvious for Junocam where the companding is done in a piecewise linear fashion and there is a slope change at 209-210 (in contrast to MSL where a smoother full table is used). At various times people have talked about intentionally adding gaussian noise back into the decompanded values, but I've never tried this. Depending on where in the processing flow you do the decompanding relative to other operations (photometric removal, for example) and what precision you are doing this to, you may or may not be quantizing the values further.


--------------------
Disclaimer: This post is based on public information only. Any opinions are my own.
Go to the top of the page
 
+Quote Post

9 Pages V  « < 7 8 9
Reply to this topicStart new topic

 



RSS Lo-Fi Version Time is now: 25th June 2017 - 07:02 AM
RULES AND GUIDELINES
Please read the Forum Rules and Guidelines before posting.

IMAGE COPYRIGHT
Images posted on UnmannedSpaceflight.com may be copyrighted. Do not reproduce without permission. Read here for further information on space images and copyright.

OPINIONS AND MODERATION
Opinions expressed on UnmannedSpaceflight.com are those of the individual posters and do not necessarily reflect the opinions of UnmannedSpaceflight.com or The Planetary Society. The all-volunteer UnmannedSpaceflight.com moderation team is wholly independent of The Planetary Society. The Planetary Society has no influence over decisions made by the UnmannedSpaceflight.com moderators.
SUPPORT THE FORUM
Unmannedspaceflight.com is a project of the Planetary Society and is funded by donations from visitors and members. Help keep this forum up and running by contributing here.