The images I posted above of the calibration dial show the differences in the hue of the four colour 'chips' when different filters are used to produce a colour-composite.

The image that is close to true colour is the L456 version.

What is also interesting about this particular image is the colour of the terrain. At this particular location it is showing up as a 'rusty' or reddish-brown colour but it has to be noted that at other locations the colour of the terrain may be a different colour.

OK OBrien,
thanks.
It would be useful to read about official document of post-stretching of rover raw images because the difference between the auto-streching and the auto-exposure is huge.
Look at the first picture, it is an image with narrow contrast range (with offset and limited brightness).



The next one is an autoexposed (by procedure of rovers):



And the third az an auto-streched variant:



It is well observable that the auto exposure has increased the upper limit of lighting while the stretching has filled out the whole range if lighting.
The auto-exp is only a simple "amplification" therefore it does not change the brightness ratio between the pixels, so an auto-exposed image is usable in color image composition unlike the stretching that changes the ratios so will distort the colors.
There is an other disturbing circumstance with the raw image found in rovers gallery. This is the nonlinearity of images. The original pixel values are 12 bit values but the rover sends them in 8-bit format which is in accordance the usual image displaying on computers. Therefore the rover's on-board computer encode them into 8 bit format by a look-up table. The encoding result will be special: the pixel values in low value range will be divided by less numbers than in the higher range, and the end-value will be divided by 16 (according to 4 bit length-difference). Thus the darker pixels will be lighter than in reality, i.e. the nonlinerized image is unusable for composition true color images. Before using they should be linearized.

OBrien,
I don't believe that the posted two images are same. (it is not your fail )
My opinion is that there is no autocorrect process that can create or remove shadow.
The shadow of the pole is in different position on the images.
Can you cite any document what writes about the post-process of rover raw images?
I have found information about on-board auto-exposure feature that causes quasi stretching.
If auto-stretch process would have been applied for all raw images then we cannot see wrong-exposured images. But we can find a lot of them.

In order to understand how the different filters affect the display of the coloured 'chips' on the calibration dial, I have produced a series of images to demonstrate the effect when different filter combinations are used.

Images L2, L3, L4, L5, L6 and L7 from Sol 50 (Spirit) were selected.

Adobe Photoshop v7.0 was used to produce the colour images.

No colour correction has been applied.


L256




L257




L356




L357




L456




L457




Reference page

http://marsrover.nasa.gov/gallery/all/spirit_p050.html

Reference images:

2P130799733ESF09BVP2111L2M1
2P130799762ESF09BVP2111L3M1
2P130799792ESF09BVP2111L4M1
2P130799823ESF09BVP2111L5M1
2P130799853ESF09BVP2111L6M1
2P130799883ESF09BVP2111L7M1

Exposure times, aperture settings and film speed

http://www.weatherscapes.com/techniques.php?cat=general&page=exposure

There was posted somewhere on this forum, a link to the site which gives exposure times and suchlike for these rover pictures. Does anyone have it saved somewhere, as it may add more information to the discussion.

Marsrock,

the differences  between the three composed images is very simple explainable.
The autoexposure feature of the rover works in the following way:
the exposition is in process until certain percent of the ccd pixels will be saturated.
This means that the exposition time is depend on the filter wavelength and the color of the object to be photographed.
In the case of your images:
in the first images the sky contain a lot of greenish and bluish color therefore the expositon time of the L5 and L6 filter imahes will be similar to the exposition time of the L4's. In these images the pixel valus of the soil-pixels are relatively low.So if you compose directly a RGB images you will obtain bearable color image, but you should to correct it because the color of the soil is false.
In the second and the third images the main part of the image-field is the soil with color  with less blue and green content. Of course the rover has exposed these L5 and L6 images with much longer exposition time because the number of the saturated pixels was very few related to the autoexposure threshold. In these images the pixel values of the soil-pixels are relativeli high. So if you compose from them directly a color image, you will obtain an absolutely false colorized image with bluish-gray colors.
This is why the color-balancing is necessary. The caltarget was usable only in the first 20-30 sols, later it became  too contamined. Now, after 2300 sol it is totally unusable for this purpose. Insted of it use the color of the soil.

Hi TW,

here is the answer and the solution for your false caltarget-strip color problem:



Consider that the blue and green color strips show extreme reflexivity beyond the wavelength 650 nm. This means that the L3 and the L2 filtered images contain unnatural pixel values of green or blue colors although the natura would be any low value. If you apply the L2 filter image as red layer in an RGB image, then you will obtain unnatural color balance , and in special cases like the caltarget the colors will be total false.
Regarding the calibration of the martian colors there is a strange thing what published by Landis from the NASA Glenn Reserarch Center (Cleveland). In the early sols  the martian global spectrum at the Opportunity landing site was very close to the earthian sunlight at noon.
See the following diagram:


The color diagram is the earthian sunlight spectrum on the surface and the dotted curves show the spectral intensity on the Meridiani Planum. They based on the caltarget white-ring reflexivity mesurement in L2...L7 , R3..R7 images.
We can say that at the optical depth value 0.94 (!!), at noon the two spectrum is very close, only at 601 nm wavelength(L4 filter) is a small peak, but at  the L3 filter wavelength the martian spectrum is a bit less than the earthian. In this case the white color consist of 100/98/95 R/G/B respectively instead of 100/100/100. The difference is very small, inappreciable. Therefore if someone create a good color rendering method balanced  the colors of the martian caltarget images to the colors of the earthian colors of the caltarget, good color images will be rendered later.
This is valid for the Opportunity images.
If the martian atmosphere is clear (at low optical depth values such as 0.2-0.3 or less) the martian sunlight spectrum will be more bluish than the earthian because of lack of the filtering particles. In this case the colors will contain more blue component  i.e. the white will be bluish-white, the martian soil will have other hue than in case of tau=0.9 etc.
I think we (and NASA) "over-mysticize" the problem of true color Mars-images. The human eye and brain is "developed" to get information in earthian spectral light. If you see anything in darkness or in disco-light you will examine it also at daylight because the other illumination served false or few information. I think the martian true color is simlar probleme to this: it would be more important to see the martian object in colors belonging to earthian sunlight spectrum instead of martian reddish spectrum. If the martian atmosphere is clear, this condition is satisfied, in other cases can be use white-balancing if it is necessary.

Cheers

xiriux wrote:

---

TW,
you have missed the image of calibration target.
Voila:

---

 

Cheers xiriux.. I posted that image in my earlier posts. I calibrated as best i could to the target but no colors due to surrounding light. Check near bottom. Also when you attempt to calibrate that image you compromise the true time of day lighting.  its darker. You see the problem that there reflective lighting readings are correct (probably).  But because the caltarget is inaccurate they cant get the color balance. Its not the brightness its the color. Its supposed to be dusk or dawn. look at the shadow on the Caltarget. the original image is much darker. The saturation of reds and browns is correct due to the position of the sun (near horizon and surrounding terrain. The air contain quite a bit of airborne dust (being in a desert environment)
.The solar panels are supposed to be much much darker. Not sandy colored. Theyre not absorbing enough light to drive the rovers talk less operate the processing systems..

The faintness in the caltarget is not good enough! You need good strong feed back to get any where close to the true colors. if its covered in dust and partially exposed its useless . the hue has changed resulting in what we have now. Unreliable, suspect, frustrating color images with absolutely no calibrated source. Its a bit like a dog chasing its tail. You think you can see the colors but with an over elaborate filtration system (completely laughable), surprisingly on board (shaking my head with disbelief). A dial that  caltarget system thats just waiting for failure. Why on gods earth would one design a robot that tries to be a processing lab at the same time? running on solar panels? let us on earth to the filtration just transmit back your default ccd data.   

Here's an example of Viking mission problems with this badly designed system...

Solving the color calibration problem of Martian lander images (Proceedings Paper)Author(s): Ron L. Levin; Gilbert V. LevinPaper Abstract

=========

If JPL paid me half what they spent on this appalling system Id of designed something soooo much better! 4 LEDs . Readings at night. leds on short stalks to avoid dust build up. They re low voltage.  Even if there's any build up on the surface they will transmit through the dust a more reliable calibration ref signal. A system of replacements can be factored in. When one fails the next led takes over in circular sector. total cost $3,500. Will last around 15 years. The whole system would sit on a 4-5 sectored disc 30cm across, with  rgb and Y in silver oxide impregnated pigments ( stops radiation bleaching)

O'brien posted

---

I guess that's why the caption on PIA101214, found at http://photojournal.jpl.nasa.gov/catalog/?IDNumber=PIA10214 reads:

"This view combines separate images taken through Pancam filters centered on wavelengths of 753 nanometers [note: a non-visible wavelength], 535 nanometers and 432 nanometers. It is presented in a false-color stretch to bring out subtle color differences in the scene."

Cheers.. FALSe color in big letters lol.. Just a problem still. The target panel does not match colors available in scene. Greens, blues. The target has just a hyper saturated / hue  blue to red????? Really cannot compute to mosaic.  If they input false colors (to the raw data) then why is that not reflected in the Caltarget?


Cheers
Guys
TW

TW,
you have missed the image of calibration target.
Voila:



You can see that it is fairly contamined with dust but the solar cells are relatively clean. It seems that they are after a cleaning cycle.
I should like to draw your attention to the caltarget: it is much less contamined that the Spirit's one.

Mike,
the "blue patch" between the Husband hill and the Home Plate is an sand-patch named as Eldorado. Its color is not blue but sand-color with some bluish and greenish hue in shadowed parts.

Look it:



OBrien,
you wrote:
"While the four images I posted were taken at different times of day, they were taken a few minutes apart, not a few hours apart as I originally thought.

Therefore there are four images taken within a few minutes of each other, all taken with the same camera through the same filter, that have significantly different sky brightnesses at 753 nm.

I don't have an explanation for that, but it is not because of different illumination at different times of day."

You are right, the illumination and the optical depth do not change in 3 minutes ( the time different from the first L2 image to the last one is about 190 seconds).

If you examine the images you can find that the position of the pancam image by image turns toward the direction of the Sun. In the first positon the direction of the illumination is right and from a bit back, in the fourth images it is less by about 50 grads (the FOV of the pancam is 16 grad), so the sunlight illuminates the atmospheric aerosol from other angle. This is the cause  why in the fourth image the sky is much brigther than in the first image.

Everyone knows this phenomenon here on the Earth: if there is clear weather,  the sky is very nice deeper blue if we look it back to the Sun. If we turn toward the Sun the sky becames more and more brighter and whiter because of the light scattering of the atmospheric aerosol. (When the sky is blue, its color contains less red component than green and blue, and when it becames whiter the intensity of the red component (and also the green's ) increases.

My opinion is that in this case we can see this phenomenon in the mentioned Opportunity images.

This  means that in the martian atmosphere are present also particles in size under 1.3 microns causing direction-dependent Mie-scattering, and not only over 1.3 microns-particles which cause direction-independent "white-scattering". It seems that the Rayleigh-scattering of the martian atmosphere can be higher than expected.

 

Just to show (left image) what it should look like and how its changed..



Cheer
Guys

And here's an image taken by Spirit. "Husband Hill" is to the North and about 800 meters away.



http://photojournal.jpl.nasa.gov/catalog/?IDNumber=PIA10214

Now this image is a better calibrated one from JPL that shows plenty of greens and blues! And a pale blue sky to boot! This was probably taken when the sky was clear with no red dust in the atmosphere. You can see the calib wheel in the foreground but alas, no colors can be made out. If you download the 133 MB version, which I don't have the time and inclination to do, you can perhaps get more wheel clarity. Can anyone check this out?

In other words, the true color of Mars is more like this one. The others with the reddish tinge are perhaps due to dust storms. In other words Mars is just not only red!

Cheers!

P.S. Check out the blue patch below the hill! Any guesses what that might be?

Well, at this point, I am convinced we won't get meaningful results from the jpl site images. The next step, one of us needs to link in the pds raw images and see if we are having the same problem working with those.

If it is a result of different auto exposures on the rover - with different exposures on each of the three shots- the same problem will exist in the pds images, and the mosaic will tell us little or nothing.  If that's the case, we'll need a single image (non-mosaic) that contains the callibration target and the sky to  truly resolve the issue.  Otherwise, we're still just guessing at some of the data  - though my experiments are indicating xiriux's guesses are prob somewhat accurate - I'd also like to see his complete shot containing the callibration target and the sky in the same image.

The image description says this was taken around 3:00 P.M. local Mars time, which I suspect is an early dusk - with the sky darkening earlier in the afternoon than it would on earth - due to the distance from the sun.  The auto exposure settings are therefore brightening the image to show us the full scene - and in effect, washing out the color.

Even so, none of this explains the muddy pictures Nasa is giving us since the color can be derived if they are properly using the calibration target.

Hi, The variation in exposures is probably down to exposure bracketing.  It saved my ass on certain difficult jobs in the past. I found this nice simple explanation.

When you expose for a scene, your camera's light meter will select an aperture / shutter speed combination that it believes will give a properly exposed picture.

Exposure bracketing means that you take two more pictures: one slightly under-exposed (usually by dialing in a negative exposure compensation, say -1/3EV), and the second one slightly over-exposed (usually by dialing in a positive exposure compensation, say +1/3EV), again according to your camera's light meter.

The reason you do this is because the camera might have been 'deceived' by the light (too much or too little) available and your main subject may be over- or under-exposed. By taking these three shots, you are making sure that if this were ever the case, then you would have properly compensated for it.

Its a simple solution to just step throuigh exposures on either side of metered exposure. This gives a wider variation for printing the final scene .

Cheers

TW

Okay, I'm getting something inexplicable here.

qmantoo, O'Brien and others can you check the times to make sure these pics were taken near the same time of day?

Something's not adding up here.

Thanks.

They're using L4 L5 and L6.  This is the set - we can check the time stamps to see if they were all taken near the same time of day:

Thanks, Mike and the watcher!  Even though it is a mosaic, it should suit our purposes perfectly - if the caption is correct.  The caption indicates the frames were taken at or near the same moment in time - which means that the lighting will be the same for all - just as if it was a single shot.  I'll work on this one and see what it reveals when I get a chance.  (Opportunity sol 2229).

pia13104 link

This is the caption:  "PIA13104: Two Worlds, One Sun

"While humans' lives unfolded on Earth, NASA's Mars Exploration Rover Opportunity paused in its southward trek and captured this photomosaic around 15:00 local Mars time on May 2, 2010. The timing for this photography with Opportunity's panoramic camera (Pancam) was coordinated with a "moment in time" simultaneous photographic event in thousands of locations on Earth, organized through New York Times photography blog, Lens. (See http://lens.blogs.nytimes.com/2010/05/04/readers-14/.)

Dusty, reddish brown dunes stretch southward to the horizon along the rover's route ahead.

The "Two Worlds, One Sun" theme is a reference to the motto inscribed on the Pancam calibration target, seen on the back of the rover deck at the bottom of this view. The target is used to properly calibrate and color-balance the Pancam images, and with its artistically styled shadow post, or gnomon, it doubles as a sundial (also known as a "Marsdial") for educational purposes. (See PIA05018.)

This scene is a three-tall by one-wide mosaic of Pancam images taken through the camera's red (602 nanometer), green (530 nanometer) and blue (480 nanometer) filters. It has been calibrated and processed to approximate the colors that would be seen by humans if they could be present for this lovely Martian view. The camera took the images during the 2,229th Martian day, or sol, of Opportunity's mission on Mars.

Image Credit:

NASA/JPL-Caltech/Cornell University

Image Addition Date:

2010-05-06

Dear TW,

I'm afraid you have misunderstood me and you are wrong.

1. This color image was rendered from the original L4, L5 and L6 filtered raw images found in gallery of Opportunity. The NASA's PIA13104 is composed from 2 part-images, and this image is the original one of the upper part of that. Ergo I did not crop the calibration target from the photo, because it is visible on an other raw image.
2. Tha raw image called as raw image because its pixels contain the original lighting values derived from CCD of the rover's camera. So it is originally black and white: the ccd pixel has no color.
3. You are wrong regarding to the L2 red-filtered image. If you make a photo with red filter, the value of the ccd-pixel will be higher (whiter on the raw image) if the object is white or red, and it will be lower (darker on raw image) if the reflected or emitted light of the object contains less red light, i.e. it is green or blue or has low intensity.
Look carefully the L2 image: the sky is very dark, but the soil and the stone in the foreground is much more lighter. This means that there is enough sunlight but there is few of red color in the sky. If you examine red-filtered river images on other sols you can see that the sky is as light as the soil, i.e. its color contains fair red component (this occurs for example during dust-storms)
4. I used as color reference the color of the "average soil" The "average soil" can be obtained blurring extremely a large part of the soil in the image. I think it gives good result because the Meridiani Planum has fairly homogen colorization on surface except the stones . The color of the soil is known from any article of Jim Bell (Cornell University) or other authors.
5. You are right I don't know wether raw images are processed or not.  I believe that they are original without changing. The tone correction was processed in the rover using auto-exposure. Although you can find relatively numerous wrong photos.

Cheers
xiriux

xiriux wrote:

---

Some consideration regarding to the imgae PIA13104 linked by Mike:

My opinion is that NASA-published color image has false colors. Why?

See the following image captured on same sol with L2 filter:



It is obvious that on the photo the sky is very dark, i.e. its color contains few red components because the atmosphere is clear. If the martian atmosphere is clear ( the optical depth (tau) is about 0.3 or below) it contains very few particles that filter the sunlight, so the color of the sunlight is very similar to the earthian (a bit more bluish because of the less Rayleigh-scattering of the atmosphere). This circumstance eliminates the color-shifting.
You can see on tha NASA image a permanent orange mist. It is impossible because the original raw images do not show any mist. It seems that this orange mist is artificial.
I have linearized the existing L4, L5 and L6 images, composed the RGB color images then I have balanced the R/G/B ratios according to the R/G/B ratios of the "average soil" (it is approximately 1/0.66/0.35 on the Meridiani Planum where Opportunity landed).  After these operation I have calculated the tristimulus values (interpolating the missing filtered-image values) of the "average soil". Finally I have calculated the  tristimulus values of some sky area and derived their RGB colors from the XYZ values, then I have corrected the color image composed from L4,L5,L6 raw photos.
Voila the result:



Without orange mist and with fairly clear sky.  I am not sure that the turquoise sky is the real color, but let us agree to blush instead of reddish.

---

Interesting.. Just one question xiriux why did u crop out the Calibration disk? I think you should include it in your enhanced images as then we can see what the disks full state. Without external calibration its just a guessing game.. Thats why photgraphers and printers need a color reference... or meter reading.. Filters wont hack it if you dont know the true readings from surface and from sky + a color chart in the said enviroment giving feed back (at source) of what the color balance is for future printing.

It is obvious that on the photo the sky is very dark, i.e. its color contains few red components because the atmosphere is clear.

I think your wrong here xiriux . What does red do in black and white??? Darken tones. There is red ish tones (more tanned ) in the sky. Thats why you have dark tones.

The color of an object depends on the wavelengths of colors reflected from the
object. A red apple is red because red wavelengths in white light are
reflected and other wavelengths are absorbed. In fact, if a red apple were to be
illuminated by light that had no red wavelengths, the apple would appear
almost black. What your black and white has done is remove the color. But thankfully red has a signature in BW.  As a phoptgrapher (many many years ago) I used red filters on blackwhite film to give me those black skies with white clouds.

You can see on tha NASA image a permanent orange mist. It is impossible because the original raw images do not show any mist. It seems that this orange mist is artificial.
I have linearized the existing L4, L5 and L6 images, composed the RGB color images then I have balanced the R/G/B ratios according to the R/G/B ratios of the "average soil" (it is approximately 1/0.66/0.35 on the Meridiani Planum where Opportunity landed).  After these operation I have calculated the tristimulus values (interpolating the missing filtered-image values) of the "average soil". Finally I have calculated the  tristimulus values of some sky area and derived their RGB colors from the XYZ values, then I have corrected the color image composed from L4,L5,L6 raw photos.
Voila the result:

Printing a color image is purely down to averages and aesthetic judgments. the metering is for the 1st generation capture.

You say :
You can see on tha NASA image a permanent orange mist. It is impossible because the original raw images do not show any mist. It seems that this orange mist is artificial.

This is my point in processing. You have a choice of 2 images from JPL. The RAW ( which was processed by a HUMAN! Or the image here in this thread, processed by a Human. Because an image a a raw identity, are you saying it hasnt been tonally corrected? printing will always throw up a wide variety of results if they dont have the CALIBRATION settings. Show me the light readings of the incidental and reflected light from all zones of the image  and I will agree with you.  Unfortunately theres no way you will no that as I think in some cases JPL dont even fully know.

I think what you've attempted is important but unfortunately fruitless as you dont have the essential variables.

Im not JPLs hugest fan but I have worked with the nightmare qualities of light for a number of years. Sometimes complicated results mask a truly simple problem.

Cheers

TW

Heres a case for dust build up..



As you can see. Taking into account color, lighting levels both incidental ambient etc  there is a definite build up of material on the disk but more tellingly on the darker surround (which is a nice gauge for dust levels between the 2 images.
The colors can just be made out but really its a bit of a disaster!
Whats a little more concerning is whats happening to the rest of the robot, especially the solar panels?
TW

mikesingh wrote:

---

 

marsrocks wrote:

---

- and further - why is there no picture from either rover showing the sky and the color callibration target in the same frame?

__________________________________________________________

 

Hi marsrocks,

Here's the image you wanted - the Martian sky and the color calib wheel both in one frame! Heck, this really sucks! No color, just red everywhere! What happened to the colors in the calibration wheel? Why weren't the color channels adjusted to produce the colors on the wheel? Otherwise what's the point in sending it up all the way to Mars?

NASA/JPL

And thanks to all the others posters who have brought out some interesting points!

Cheers!
Mike.

---

Hi guys,
I calibrated as best I could to the color wheel. Unfortunately in the current light in shot its washed out by color temperature of  incidental light. When calib disc is compromised by dust or extreme monochromatic light an incidental dome will help a great deal.. The rovers metering is a bit hit and miss in these conditions. They shoulda got a pro photographer to help setup metering system..  Or the guy they used was hopeless!

In this shot it looks like the Clib disc is covered in dust?  Ive tried to get color from it but its completely washed out!

Cheers
TW

O'Brien, I can see what you are saying and there it appears to make sense, however, I believe it was pointed out previously that the colour patches have been removed or are not there or are not shown now. This was the original way that the colours could be matched because we should have the images of the patches in earth light and the images on Mars. As I understand it, this was the idea to give us a benchmark to go by.

The red filter you used to illustrate the yellow glove is probably extremely red and is not a very good comparison to the red dust either in the air on Mars or a general reddish reflection from the surface. You could also tell us just how 'red' that filter is (although it probably would not mean anything to me, at least)

The last point is this
Measurements are frequently made of the sun with the Pancam and solar filter, to determine the Martian atmosphere's optical thickness. Knowing the amount of dust in the air can be used to calculate the spectral attenuation for "correcting" images...
I have asked before but maybe you did not see it - where can I find information of these sun measurements and how to interpret the tiny pictures that we get back? I assume that the pictures are, in reality, still the 1024 x 1024 pixels that the other pictures are? So why are we getting these tiny representations of a sun as a white disc in the middle of a black background?

xiriux, I agree to a certain extent. But you have not answered the question of how that yellow colored cable on the Rover became red on Mars? I have not 'auto' color corrected the second image where I have adjusted the colors globally on that Mars image to match the cable to yellow. Doing that brought out a set of different colors compared to the NASA/JPL image. The former corrected one looks more 'natural' compared to the latter NASA image.

So the question is, how has the yellow cable on the Rover turned to red in the NASA image on Mars? Is it the red dust? Obviously not! Or the more pertinent question would be, why?

Cheers!

Mike,

You've got some great examples.  Sir Charles also has some examples like this in his CD-book.
http://xenotechresearch.com/bookpromo.htm


One of the most compelling Sir Charles points out is the removal of the mirrors from the color callibration target.  Careful examination reveals that the mirrors - (which would have shown us the true color of the sky as compared with the color chips) - have been replaced by pink or salmon colored rectangles.  Why did they do that?  And why has there been no announcement telling us the mirrors were removed? - and further - why is there no picture from either rover showing the sky and the color callibration target in the same frame?