| Author | Thread |
|
|
03/18/2009 09:36:41 PM · #26 |
| great minds think alike. I just posted a comment on the photo doing something similar. |
|
|
|
03/18/2009 09:44:58 PM · #27 |
Originally posted by jeger:
Okay.
Here is my quick edit.
1 - I converted the profile to sRGB. You should make sure your colour profile is sRGB because otherwise when you save for web, the colours will not be displayed the same.
2 - Noise Ninja to remove some of the noise.
3 - Adjustment Layer for Colour Balance ( -25 Cyan, +30 Blue)
I purposely left it a little warmer because I know that was the look you were going for.
 
What do you think? Can you see the difference? Do you think it's an improvement? |
I do see the difference, I guess for me it looks a little more like the contrast/exposure has more to do with it that the yellow but it is definitely an improvement, seen against the edit the original seems somewhat washed out, i dont notice the color change so much as the contrast difference, is the color adjustment what changed the contrast like that or was that the noise ninja that shifted it(incidently I LOVE the improvement in the noise in the background, I have huge issues with the Bokeh for this lense as it does not translate well down into the jpgs needed for submission, Raws and PSDs look great then when saved down they go to pot, normally only on the rendering of the rokkor lense. How much is the noise ninja tool?)
|
|
|
|
03/18/2009 09:46:19 PM · #28 |
Originally posted by Nobody:
great minds think alike. I just posted a comment on the photo doing something similar. |
I see in yours as well, but its seems kind of harsh at first glance almost TOO white, just my impression though.
|
|
|
|
03/18/2009 09:51:53 PM · #29 |
Yes, the shadows did become darker after I made the colour balance adjustments. Ideally I should have played with the levels afterward to bring out more details in the eyes and the nose.
As for Noise Ninja, you can get it for $45. I think it is worth it. |
|
|
|
03/18/2009 09:57:58 PM · #30 |
| Its great to see how those little tweaks can change the feel of an image, the little details can make or break a shot. |
|
|
|
03/18/2009 10:13:19 PM · #31 |
I did a quick edit, too. No doubt overdone, and too light for your taste, but just to show a couple things. There is a reflection in the eyes that gives them some dimension. There is a fair amount of noise, grain, even after a noise reduction pass.
I think the comment about verifying your color space might be key--it could explain why it looks different to you than to us. BTW, my monitor tends toward too bright in general, and the original image looks very dark to me.
I think the image (as posted) might have been underexposed some, which would contribute to the noise levels. If you can see the histogram on your unedited original, that might be a clue. It was clearly a challenging environment/situation in which to capture the image :-)
Message edited by author 2009-03-18 22:50:31. |
|
|
|
03/18/2009 10:26:41 PM · #32 |
Here is a good idea what noise software can do to change an image, I used Neat Image on a duplicate layer then masked out the area of the face to minimize the effects of the background and make the nose less noticeable. After that only made very slight changes to color and contrast. I wanted to go lighter but you had said you felt this was the color of your dog so my attempt was to lesson the color impact by the background and focus attention on the eyes. |
|
|
|
03/18/2009 10:29:41 PM · #33 |
I understand your frustrations about low scores. But realize a couple of things:
1. With your voting average, you give more votes under 5 than 5 and over. There are people voting that give even lower scores, which explains some of the low scores.
2. I like very shallow depth of field shots, but there are many people who can't stand them. They want realistic photos and don't score photos with any artistic attempts very well--which explains a few more low scores.
3. Some people are continuously complaining about dog and cat (although mostly cat) shots. They're not going to score a dog photo well unless it's something completely different or unusual.
4. Some people are just out to throw things off. (the all white photo had something like ten 10s.) Unfortunately, there's really not that many of these.
Also, you had said:
"as i said before i didnt expect a higher or lower score, this was an emotive shot not a wow shot, but you seemed to imply that the image had a problem of not having any wow to it? is that something to do with the way I shot it(ie is there some way to give it wow that is beyond my knowledge or was it just that it was not a wow subject) "
People were just trying to help you figure out how to do better in the challenge. They didn't realize that you knew where this would place. They assumed that you entered with hopes to win. And they knew that a soft, comfortable image doesn't do as well against and image that will really stick out in someone's mind. I don't think anyone would claim that there's something wrong with a comfortable, cozy shot.
|
|
|
|
03/18/2009 10:36:13 PM · #34 |
Originally posted by jhomrighaus:
1, 1
2, 11
3, 29
4, 82
5, 174
6, 127
7, 57
8, 14
9, 4
10, 8
I mean seriously, how can you be getting 7, 8 , 9 , 10s and be getting 3s, 2s, and 1s, I have never seen such a heavy set of low votes with the curve skewed to the higher side like this. Sure a token few trolls that have some kind of issue but 10% voting below 3, this implys to me at least that the photo was horrible, but the 5 6 7 8 scores say its an average to above average photo.
Arrrggg this place can be soo annoying. |
===================
Back to your very first posting....I understand your initial frustration with the wide range of scores. I had a similiar question about the 1's, 2's and 3's I received with my cardinal entry... It "feels" like some are forcing the average lower...maybe to reduce the competition?? But then I keep telling myself that everyone has their own "interpretation" of a photo - and that for the most part this is a great group of people that really want to help each other improve their skills.
Just keep shootin...
Drake |
|
|
|
03/19/2009 11:48:24 AM · #35 |
Originally posted by jhomrighaus:
if I take the camera into the room and shoot a white balance on a white card before I start shooting then I am very confused at the comments stating that the color is off. |
I was just looking at the EXIF data of your shot, and it says:
White Balance = auto (0)
Unless I'm reading that wrong, it would appear that you had it set on auto white balance. You need to set the camera to "PRE" and then use the preset feature to set the white balance with a balance card or ExpoDisc, and keep it on that setting. If you indeed shot this in auto under room lighting, it's not surprising that the colors were off. |
|
|
|
03/19/2009 12:07:14 PM · #36 |
Originally posted by jhomrighaus:
Here is an outtake from the same session with a wider view of the room/dog, is this photo too yellow, too orange? How would you adjust it? |
All you need to know about your color balance is in this shot: see the OOF light source in the BG? See the yellowish cast to it? If WB were correctly adjusted to this light, that cast would not be there. And that you can see the cast at all int he actual light *source*, which is totally blown out, is a measure of how severe the cast is. Here's the image neutralized:
Now, I opened your submitted shot in CS3, and opened the edited outtake, where I balanced the color off the visible light source. I duped the base layer on the submitted shot, then did a "match color" adjustment, specifying as the source the color-balanced outtake version. I thens et the match-color layer to "color" mode, and this is the result:
I have owned goldens myself, and this looks MUCH more natural to me. In fact, toggling between them, the original looks more like an irish setter, color-wise, than a golden retriever, to me. Lovely dog, btw :-)
R.
Message edited by author 2009-03-19 12:19:22.
|
|
|
|
03/19/2009 12:11:22 PM · #37 |
| In the submitted shot, looking at the catch-lights made me think the wb was off. They look yellowish, not white. |
|
|
|
03/19/2009 12:21:30 PM · #38 |
I actually gave the shot a 7. A few thought about my opinion.
1) There is a cast, of course, but I think it works as well as the color corrected shot. It's warm and warm colors are intimate, friendly, and loving. The cast also works fairly well because there are no true whites in the shot. Karmat just mentioned the catchlights, but they are small. The blacks are dark enough to appear truly black.
2) The noise is mild and gives texture to the picture. For those who think the image is degraded by it, I'd counter and say your "noise radar" is set a little too high.
Anyway, everybody has their opinion and the voters have their say at the end of the day. 5.3? I can see it. It's a subtle, intimate shot that clearly means more to the owner of the dog than to anybody else. |
|
|
|
03/19/2009 12:27:16 PM · #39 |
You've gotten a lot of good feedback on the shot itself; let me give you a little feedback on the statistics of the distribution... when we look at voting distributions, they are most often "Gaussian" (normally distributed) which means they are random processes.
The good news here is, you're right, this distribution has signs that it is non-Gaussian.
The bad news is, it's because you got too many tens.
The bottom end of it (from one through nine) fits a normal distribution almost perfectly, it's just that four folks gave you a ten for some undetermined reason, when we'd expect zero. The fact that the votes are "all over the map" just means that voters disagreed strongly on the merits of the shot. |
|
|
|
03/19/2009 12:33:40 PM · #40 |
| the problem with all complaints about "distribution" of votes is that the notion of a Gaussian or any sort of balanced distribution does not apply to voting. Why should it? Read Taleb's The Black Swan to learn more about this. |
|
|
|
03/19/2009 12:45:16 PM · #41 |
If you think your scores are warped towards the lows.. look at the scores on my submission.. though I said that in another thread. I got 15 1's and 15 10's
Some great advice here.
I agree with the colour balance being too orange.
When photographing indoors under tungsteon light, either set your cameras white balance to this, or as mentioned, use the custom white balance setting. Check your handbook as to how this works as you obviously did it wrong.
Otherwise photograph in RAW and adjust the white balance in post processing as mentioned above.
|
|
|
|
03/19/2009 12:51:07 PM · #42 |
If you plot this as a 2-dimensional function, I can already see a damn close approximation to a Gaussian curve: one of the most popular probability distribution functions. To me that looks normal - the median lobe around 5/6. If it were a ribbon winner's image, the lobe would be slightly shifted towards the right side. But I dont think you can get a partial lobe which starts abruptly at 5/6 and ends at 10.
Thats what happens when we randomly sample people's votes:-)
ETA:
For a sample winner of one of the recent challenges, you can still see a lobe, but loomed towards right:
1 - 1
2 - 1
3 - 2
4 - 11
5 - 24
6 - 74
7 - 134
8 - 126
9 - 79
10 - 79
Originally posted by jhomrighaus:
1, 1
2, 11
3, 29
4, 82
5, 174
6, 127
7, 57
8, 14
9, 4
10, 8
I mean seriously, how can you be getting 7, 8 , 9 , 10s and be getting 3s, 2s, and 1s, I have never seen such a heavy set of low votes with the curve skewed to the higher side like this. Sure a token few trolls that have some kind of issue but 10% voting below 3, this implys to me at least that the photo was horrible, but the 5 6 7 8 scores say its an average to above average photo.
Arrrggg this place can be soo annoying. |
Message edited by author 2009-03-19 12:54:41. |
|
|
|
03/19/2009 12:57:49 PM · #43 |
Originally posted by posthumous:
the problem with all complaints about "distribution" of votes is that the notion of a Gaussian or any sort of balanced distribution does not apply to voting. Why should it? Read Taleb's The Black Swan to learn more about this. |
More appropriately, "why shouldn't it?"
Normality is something that is relatively easy to test for, even if the distribution is "truncated" (with some votes "squashed" up against an arbitrary limit). I have a spreadsheet I use to do this, and Ive analyzed the voting pattern on literally hundreds of DPC images (mine and others). In nearly all cases, voting closely approximates a Gaussian distribution, usually with a very few extreme outliers.
In this specific case, there are four outliers (all of the votes of "10") and if those are removed, the distribution matches a Gaussian curve with a r-squared value of 0.998.
In God we trust, all others bring data... O wait, I'm an atheist... just bring data ;-) |
|
|
|
03/19/2009 01:02:09 PM · #44 |
Originally posted by posthumous:
the problem with all complaints about "distribution" of votes is that the notion of a Gaussian or any sort of balanced distribution does not apply to voting. Why should it? Read Taleb's The Black Swan to learn more about this. |
Exactly. However fancy the mathematics, statistics are always only approximate. It is strange that people with digital cameras fail to understand that. If all were predictable there would be no consciousness/need for consciousness. |
|
|
|
03/19/2009 01:04:27 PM · #45 |
Originally posted by posthumous:
the problem with all complaints about "distribution" of votes is that the notion of a Gaussian or any sort of balanced distribution does not apply to voting. Why should it? Read Taleb's The Black Swan to learn more about this. |
I'm gonna have to echo Fritz on this one. We SHOULD expect a gaussian distribution in a 500 vote sample. At the least we should expect an approximation. |
|
|
|
03/19/2009 01:06:41 PM · #46 |
| Maybe the scoring scale should be from 1 to 100 for added accuracy. |
|
|
|
03/19/2009 01:07:32 PM · #47 |
|
|
|
03/19/2009 01:08:23 PM · #48 |
Originally posted by K10DGuy:
my head hurts. |
Take two aspirin and call Sneezy in the morning.
R.
|
|
|
|
03/19/2009 01:14:32 PM · #49 |
Originally posted by DrAchoo:
Originally posted by posthumous:
the problem with all complaints about "distribution" of votes is that the notion of a Gaussian or any sort of balanced distribution does not apply to voting. Why should it? Read Taleb's The Black Swan to learn more about this. |
I'm gonna have to echo Fritz on this one. We SHOULD expect a gaussian distribution in a 500 vote sample. At the least we should expect an approximation. |
Ditto. And read my previous post.
There is always some order in randomness if we look at the bigger scale.. i.e. have enough samples. Pretty much all of the drugs that you take, on/off prescription, are tested on sparsely and randomly sampled population, and the statistics are considered reliable enough to trust them with human lives.
Message edited by author 2009-03-19 13:15:04. |
|
|
|
03/19/2009 01:15:20 PM · #50 |
Originally posted by kirbic:
In nearly all cases, voting closely approximates a Gaussian distribution, usually with a very few extreme outliers.
In this specific case, there are four outliers (all of the votes of "10") and if those are removed, the distribution matches a Gaussian curve with a r-squared value of 0.998.
|
So we could have 2 results, the current one and the one which excludes outliers individually until it achieves a target r-squared value, such as 0.995 or 0.997? |
|
