 Posted 4/12/2007 7:47:52 AM
|
|
|
|
| Okay here is another question...... Will the histogram tell the "truth" so to speak...no matter what settings you have?.. okay I will rephrase that.... If I am experimenting and I want certain effects......like let's just say light or dark...and the histogram has values between the extremes does this mean I should be okay when I upload and view ? as long as those values are between the extremes... Okay I am confusing myself..I know what I want to ask but not how to ask it... ....can anyone decode what I am trying to say?.. Okay I will try one more time...If I choose some "odd" or "crazy" settings cause I found I liked what the image looked like on the LCD....and I look at my histogram and it shows to be between the extremes..should the lighting be okay when I upload and view?...or is this just something I have to see upon upload.. lol.......I shot some and I don't want to be dissappointed with the results so I am procrastinating at this moment.
I may not always be wRIGHTbut I am never wrong Profile -  Top 24 - 
|
|
|
Posted 4/13/2007 8:58:18 AM
|
|
|
|
Okay forget that last post.. ...after the last couple of days going over the histogram and the photos I uploaded after studying the grams on the lcd....I can see the effects and understand them... that is for the bright/dark factor...Examples were given in that article and one situation in particular was given for what you can do when you are not in optimum light conditions....one tool was taking to pics and converging them so to speak... or have filters... I suppose if one does not know how to do this in editing then one must just try and get as good as they can in poor light situations...hope I have that right.. Now on the flip side....I was wondering about the rgb values...it is my understanding that an optimum image must fall between the extremes in both color and lighting....but that for the best to be gotten both lighting and color values must be in sync not mattering so much where exactly they are on this scale but that they just more so "mirror" each other.. So my next question is ..do I have that right?...also, as far as colors are concerned, will settings have just as much the same effect on these as on the brightness ?....I imagine that no one keeps toggling the menu to keep switching back and forth looking at brightness grams versus RGB grams... Hope you understood that ...and Hey! Thanks!
I may not always be wRIGHTbut I am never wrong Profile - Top 24 -
|
|
|
Posted 5/26/2007 8:17:51 AM
|
|
|
|
| Alrighty then,,,histograms again Now this morning I was energized enough after being woke up by the pain in butt beast to try and get some nice shots of some stuff as the sun was coming up just after 6:00 this morning. I don't have time to upload what I shot this morning but I am curious and almost afraid to see what I got.. I found it very odd, that while I was relying heavily on my meter...I noticed something odd.. I am not still too sure on what makes a good histogram but my theory from all I have read is that if the pixels are evenly distributed between the extremes,,the colors should be very eye pleasing.. I do not know if this should include any "waves"between these extremes or if it is even better to have totally even levels across the graph excluding the extremes.. perhaps I am dwelling too much on this and anything between the extremes and non spiking is good to go but,, If I relied on my meter, I found the histogram to be less desirable with more spikes than if I chose a speed myself according to the histograms I was getting..So I am just wondering if the meter is more important than the histogram.. any thoughts are welcome. Beautiful light for the peonies HOWEVER, they had not yet woke up from their sleep,,at someone is getting some sleep..  time to get my tomato plants planted...will be wonderful if the little ones dont pluck all the baby green tomatoes again this year...ugh.
I may not always be wRIGHTbut I am never wrong Profile - Top 24 -
|
|
|
Posted 5/26/2007 8:43:03 AM
|
|
|
|
While there is no such thing as a "perfect" histogram, some are better than others. And, yes, I do believe a good histogram distribution correlates with eye-pleasing images.
Below are the two most recent histograms I worked on:
Landscape scene at sunrise
Snowshoe Hare against white snow background
Eye-pleasing images because:
1) Good tonal range, as indicated by distribution of pixels.
2) Not too many pixels at either end of the distribution. 0 is all black and 255 is all white. It's okay to have a few pixels (low percentage of total) in the extreme area, but too many indicates loss of image detail in the highlights and/or shadows.
|
|
|
Posted 5/26/2007 9:21:15 AM
|
|
|
|
| What if they are all on the lower side ,,like your second one without the peak on the right and pretty even.. ?.. this would be good because they are even right? and your first one shown, this is a good one because there is a nice smooth curve?... or does this not really matter.. I get the basic concept but now that I have them pretty much in between/not all the time but alot whereas I never did before, the extremes,,,I am just trying to discern exactly which kinds of dispersion are better than others if there is such a thing....geesh I wish I knew how to put a pic of one of mine lol.. The one article I read , this was said in a few of them, I specifically remember it saying that there are no such things as a good or bad histogram they just are...but many examples of good ones have been like your first one you just posted Richard..so it seems that really there are better ones to shoot for. So in order for me to get this to click in my head,,tell me if this is correct...I am thinking a bad histogram or not eye pleasing might be one that has medium levels towards the outsides but not touching the ends ,,but few pixels right in the middle. Or has pixels over in the middle , again not touching the ends but very high peaks scattered throughout. is this correct that these would not be optimum histos?
I may not always be wRIGHTbut I am never wrong Profile - Top 24 -
|
|
|
Posted 5/26/2007 9:45:41 AM
|
|
|
|
I am going to see if I can shed some more light (no pun intended) on how the histogram works.
I have several screenshots below. The first, contains three squares. Pure black, pure white and in the middle 50% gray.
Notice the three spikes in the histogram. One far left, far right and middle. Each represents the "colors" or tonal values presented in the image.
Here is the second. I selected an area in the white.
The histogram at this point is only showing me tonal ranges selected. Since only white is selected, you get a spike only in the right side of the histogram.
Here is the next image with a selection on gray only.
See the spike in the middle (50% of the way across the image or "middle gray")
The black selected next. Left side spike.
In the next image, I added a small square of 60% black. Middle gray plus a small amount of white.
Notice in the histogram, the spike about 60% of the way across now. 10% closer to white.
Also notice the middle spike is shorter than it was and the spike of 60% gray is even smaller. This is due to the averaging of the number of those "colored pixels" across the entire image. There are less 50% gray pixels, thus a smaller spike. There are less 60% white (or 40% black) pixels so that spike is smaller, indicating less total average pixels of that color.
In the next image, I added 50% green. Since a histogram basically sees everything is ranges of tone, the color is unimportant. But since it is 50% green, it added a tiny bit of height to the middle spike. You probably can't see it as it is a very small difference. The spike grew because the number of 50% tone pixels grew.
In the next image, I added a larger square and in the RGB color pallet, I brought green up to 100% a value of 255. In doing that the color had to balance out and the red and blue both dropped to a value of 26. If they hadn't dropped, the color would be white again.
So although green is at 100% the new spike should be all the way to the left, correct? No, because tonally it's a little closer to the right side (towards black) because the 26 red and 26 blue make it a bit darker tonally. What's interesting is the new square is larger than the first green square yet it's spike is smaller. Actually, that's not true. The 50% green square just got stacked on top of the 50% tonal range, remember? So, because this isn't a 50% tone, you see it's spike elsewhere on the graph.
Now, in the next image, I did not alter the color, all I did was make the square bigger. So, there is no new spike the old spike for that tone just got taller. That's because the number of total pixels of that tone increase on average. Notice the black pixel bar shrunk as well. This doesn't seem to make any sense because there are still the same number of black pixels, right? Well yes and no. Since the new big green box was added, that tone consists of more lighter colors (close to the white side), thus taking away darkness from the average pixel color of the image.
So as Hahn suggested, there is no such thing as a perfect histogram. A histogram is more or less dumb. All it does is report to you the number of pixels per tone in your image.
What happens in the instance of "blown highlights" is kind of like this. Imagine you have a slider at the bottom the the histogram that is 256 units wide. The length of the slider represent the range of tones that your computer will recognize and something in between pure white and pure black. If you move that slider in one direction or the other, something on one end will be "cut off". In other words, since you computer only recognizes 256 tones anything above that will JUST STAY pure white. Anything below 0 will JUST STAY pure black.
If you slide one of the "levels" sliders (let's say the one on the right towards the left. Your computer will not recognize anything on the right of that slider any longer as an acceptable tone and will default those pixels to the NEAREST acceptable tone which is Pure Black, a tone value of 0. There is no color below 0.
Hopefully this was able to help you understand the histogram a little better. In terms of "getting a good one" on your images, that is something you are responsible for. This is controlled by exposure for the most part. Sometimes it is acceptable to have tones outside the range of the acceptable 0 - 255 values. That's an artistic choice and can be pleasing but I think it has to be done right.
|
|
|
Posted 5/26/2007 9:46:02 AM
|
|
|
|
It depends on the subject and background, of course. My goal is to set my camera to capture an image with a histogram like this:
In other words, I adjust the exposure settings on the camera, after looking at the histogram of an initial exposure, to move the distribution of pixels to something that looks like the histogram above. I tend to look at the histogram first, before looking at the image itself. You know you can put up the combined view of shot data, image and histogram all on one LCD display, right?
I do want to address some potential problems when a histogram shows a lot of peaks, rather than a smooth distribution. First of all, a smooth distribution means there will be nice gradients and smooth solid areas with detail depth and color depth. The peaked histogram could be a sign of posterization, which usually comes from jpg compression. I see posterization quite a lot in images less than 50 kb in file size. Like the one below:
More info on Posterization
|
|
|
Posted 5/26/2007 10:11:56 AM
|
|
|
| | |
