Photoshop Levels causes spikes

Why after applying levels and then going back to check level's

histograms,I notice vertical lines,does that mean some pixels lost

somewhere in process?Or is there a way to avoid this "loss"?Please

contribute.Many thanks.

I don't know why this happens but it means you lost data. Your image won't look different.

 

Think about it this way: Every color chanel has 256 (when 8bit i think) differnt shades of gray. If you have shades which fill up 75% of the range and you pull it to 100%, 25% of the image has to be created, but since thee are only 256, they can't be created so the software doens't have dta for that one part of the range.

 

I think i am right but if this is wrong, will the moderator please delete this post.

A gap line indicates a sudden drop in pixels of a certain tone, a spike line indicates a sudden increase. In both cases, pixels have shifted to adjacent tones, producing posterization, or steps, in the tone continuum. If the clipping you're applying in levels is minor, the spiking won't be too pronounced, and you've improved contrast. If you're clipping heavily, the degradation will be more obvious.

 

Things that should help:

 

1. Only clip once, be conservative, and do it in the scanner software.

 

2. Output clipped 16 bit files, then convert to 8 bit.

I don't think it's necessarily lost pixels. Consider greyscale, each pixel can be 0-255. If you picture happens to only have pixels from 0 to 127, and you use levels to redistribute from 0-255, you wind up doubling the brightness of each pixel. So 0 becomes 0, 1 becomes 2, 2 becomes 4, etc.

 

Nothing is "lost," but there are gaps in the histograms, there'd be no odd valued pixels. Color is the same, just more dimensions.

If you can't see a degradation, don't worry about it. If you can, do your editing in 16 bit mode. This applies if you are talking about spikes not at the extremes of the tonal scale. Otherwise, set the black/white points more conservatively; i.e., move them away from the midtones.

Try using Curves instead of levels, as you tend to do less damage to the image that

way.

 

Another way of working is to make a duplicate of your image, do a save as with copy

in the name ofthe new file, and then do your adjustments on a seperate layer

(Layers> Adjustment Layers > Levels), make your adjustments and once you are

satisfied, flatten the layers. This perserves the original file (you are not working on it)

and allows you to make changes that are more easily reversable.

You can close the gaps and make the histogram smooth again by

slightly increasing the overall saturation of the image (say, +10%)

after applying Levels.

As Emre states, if you can't see it, it's not a problem. In case you're not already familiar with what 'it' looks like, the gaps in the histogram result in what would be smooth gradations appearing both on screen and what's worse, in print, as staggered puddles of color. If this is a new concept to you, I recommend that you purposefully exaggerate a levels adjustment of an image with clear blue sky and view the results at 100%.

Applying unsharp mask right after levels/curves will clean up most of the holes/spikes in the histogram.

There's a good tutorial on this at luminous-landscape.com. It's "Understanding Bit Depth" in the "Understanding" section. Also check out the section on histograms.

Many thanks to all contributors,I shall try all different remedies shown above. And yes, I notice no degradation but it was an issue I could not unerstand.

Every adjustment in Photoshop results in some lost information.

 

Once the information is gone, there is nothing you can do to bring it back. If you make more than one adjustment, each incremental adjustment results in a increasing rate of lost information.

 

Information loss can result in visible noise, posterization, banding, or lost detail in highlights and/or shadows.

 

Whether or not that information loss is visible depends on many factors, including the amount of information in the original image file, the absolute differences between adjacent pixels in the file, the size of the image file, and the size of the print.

 

Although the problem is impossible to avoid, there are several things you can do to minimize the impact of lost information:

 

- Scan in the highest bit resolution supported by your scanner (or digital camera), and edit using 16-bit files in Photoshop.

 

- Minimize the number of tonal adjustments done in Photoshop.

 

- Minimize the number of times you change color space.

 

- Assign ICC profiles instead of converting to a new profile, when possible.

 

- Minimize the number of times you resample an image (size adjustments, rotation, etc.)

 

Also, let me clarify a couple of points that others have made about the difference between levels and curves.

 

A Levels adjustment is a linear adjustment. That means that every pixel is adjusted in the exactly the same way.

 

If you are working in 8-bit mode in Photoshop, the peaks and gaps that you see after making a Levels adjustment is due to rounding errors. The peaks and gaps represent loss of detail. Whether or not this loss is visible depends on the factors mentioned above.

 

You cannot recover this lost information by performing additional adjustments. You might be able to fill in the gaps and level out the peaks by using a filter (such as unsharp mask), but you are not restoring lost data. Instead, you are massaging the damaged data, making it even more damaged.

 

A Curves adjustment is non-linear. In other words, the changes made from pixel to pixel are not necessarily the same. You can use Curves to create an adjustment that is linear and identical to a Levels adjustment. The reverse is not true.

 

Using curves gives you a great deal more control than does Levels. The choise of which to use depends on what you're trying to achieve. One is not always better than the other.

 

For more information, see Real World Adobe Photoshop by David Blatner and Bruce Fraser.

Thank you Michael. Regards.

Michael is giving you lots of good advice. I´d like to comment on

the one about assigning ICC-profiles instead of converting to a

new profile. This advice should be don´t change profile unless

it´s necessary. Don´t assign new profiles to the images you

know the color space of.

 

If you want the colors in your image to be interpreted correctly in

the new color space you should *not* just assign the new profile

instead of converting. This is because the same RGB-value

represents different colors in different RGB-spaces. For example

the skin tones in an image created in the sRGB space will turn

flaming red if you just assign the Adobe RGB-profile to it instead

of converting the image to Adobe RGB.

 

When I make color corrections in an image I always use layers

and keep a copy before merging the layers. This is because

there is less "generation loss" if you use this method.

Photoshop will only calculate the new pixel values once when

the image is merged. The layer method is also useful for further

corrections of an image since you can go back to your original

adjustment layers and adjust the original pixels instead of

correcting a second generation image.

I´d like to change my previous advice about assigning new

profiles to images. My advise is only true when the original

icc-profile is unknown. (The image is for example sRGB but the

profile is not embedded into the image. If you assign the Adobe

RGB-profile to this image the colors will be wrong).

Thanks Daniel..Regards.