Noise
[Abstract]:
A thorough investigation and presentation of what noise has to offer visually upon digital imagery; also introducing the different and features of noise. Noise is an interesting and important element for digital artists and photo enhancers. This research will demonstate how noise can be used, what the noise controls are, then how to apply, adjust, and simulate noise.
[Definition]:
Noise is unwanted electrical or electromagnetic energy that degrades the quality of signals and data. Noise occurs in digital and analog systems, and can affect files and communications of all types, including text, programs, images, audio, and telemetry. Noise is quite simple, an effect where every pixel gets a random value added to it. This may sound simple, but it can be the first step towards just about any surface that has that "natural chaos" look to it. There is complicated math behind noise, but you don't really need to understand the math to use the techniques.
[Features]:
applying and adjusting noise to images
to simulate and replace film grain
to add a natural sparkle to skin
to prevent banding in gradients
to eliminate blocking in highlights
to eliminate blocking in shadows
to prepare for texturizing
[Controls]:
Various noise filters come with one or more of these controls. Their effects can be seen in the filter preview window.
Grain size: which should be adjusted to match that of the film or be appropriate to the scale of artwork.
Grain density: which should be adjusted match that of the film or be appropriate to the artwork.
Grain color: Multi-color is used for most full color images and adding sparkle to skin. The grain of color film is multi-color.
Monochromatic is used for grayscale and toned images to avoid adding unwanted speckles of color.
Selective hue is available in some third party filters and can be used for any of the above.
Grain distribution:
* Normal or Even distribution gives a mechanical appearance and is less visible than Gaussian.
* Gaussian gives a more clumped appearance that is more visible and more closely resembles the clumping found in film grain.
Grain blur
* Grain may need to be blurred in another filter to match that of the film, or to otherwise make it more suitable to the image, such as skin tones.
* Try various blur filters, including Gaussian, to soften noise.
[Examples | Illustrations]:
Above, we see Uniform noise. Values here are purely random and unbiased, so that any value is (in theory) just as likely as any other value. This was done with a value of 56 in the Amount slider, so all values fall randomly between (128-56)/256 and (128+56)/256.
This is Gaussian noise. Here, the randomness falls into more of a curve, where midtones are most likely, and darks and lights are rare. You'll notice that values here are found beyond the +/-56 border that the Uniform noise fell into, even though this was also made with an Amount of 56. This value, in this case, is acting more as a midpoint, and the curve continues to approach 0% probability as it reaches positive and negative infinity, in theory. In reality, it bumps into solid white and solid black and bunches up there. You can see the spikes at the very edges of the graph.
[Links]: