Overview of Noise and Artifacts¶
Noise and artifact are terms used to describe speckles, spikes, reseaus, missing data, and other marks, blemishes, defects, and abnormalities in image data created during the acquisition, transmission, and processing of image data. The line between the definitions of artifact and noise is fuzzy (and often subject to opinion), and often the terms are used interchangeably. Some noise and artifacts are expected, even purposefully added, and can be removed during the radiometric calibration process.
What is noise?¶
In image processing, noise is a type of flaw or blemish in the image caused by:
- Telemetry data dropouts or transmission errors
- Malfunctioning or dead detectors
- Read noise native to the CCD system
- Coherent noise caused by spurious electronic signals from the operation of instruments onboard the spacecraft
Noise can take the appearance of speckling, missing data, random or orderly patterns, and other variations that cause the image to have a muddled appearance, or visually distracting blemishes or patterns. There are three categories of noise:
- Fixed-location noise always exists at the same location in the image array, with predictable positions. Fixed location noise can be cosmetically corrected by replacing the bad pixels with the weighted average of the unaffected neighborhood pixels. Fixed-location noise can result from malfunctioning or dead detectors.
- Randomly occurring noise results from data transmission errors causing data bits to be altered at random intervals in the image. The random noise produces discrete, isolated pixel variations or "spikes" and gives an image a "salt-and-pepper" appearance. Additionally, telemetry drop-outs can cause portions of an image to be completely missing. This type of noise is generally corrected using filtering techniques that recognize missing or anomalous data pixels and replaces these data points with a weighted average of the unaffected neighborhood pixels.
- Coherent noise can be introduced by spurious electronic signals produced by the operation of instruments onboard the spacecraft during image observations. The spurious signals interfere with the electronics of the imaging system, causing coherent noise patterns to be electronically "added" to the images. For example, the shuttering electronics of the Viking cameras introduced a spurious "herring-bone" pattern at the top and bottom of the image. Noise-removal algorithms are designed to correct specific coherent noise problems such as this one.
What are artifacts?¶
Generally, image artifacts are a type of flaw or blemish in the image introduced during processing, intentionally introduced due to the design of system or unintentional introduction of debris or energy external to the system. Examples of artifacts include:
- reseaus etched on the camera lens
- reseaus exposed on photographic film during pre-flight preparations for a mission
- minute dust specks located in the optical path or on the focal plane array
- cosmic rays and other charged particles impacting the sensor (particularly CCDs)
- fringe, ring, or visible patterns created during filtering, ratio analysis, and other enhancement processes
- quantization, checkerboarding, and other artifacts introduced by image compression algorithms during conversion from Isis cube format to a lossy image format or bit-type reductions that reduce the tonal resolution of the data
Most artifacts fit neatly into the categories of noise listed earlier and are corrected using many of the same processes. For example, dust specks create fixed-location blemishes, and cosmic rays cause random spikes. Reseaus are useful blemishes that are removed once they are analyzed for their locations within an image and the information saved for later processing.
Reseaus, dropped data, and salt-and-pepper¶
This example shows a number of noise and artifact types. The large, regularly spaced black dots across the image are caused by the reseaus on the camera lens. The pattern of vertical black lines across the bottom of the image was caused by transmission data loss. The black and white speckles, called salt-and-pepper, are random noise.
Instrument error¶
The black streak running down the right side of this image may have been caused by an unusual error in the line-scanner camera that acquired this image.
Transmission error¶
A glitch during the transmission of this image caused the data to become garbled (upper right) and some data was completely lost (the black area across the middle).
Image compression¶
Exporting the image to a lossy, compressed image format (using very high compression), the compression algorithm introduced a tiled pattern across the image.
