Noise Reduction: Enhancing Clarity, Advanced Techniques for Noise Reduction in Computer Vision

By Fouad Sabry

What is Noise Reduction

Noise reduction is the process of removing noise from a signal. Noise reduction techniques exist for audio and images. Noise reduction algorithms may distort the signal to some degree. Noise rejection is the ability of a circuit to isolate an undesired signal component from the desired signal component, as with common-mode rejection ratio.

How you will benefit

(I) Insights, and validations about the following topics:

Chapter 1: Noise reduction

Chapter 2: Dolby noise-reduction system

Chapter 3: Dbx (noise reduction)

Chapter 4: Digital image processing

Chapter 5: Image noise

Chapter 6: Wavelet

Chapter 7: Difference of Gaussians

Chapter 8: Bilateral filter

Chapter 9: Non-local means

Chapter 10: Block-matching and 3D filtering

(II) Answering the public top questions about noise reduction.

(III) Real world examples for the usage of noise reduction in many fields.

Who this book is for

Professionals, undergraduate and graduate students, enthusiasts, hobbyists, and those who want to go beyond basic knowledge or information for any kind of Noise Reduction.

• Language: English
• Publisher: One Billion Knowledgeable
• Release date: Apr 28, 2024

Book preview

Chapter 1: Noise reduction

Eliminating unwanted background noise from a signal is known as noise reduction. Both visual and aural noise can be reduced using various methods. The signal may be distorted slightly by noise-reduction algorithms. Common-mode rejection ratio and noise rejection are two measures of a circuit's ability to filter out unwanted signals.

There are characteristics shared by both analog and digital signal processors that make them vulnerable to noise. A device's mechanism or signal-processing algorithms can introduce frequency-dependent noise, in addition to the random, uniform noise that we call white noise..

Hiss is a common form of electronic noise caused by the chaotic motion of electrons in response to fluctuations in temperature. Rapid addition and subtraction by these agitated electrons to the output signal results in audible noise.

The grain structure of photographic film and magnetic tape introduces noise (both visible and audible) into the recording. The sensitivity of a photographic film is determined by the size of its grain; more sensitive film will have larger grains. Noise in magnetic tape increases in proportion to the size of the magnetic particles (typically ferric oxide or magnetite). The noise can be reduced to a more tolerable level by using more film or magnetic tape.

In order to reduce noise, algorithms typically modify signals in some way. Avoiding modifications to the signals is possible with the help of the local signal-and-noise orthogonalization algorithm.

For seismic imaging, signal enhancement is especially important because the useful signal buried in the background noise is often overlooked, leading to potential for spurious discontinuities in seismic events and artifacts in the final migrated image. Reducing interpretation difficulties and false alarm risks in oil and gas detection can be accomplished by boosting the useful signal while attenuating random noise in seismic profiles.

Noise reduction example

Using Audacity, here is an example of noise reduction at 0 dB, 5 dB, 12 dB, and 30 dB with frequency smoothing at 150 Hz and an attack/decay time of 0.15 seconds.

---

Is there a problem with this file? For more information, check out the media.

In analog tape recording, tape hiss is a major performance bottleneck. This is associated with the relative tape velocity across the tape heads and the particle size and texture of the magnetic emulsion sprayed onto the recording media.

Single-ended pre-recording, hiss reduction, surface noise reduction, and codec or dual-ended systems are the four main methods of noise reduction. Single-ended pre-recording systems (like Dolby HX Pro) aim to modify the medium itself during the recording process. Scratches, pops, and surface non-linearities can all be mitigated during phonograph record playback with the help of single-ended hiss reduction systems like DNL. The Phase Linear Autocorrelator Noise Reduction and Dynamic Range Recovery System (Models 1000 and 4000) is a single-ended dynamic range expander that can be used to remove a wide variety of noise from vintage recordings. During the recording phase of dual-ended systems (like Dolby noise reduction system or dbx), a pre-emphasis process is applied, and during the playback phase, a de-emphasis process is applied.

During recording, dual-ended compander noise reduction systems apply pre-emphasis, and during playback, they apply de-emphasis.

Dolby A and UC, used by professionals in vinyl recordings, and Dolby FM are all examples of such systems, High Com FM and FMX used in FM radio broadcasting.

Ray Dolby created the first successful method for reducing background noise in recorded music in 1966. Dolby Type A was an encode/decode system that was designed for industrial applications. During recording (encoding), the amplitude of frequencies in four bands was increased, and during playback, the amplitude was decreased correspondingly (decoding). In particular, above 1 kHz of an audio signal would be amplified during the recording of quieter sections of the signal. This improved the signal-to-noise ratio on tape by as much as 10 dB, depending on the signal strength to begin with. The decoder worked in reverse during playback, effectively decreasing the background noise by up to 10 decibels.

Dolby B was a consumer-focused single-band system developed in collaboration with Henry Kloss. Dolby B was not as powerful as Dolby A, but it could be played back without a decoder, which was an advantage.

The Telefunken High Com integrated circuit U401BR could be utilized to work as a mostly Dolby B–compatible compander as well.

In various late-generation High Com tape decks the Dolby-B emulating D NR Expander functionality worked not only for playback, but, because it's not in the manual, additionally while recording.

Dbx, created by David E. Blackmer of Dbx, Inc., was an alternative analog noise reduction system. The high-noise frequencies were amplified and a 2:1 compander was applied to the entire signal before it was encoded and decoded using a root-mean-squared (RMS) algorithm. Unlike Dolby B, which could be played without a decoder, dbx used the entire audible spectrum. However, up to 30 dB of noise reduction is possible with this method.

There's no need for audio-style noise reduction in analog video recordings because the luminance part (composite video signal in direct color systems) is modulated with frequency to maintain saturation.

Philips introduced the dynamic noise limiter (DNL) in 1971 for use on cassette decks as a method of reducing background noise.

A second group of algorithms, time-frequency filters, operate in the time-frequency domain using linear or non-linear filters