Exploring Noise Filtering in Image Processing: A Deep Dive into Four Methods
Have you ever snapped what you thought was the perfect photo, only to discover some unwanted guest dots, patterns, and unruly pixels spoiling the scene? Welcome to the world of image noise, a sneaky intruder that can sneak into your pictures without an invitation.
What is Image Noise? Image noise is like unexpected noise in a quiet room. It’s the stuff that creeps into your photos when you’re trying to capture a moment. Think of it as uninvited guests that crash your visual party.
How Does Noise Show Up? Imagine looking at a beautiful landscape photo, but wait! Hidden among the pixels are unwanted artifacts messing up the view. These unexpected changes in pixel values create disturbances, making your photo less awesome than it could be.
Meet the Main Culprits: Exploring Different Types of Noise
Gaussian Noise — The Trouble with Dots
Picture your image with scattered dots, each with its own intensity. This is Gaussian noise the troublemaker that crashes your visual party. These dots are like unwanted guests, and getting rid of them is a bit tricky.
Structural Noise — Patterns That Won’t Leave
Structural noise brings patterns that stick around. It’s like having an unwanted design in your photo. Imagine a visual hum that turns a nice scene into something weirdly abstract.
Speckle Noise — The Salt and Pepper Show
Meet speckle noise, also known as the salt and pepper disruptor. It adds extremely bright or dark pixels to your photo, making a mess of your visual story. These unruly pixels grab attention and mess up what you wanted to say.
Identifying Image Noise: Unmasking the Unseen Culprit
When we look at an image, we don’t actually “see” noise itself; rather, we observe its impact as corrupted pixels. To tackle noise effectively, we rely on understanding its properties, and there are two main ones that guide us in this quest.
1. Locality: The Neighbor Influence
Noise often behaves like that mischievous neighbor affecting only one pixel or a small region. In an image, pixels are buddies with their nearby pixels, creating a sense of harmony. So, if one pixel goes rogue, it’s highly likely that noise is behind the mischief. Imagine it as a ripple effect where one disturbance in the water reveals the presence of an unseen force.
2. Randomness: The Unpredictable Intruder
Noise, at its core, is a bit like unpredictability personified. It follows the rules of probability distributions, based on its source properties. This randomness is what makes noise challenging to pin down. It could be a bright speck here or a dark speck there, throwing off the balance of your image. Understanding this randomness is the key to identifying and combating the elusive intruder.
Noise Distribution Models:
Different types of noise follow distinct patterns, and three prominent models shed light on this diversity.
1. Gaussian Noise: The Common Presence
Gaussian noise is the VIP (Very Important Player) in the noise world. It’s the most prevalent type, affecting pixels with either zero noise or smaller variations. This often stems from source damages, like sensor or circuit malfunctions. Picture it like a gentle breeze, creating subtle variations. The magic lies in its normal distribution, as shown in Figure 1. This distribution makes Gaussian noise a familiar face in the world of image processing.
2. Uniform Noise Distribution: Equal Chaos Everywhere
Uniform noise distribution is a bit of an oddball. All noise levels are equal, making it challenging to differentiate between what is noise and what isn’t. It’s not very common, but when it appears, it throws a uniform chaos into the mix. Imagine it as a steady rainfall where every drop is as likely as the next. Although less frequent, uniform noise can create a perplexing visual puzzle.
3. Salt and Pepper Distribution: The Outlier Extravaganza
Salt and pepper distribution is like the rockstar of noise extreme, noticeable, and easy to identify. Imagine random extreme outliers, like grains of salt and pepper, disrupting the visual harmony. This distribution is often observed after image processing techniques or due to sensor artifacts. While not as common as Gaussian noise, its impact is strikingly apparent. (Sudden white pixels — Salt, sudden black pixels — Pepper, that’s how the name came up)
I’ll talk about how to get rid of these noise types in another blog post. Mainly focusing on the Noise Filters including Mean Filter, Median Filter and such that. Link to that article here. https://medium.com/@epcm18/exploring-noise-filtering-in-image-processing-part-2-noise-filters-53ae69dc0d68
Thank you for joining us on this exploration. Until our pixels meet again, happy capturing and processing!
