tutorials

Photo Denoise: Remove Grain and Noise from Your Images

Q: is Image Noise?

Image noise refers to random variations in brightness or color that don't represent the actual scene. It appears as grain, speckles, or a "sandy" texture that degrades image quality.

Q: Causes Photo Noise?

Understanding noise causes helps you prevent it in the first place and choose the right reduction method.

January 16, 2025

8 min read

intermediate

Before and after comparison showing noisy photo transformed into clean, noise-free image

What is Image Noise?
What Causes Photo Noise?
Types of Noise: Luminance vs Chroma
Gaussian Denoise: The Universal Solution
Median Filter: Edge-Preserving Denoise
NLM: Highest Quality Noise Reduction
Choosing the Right Denoise Strength
Step-by-Step Denoise Guide
Common Denoise Mistakes
FAQ

Close-up comparison showing noise reduction in low-light photography

What is Image Noise?

Image noise refers to random variations in brightness or color that don't represent the actual scene. It appears as grain, speckles, or a "sandy" texture that degrades image quality.

Visual Characteristics of Noise

What noise looks like:

🔍 Grain texture: Random dots throughout the image
🎨 Color speckles: Random colored pixels (especially in shadows)
📉 Loss of detail: Fine textures become muddy
💫 Reduced sharpness: Overall "soft" appearance

Professional photo editing example
Left: Clean, noise-free image | Right: Noisy image showing grain and color speckles

Why Noise is a Problem

Impact on your photos:

❌ Unprofessional appearance: Looks amateur or low-quality
❌ Print issues: Noise becomes very visible in large prints
❌ Detail loss: Fine textures and details get obscured
❌ Color degradation: Colors appear muddy or incorrect
❌ Social media: Compression makes noise even worse

Pro Tip: A small amount of noise is often acceptable (even desirable for film-like aesthetic), but excessive noise always degrades quality.

What Causes Photo Noise?

Understanding noise causes helps you prevent it in the first place and choose the right reduction method.

Night photography demonstrating effective grain removal and detail preservation

Primary Causes of Image Noise

1. 🌙 High ISO Settings

The main culprit:

Higher ISO = more amplification of sensor signal
More amplification = more amplified electronic noise
ISO 100-400: Minimal noise
ISO 800-1600: Noticeable noise
ISO 3200+: Significant noise

Why it happens: Your camera sensor is boosting a weak signal, which amplifies both light information and random electronic interference.

2. 🕶️ Underexposure

Dark photos = more noise:

Insufficient light reaching sensor
Brightening in post reveals hidden noise
Shadow areas particularly affected

Solution: Expose as bright as possible without blowing highlights (ETTR - Expose To The Right).

3. 🌡️ Long Exposures

Heat-generated noise:

Sensor heats up during long exposures
Heat causes "hot pixels" and general noise
Especially problematic above 30 seconds

Examples: Night photography, astrophotography, light trails

4. 📱 Small Sensors

Physics problem:

Smaller sensors = smaller pixels
Smaller pixels = less light gathering ability
Less light = more noise at same ISO
Smartphone cameras particularly affected

5. 🔥 High Temperatures

Environmental factor:

Hot weather increases sensor temperature
Thermal noise increases
Compounded with other factors

Noise by Camera Type

Camera Type	ISO 100-400	ISO 800-1600	ISO 3200+
Full-Frame DSLR/Mirrorless	Excellent	Very Good	Good
APS-C DSLR/Mirrorless	Very Good	Good	Noticeable
Micro Four Thirds	Good	Noticeable	Significant
High-End Smartphone	Good	Noticeable	Heavy
Budget Smartphone	Noticeable	Significant	Severe

Types of Noise: Luminance vs Chroma

Not all noise is the same. Understanding the two main types helps you choose the right reduction approach.

Luminance Noise (Brightness)

What it is: Random variations in brightness/lightness

Appearance:

Looks like film grain
Black and white speckles
More noticeable in shadows
Actually somewhat acceptable

Visual example: Like sand or grain texture

Professional photo editing example
Luminance Noise: Grainy texture, similar to film grain

Impact:

✅ Less objectionable than chroma noise
✅ Can create "film-like" aesthetic
⚠️ Still reduces sharpness and detail

Chroma Noise (Color)

What it is: Random variations in color information

Appearance:

Red, blue, green speckles
"Confetti" or colored dots
Very noticeable and distracting
Particularly bad in shadows and blue skies

Visual example: Like random colored pixels

Professional photo editing example
Chroma Noise: Colored speckles, very distracting

Impact:

❌ Extremely objectionable
❌ Looks very unprofessional
❌ Should always be removed
❌ Cannot be stylistically acceptable

Denoise Priority

Always tackle noise in this order:

First: Remove chroma (color) noise completely
- Most visually objectionable
- No downsides to aggressive removal
Second: Reduce luminance noise carefully
- Balance between smoothness and detail
- Too much = loss of sharpness
- Too little = grainy appearance

Pro Tip: It's better to have some luminance noise than to lose detail through over-denoising!

Gaussian Denoise: The Universal Solution

Gaussian blur-based denoise is the most common and versatile noise reduction method.

How Gaussian Denoise Works

The technique:

Applies a controlled blur to the image
Smooths out random noise variations
Uses Gaussian distribution for natural look

Strengths:

✅ Fast processing
✅ Works well for general noise
✅ Predictable results
✅ Good for moderate noise levels

Weaknesses:

⚠️ Can soften edges
⚠️ May reduce sharpness
⚠️ Less sophisticated than modern methods

When to Use Gaussian Denoise

Best for:

General-purpose noise reduction
Web-sized images (less detail needed)
Photos where slight softness is acceptable
Quick processing needs
Moderate noise levels

Avoid for:

Portraits requiring sharp detail
Images with fine textures
Heavy noise requiring aggressive reduction
When you need the absolute best quality

Gaussian Denoise Settings

Strength guidelines:

Noise Level	Gaussian Strength	Result
Light	1-3	Subtle smoothing, minimal detail loss
Moderate	3-5	Noticeable reduction, slight softening
Heavy	5-8	Significant reduction, some detail loss
Severe	8-10	Maximum reduction, considerable softening

Warning: Above strength 7, you'll start losing important detail. Consider other methods for heavy noise.

Professional photo editing example
Gaussian Denoise: Strength 0 (original), 3 (light), 6 (moderate), 9 (heavy)

Median Filter: Edge-Preserving Denoise

Median filter is a specialized denoise method that excels at preserving edges while removing noise.

How Median Filter Works

The technique:

Analyzes neighborhood of pixels
Replaces each pixel with the median value
Preserves edges better than Gaussian blur
Excellent for certain noise types

The magic: Because it uses median (not average), it's less affected by outliers (noise spikes).

Strengths of Median Filter

Why use Median:

✅ Superior edge preservation: Sharp boundaries maintained
✅ Excellent for salt-and-pepper noise: Random black/white speckles
✅ No edge halos: Clean transitions
✅ Good for chroma noise: Effective at color speckles

When to Use Median Filter

Best for:

Images with important sharp edges (architecture, products)
Salt-and-pepper noise (random black/white dots)
Chroma (color) noise reduction
Photos where sharpness is critical
Graphics and illustrations with flat colors

Less ideal for:

Photos with fine textures
General luminance noise
Very heavy noise (may need multiple passes)

Median Filter Settings

Kernel size determines strength:

Kernel Size	Effect	Best For
3x3	Minimal, preserves detail	Light noise, careful reduction
5x5	Moderate reduction	General use, balanced approach
7x7	Strong reduction	Heavy noise, less concern for fine detail
9x9+	Very aggressive	Severe noise, illustrations

Pro Tip: Start with 3x3 or 5x5. Larger kernels can create "blocky" artifacts.

Photo filter examples
Edge Preservation: Gaussian (left) vs Median (right) - notice sharper edges with Median

NLM: Highest Quality Noise Reduction

Non-Local Means (NLM) is the most advanced denoise algorithm, offering superior quality at the cost of processing time.

How NLM Works

The advanced technique:

Looks at patterns across the entire image
Finds similar patches and averages them
Preserves texture while removing noise
Much more sophisticated than simple blur

Why it's special: NLM distinguishes between noise (random) and texture (pattern), preserving real detail while removing randomness.

Strengths of NLM Denoise

Why NLM is superior:

✅ Best detail preservation: Maintains fine textures
✅ Natural appearance: No artificial smoothing
✅ Excellent for heavy noise: Handles high ISO well
✅ Preserves edges and textures: Both sharp and soft details
✅ Professional quality: Best possible results

When to Use NLM

Best for:

Professional work requiring maximum quality
Portraits (preserves skin texture while reducing noise)
Landscapes with fine detail
Heavy noise from high ISO
Any image where quality is paramount
Large prints or detailed viewing

Trade-offs:

⏱️ Slower processing (worth the wait!)
🔧 More parameters to understand
💻 More computationally intensive

NLM Settings Explained

Key parameters:

Filter Strength (h):
- Controls noise reduction intensity
- Range: 3-15 (higher = more reduction)
- Start with 10 for typical noise
Template Window Size:
- Size of patch being compared
- Typical: 7x7 pixels
- Larger = smoother, slower
Search Window Size:
- How far to look for similar patches
- Typical: 21x21 pixels
- Larger = better quality, much slower

Recommended settings by noise level:

Noise Level	Filter Strength (h)	Result
Light	3-5	Subtle, natural
Moderate	6-8	Balanced
Heavy	9-12	Strong reduction
Severe	12-15	Maximum reduction

Professional photo editing example
NLM Quality: Notice how skin texture is preserved while noise is removed

Choosing the Right Denoise Strength

Balance is key: Too little leaves noise, too much loses detail. Here's how to find the sweet spot.

Professional photographer working in modern photography studio

The Detail vs Smoothness Trade-off

Every denoise method involves compromise:

More Noise Reduction ←→ Less Detail
Smoother Appearance ←→ Natural Texture

The goal: Remove noise while preserving as much real detail as possible.

Guidelines by Image Type

Portraits

Skin texture is critical:

Face close-ups: Low strength (2-4), prefer NLM
Full body: Moderate strength (4-6)
Group photos: Can handle more (5-7)

Priority: Smooth skin while keeping texture natural

Landscapes

Detail preservation crucial:

Mountain/forest details: Low-moderate (3-5), use NLM
Sky: Can handle more aggressive (6-8)
Water/clouds: Moderate (4-6)

Priority: Keep textures in foliage, rocks, etc.

Architecture & Urban

Sharp edges important:

Buildings: Moderate (4-6), use Median for edges
Interiors: Low-moderate (3-5)
Night cityscapes: Higher okay (6-8)

Priority: Preserve straight lines and edges

Product Photography

Texture and detail critical:

Smooth products: Can handle more (5-7)
Textured products: Low strength (2-4)
Jewelry/watches: Very low (1-3), use NLM

Priority: Maintain product textures exactly

The Before/After Test

How to know if you've got the right strength:

Zoom to 100% (actual pixels)
Compare before and after
Ask yourself:
- Is noise sufficiently reduced?
- Are important details preserved?
- Does it look natural or artificial?
- Can you see a texture pattern or is it smooth?

Rule of thumb: If you can't easily tell the image was denoised, you've nailed it!

Step-by-Step Denoise Guide

Follow this professional workflow for optimal results:

Step 1: Assess the Noise

Before you start:

Zoom to 100% (actual pixels)
Look at shadows, blue skies, and uniform areas
Identify noise type:
- Luminance (grain)?
- Chroma (color speckles)?
- Both?
Assess severity: light, moderate, heavy, or severe?

Step 2: Choose Your Method

Decision matrix:

Situation	Best Method	Why
Light noise, need speed	Gaussian	Fast, sufficient
Moderate noise, sharp edges needed	Median	Edge preservation
Heavy noise, quality critical	NLM	Best quality
Portrait	NLM	Texture preservation
Web image, quick job	Gaussian	Speed
Print/professional	NLM	Maximum quality

Step 3: Remove Chroma Noise First

If you have color noise:

Go to Denoise Tool
Upload your photo
Navigate to Professional Tools → Denoise
Start with chroma noise reduction
Use higher strength (you can be aggressive)
Remove color speckles completely

Noise reduction comparison showing grain removal

Step 4: Reduce Luminance Noise

After chroma is gone:

Select your denoise method (Gaussian, Median, or NLM)
Start with moderate strength (5 for Gaussian, 7-8 for NLM)
Observe real-time preview
Adjust strength up or down
Find the sweet spot

Key balance: Smooth enough to look clean, but detailed enough to look real.

Step 5: Check Critical Areas

Zoom in and inspect:

✅ Faces: Skin texture preserved?
✅ Edges: Still sharp?
✅ Fine details: Hair, foliage, fabric texture?
✅ Shadows: Smooth but not plastic?
✅ Sky: Clean but natural?

Step 6: Sharpen (Optional)

Denoise can soften images:

After denoising, consider light sharpening
Amount: 10-30% depending on softening
Radius: 0.5-1.0 pixels
Brings back edge definition

Pro Tip: Always denoise first, sharpen last!

Step 7: Final Comparison

Toggle before/after multiple times
View at different zoom levels:
- 100% (pixel-level quality)
- 50% (typical viewing)
- Fit to screen (overall impression)
Verify improvement outweighs any softening