erode

Apply morphological erosion to expand dark regions and shrink bright areas.

Usage

Terminal

$ mediaproc image erode <input> [options]

The erode command applies morphological erosion, a fundamental operation in image processing that expands dark (low-value) regions while shrinking bright (high-value) regions. This is the opposite of dilation. Erosion is useful for:

Noise removal: Remove small bright spots and noise
Object separation: Separate touching or connected objects
Thinning: Thin bright features like text or lines
Border creation: Create borders around objects
Mask refinement: Clean up binary masks for segmentation

Uses a 3×3 kernel that examines each pixel and its 8 neighbors, replacing the center pixel with the minimum value in the neighborhood.

Options

Option	Type	Default	Description
`<input>`	string	required	Input image file path
`-o, --output`	string	`<input>_eroded.<ext>`	Output file path
`-q, --quality`	number	`90`	Output quality 1-100
`--dry-run`	boolean	`false`	Show what would be done without processing
`-v, --verbose`	boolean	`false`	Enable verbose logging

Examples

Basic Erosion

Apply erosion to shrink bright regions:

# Erode binary mask
mediaproc image erode mask.png -o eroded_mask.png

# Remove bright noise
mediaproc image erode noisy.jpg -o clean.jpg

# Thin text or lines
mediaproc image erode text.png -o thinned.png

# Shrink white regions
mediaproc image erode binary.jpg -o shrunk.jpg

Quality Control

# High quality output
mediaproc image erode mask.png -q 95 -o high_quality.png

# Standard quality
mediaproc image erode image.jpg -q 90 -o eroded.jpg

# Lower quality for previews
mediaproc image erode binary.png -q 80 -o preview.png

Multiple Erosions

Apply erosion multiple times for stronger effect:

# Single erosion
mediaproc image erode mask.png -o eroded1.png

# Double erosion (stronger effect)
mediaproc image erode mask.png -o temp.png
mediaproc image erode temp.png -o eroded2.png
rm temp.png

# Triple erosion
mediaproc image erode mask.png -o temp1.png
mediaproc image erode temp1.png -o temp2.png
mediaproc image erode temp2.png -o eroded3.png
rm temp1.png temp2.png

Morphological Opening

Combine erosion then dilation to remove noise:

# Opening operation: erode → dilate
# Removes small bright spots while preserving size
mediaproc image erode noisy.png -o temp_eroded.png
mediaproc image dilate temp_eroded.png -o opened.png
rm temp_eroded.png

# Stronger opening (multiple iterations)
mediaproc image erode noisy.png -o temp1.png
mediaproc image erode temp1.png -o temp2.png
mediaproc image dilate temp2.png -o temp3.png
mediaproc image dilate temp3.png -o opened_strong.png
rm temp1.png temp2.png temp3.png

Separating Connected Objects

Break apart touching objects:

# Separate touching characters
mediaproc image erode text_blob.png -o separated_text.png

# Split merged objects
mediaproc image erode connected.jpg -o split.jpg

# Break connected components
mediaproc image erode merged.png -o individual.png

Batch Processing

Apply erosion to multiple images:

# Erode all masks
for file in masks/*.png; do
  mediaproc image erode "$file" -o "eroded/${file##*/}"
done

# Process binary images
for img in binary/*.jpg; do
  mediaproc image erode "$img" -o "processed/${img##*/}"
done

# Erode with quality setting
for file in *.png; do
  mediaproc image erode "$file" -q 95 -o "output/${file}"
done

Edge Detection

Detect edges by subtracting eroded from original:

# Edge detection workflow
mediaproc image erode image.jpg -o eroded.jpg
# Then use external tool to subtract eroded from original
# Result shows object boundaries

Text Processing

Process text in images:

# Thin bold text
mediaproc image erode bold_text.png -o thin_text.png

# Separate touching characters
mediaproc image erode connected_text.jpg -o readable.jpg

# Remove text artifacts
mediaproc image erode noisy_text.png -o clean_text.png

Document Cleanup

Clean scanned documents:

# Remove scanning noise
mediaproc image erode scan.jpg -o clean_scan.jpg

# Thin thick text
mediaproc image erode thick_document.png -o readable.png

# Remove dust and spots
mediaproc image erode dusty.jpg -o cleaned.jpg

Dry Run Testing

Test erosion before processing:

# Preview erosion
mediaproc image erode mask.png --dry-run

# Test batch operation
for file in *.png; do
  mediaproc image erode "$file" --dry-run
done

Verbose Processing

Monitor erosion operations:

# See processing details
mediaproc image erode mask.png -v -o eroded.png

# Debug batch processing
for file in *.jpg; do
  mediaproc image erode "$file" -v -o "output/${file}"
done

How Erosion Works

Morphological Operation

Erosion examines each pixel and its 8 neighbors in a 3×3 grid:

Kernel:
-1 -1 -1
-1  8 -1
-1 -1 -1

Operation:
For each pixel, replace with MINIMUM value in 3×3 neighborhood

Effect:

Dark pixels expand into neighboring bright pixels
Bright regions shrink
Small bright spots disappear
Objects decrease in size

Visual Example

Before (binary):           After erosion:
0 0 0 0 0                 0 0 0 0 0
0 1 1 0 0                 0 0 0 0 0
0 1 1 1 0    →           0 0 1 0 0
0 0 1 1 0                 0 0 0 0 0
0 0 0 0 0                 0 0 0 0 0

Notice: Size decreased, small features removed

Common Use Cases

Noise Removal

# Remove small bright spots (salt noise)
mediaproc image erode noisy.png -o denoised.png

# Clean up scanning artifacts
mediaproc image erode scan.jpg -o clean.jpg

Object Separation

# Separate touching components
mediaproc image erode connected_objects.png -o separated.png

# Break apart merged regions
mediaproc image erode merged.jpg -o individual_objects.jpg

Text Processing

# Thin thick text lines
mediaproc image erode bold_text.png -o thin.png

# Separate connected characters
mediaproc image erode touching_chars.jpg -o readable.jpg

Medical Imaging

# Remove noise from medical scan
mediaproc image erode xray.png -o clean_xray.png

# Separate connected tissue regions
mediaproc image erode tissue.jpg -o separated_tissue.jpg

Best Practices

Work with Binary/Grayscale: Erosion works best on high-contrast or binary images
Multiple Iterations: Apply multiple times for stronger effects
Combine with Dilation: Use opening (erode→dilate) to remove noise without size change
Quality for Masks: Use PNG format for binary masks to preserve sharp edges
Test First: Use dry-run to verify operation before batch processing
Preserve Originals: Always keep original files for comparison
Monitor Results: Use verbose mode for troubleshooting

Morphological Workflows

Opening (Remove Noise)

# Remove small bright noise while preserving size
mediaproc image erode input.png -o temp_eroded.png
mediaproc image dilate temp_eroded.png -o opened.png
rm temp_eroded.png

Morphological Gradient

# Gradient = dilated - eroded
# Highlights object boundaries
mediaproc image dilate input.png -o dilated.png
mediaproc image erode input.png -o eroded.png
# Subtract eroded from dilated using external tool

Black Top-Hat

# Black top-hat = closed - original
# Extracts small dark features
# (closing = dilate then erode)
mediaproc image dilate input.png -o temp1.png
mediaproc image erode temp1.png -o closed.png
# Subtract original from closed

Performance Tips

Format Selection: PNG for binary masks, JPEG for grayscale
Quality Settings: Use lower quality for intermediate results
Batch Efficiency: Process similar images together
Iteration Control: More iterations = more processing time
File Size: Erosion doesn't significantly increase file size

Troubleshooting

Objects Too Small After Erosion

Problem: Eroded objects are too small or disappeared.

Solutions:

# Use single erosion instead of multiple
mediaproc image erode mask.png -o eroded.png

# Apply dilation after to grow back
mediaproc image erode mask.png -o temp.png
mediaproc image dilate temp.png -o balanced.png
rm temp.png

Objects Still Connected

Problem: Touching objects still not separated.

Solutions:

# Apply multiple erosions
mediaproc image erode mask.png -o temp1.png
mediaproc image erode temp1.png -o temp2.png
mediaproc image erode temp2.png -o separated.png
rm temp1.png temp2.png

# Then dilate back to restore size
mediaproc image dilate separated.png -o restored.png

Not Enough Effect

Problem: Noise still visible after erosion.

Solutions:

# Apply multiple erosions
mediaproc image erode noisy.png -o temp1.png
mediaproc image erode temp1.png -o temp2.png
mediaproc image erode temp2.png -o clean.png
rm temp1.png temp2.png

# Use morphological opening
mediaproc image erode noisy.png -o temp_e.png
mediaproc image dilate temp_e.png -o opened.png
rm temp_e.png

Quality Loss

Problem: Output looks pixelated.

Solutions:

# Increase quality
mediaproc image erode mask.png -q 95 -o high_quality.png

# Use PNG for binary images
mediaproc image erode binary.jpg -o eroded.png

# Ensure source is high resolution

Too Much Erosion

Problem: Important features removed.

Solutions:

# Use single pass only
mediaproc image erode mask.png -o eroded.png

# Apply gentler erosion then dilate
mediaproc image erode mask.png -o temp.png
mediaproc image dilate temp.png -o balanced.png
rm temp.png