1.
Image Degradation & Restoration Model
(Model: g(x,y) = h(x,y)*f(x,y) + n(x,y))
This model explains how an image gets corrupted and how we recover it.
Application 1: Astronomy Imaging
Historic Example: Early space telescopes like Hubble (1990) produced blurred images due to mirror defects. Scientists used
restoration algorithms to correct them.
Experimental Example:
• Take a sharp image
• Apply blur + noise artificially
• Use inverse/Wiener filtering to recover original
Application 2: Medical Imaging
Historic Example: Early CT scans had heavy noise due to low sensor quality. Restoration improved tumor detection
accuracy.
Experimental Example:
• Add Gaussian noise to an MRI image
• Apply restoration model
• Compare before/after clarity
Application 3: Surveillance Systems
Historic Example: Criminal identification from blurry CCTV footage (e.g., early 2000s investigations).
Experimental Example:
• Simulate motion blur using convolution
• Apply deblurring filter
• Measure improvement in edge sharpness
Application 4: Satellite Imaging
Historic Example: Landsat satellites used restoration to improve earth observation data.
Experimental Example:
• Introduce atmospheric blur model h(x,y)
• Restore using frequency-domain filtering
2. Noise Models
(Gaussian, Rayleigh, Salt & Pepper, etc.)
Noise models help us understand the type of disturbance.
Application 1: Camera Sensor Design
Historic Example: Early CCD sensors had thermal noise (Gaussian).
Experimental Example:
• Capture dark image (no light)
• Analyze histogram → Gaussian distribution
Application 2: Wireless Transmission
Historic Example: Satellite communication affected by atmospheric noise.
Experimental Example:
• Add random noise during transmission simulation
• Compare received vs original image
Application 3: Medical Imaging
Historic Example: X-ray images affected by impulse noise due to hardware faults.
Experimental Example:
• Add salt-and-pepper noise
• Test median filter performance
Application 4: Laser Imaging
Historic Example: Used in industrial measurement systems with gamma noise.
Experimental Example:
• Simulate exponential/gamma noise
• Compare histogram with theoretical PDF
3. Spatial Filtering
(Mean, Median, Adaptive filters)
Used for noise removal directly in image space.
Application 1: Photo Noise Reduction
Historic Example: Early Photoshop used mean/median filters.
Experimental Example:
• Add salt-and-pepper noise
• Apply median filter
• Compare pixel values
Application 2: Medical Imaging
Historic Example: Adaptive filters improved CT scan diagnosis accuracy.
Experimental Example:
• Compute local mean & variance
• Apply adaptive filter
• Observe edge preservation
Application 3: AI Preprocessing
Historic Example: Early face recognition systems failed without noise removal.
Experimental Example:
• Feed noisy image to ML model
• Compare accuracy before/after filtering
Application 4: Document Processing
Historic Example: OCR systems required clean text images.
Experimental Example:
• Scan noisy document
• Apply mean filter
• Run OCR and compare results
4. Frequency Domain Filtering
(Band reject, notch filters)
Used to remove periodic noise using Fourier transform.
Application 1: Periodic Noise Removal
Historic Example: TV transmission interference causing stripes.
Experimental Example:
• Convert image to Fourier domain
• Identify spikes
• Apply notch filter
Application 2: Radar Imaging
Historic Example: Military radar affected by interference signals.
Experimental Example:
• Add sinusoidal noise
• Apply band-reject filter
• Compare frequency spectra
Application 3: Biomedical Imaging
Historic Example: ECG interference in imaging systems.
Experimental Example:
• Add periodic noise
• Remove using frequency filtering
Application 4: Satellite Communication
Historic Example: Periodic interference in space communication.
Experimental Example:
• Simulate sinusoidal noise
• Remove using notch filter
5. Wiener Filtering (MMSE)
(Minimizes mean square error)
Most advanced restoration technique.
Application 1: Old Image Restoration
Historic Example: Restoration of old photographs in archives.
Experimental Example:
• Add blur + noise
• Apply Wiener filter
• Measure MSE reduction
Application 2: Audio Processing
Historic Example: Telephone systems used Wiener filtering for noise reduction.
Experimental Example:
• Add noise to audio signal
• Apply Wiener filter
• Compare waveform clarity
Application 3: Medical Imaging
Historic Example: Used in ultrasound imaging improvement.
Experimental Example:
• Add noise
• Apply Wiener filter
• Compare SNR
Application 4: Communication Systems
Historic Example: Used in early digital communication receivers.
Experimental Example:
• Simulate noisy channel
• Apply Wiener filter
• Measure error rate