🖼️ Image Processing Utilities Documentation

This documentation covers two Python modules: error-metrics.py and reconstruction.py. Both modules provide essential tools for image analysis and restoration. Read on for detailed explanations, workflows, and usage.

---

error-metrics.py

This module provides utilities for quantitatively comparing two images. It computes standard error metrics commonly used in image processing and computer vision.

Main Features

• Image loading & conversion to grayscale
• Calculation of image error metrics:
  • Mean Squared Error (MSE)
  • Root Mean Squared Error (RMSE)
  • Peak Signal-to-Noise Ratio (PSNR)
  • Structural Similarity Index (SSIM)
• Automatic resizing for metric compatibility
• Command-line interface for quick metric evaluation

---

📦 Functions Overview

Function	Description
load_img	Loads an image, converts to grayscale, normalizes to float [0,1]
calc_mse	Computes the Mean Squared Error between two images
calc_rmse	Computes RMSE between two images
calc_psnr	Computes the PSNR between two images
eval_ssim	Computes the SSIM between two images
run_img_comp_metrics	Loads two images and prints all error metrics

---

---

Function Details

load_img(img_path)

• Loads a grayscale, float-precision image from disk.
• Handles RGB and grayscale files.
• Prints shape and dtype info.

def load_img(img_path):
    ...

calc_mse(img1, img2)

• Computes pixel-wise Mean Squared Error.
• Resizes images if shapes differ.

def calc_mse(img1, img2):
    ...

calc_rmse(img_a, img_b)

• Calculates RMSE using the MSE function.

def calc_rmse(img_a, img_b):
    ...

calc_psnr(img_one, img_two)

• Calculates the Peak Signal-to-Noise Ratio in dB.
• Handles identical images as a special case.

def calc_psnr(img_one, img_two):
    ...

eval_ssim(img_src, img_tgt)

• Uses the SSIM metric from scikit-image.
• Resizes if dimensions do not match.

def eval_ssim(img_src, img_tgt):
    ...

run_img_comp_metrics()

• Reads two images and prints all metrics.

---

Typical Usage

To use this module:

1. Set the paths in orig_img_loc and comp_img_loc inside run_img_comp_metrics().
2. Run the script.

Example Output:

--- Image Comparison Results ---
MSE ('original.png' vs 'compressed.png'): 0.0032
RMSE: 0.0566
PSNR: 32.12 dB
SSIM: 0.92
-----------------------------------

---

Error Metric Definitions

Metric	Description	Output Range
MSE	Mean squared pixel error	0 (identical) +
RMSE	Square root of MSE	0 (identical) +
PSNR	Ratio of max signal to noise (in dB)	0 - inf; higher=better
SSIM	Structural similarity (perceptual)	-1 to 1; 1=identical

---

---

reconstruction.py

This module provides a pipeline for image restoration and enhancement. It applies several advanced techniques to denoise, deblur, sharpen, and upscale grayscale images.

Main Features

• Image loading & normalization
• Fourier-based denoising
• Radon streak removal (sinogram smoothing)
• Unsharp masking for sharpening
• Upscaling + Wiener deconvolution to enhance resolution
• Batch pipeline for automated enhancement

---

📦 Functions Overview

Function	Description
load_image	Load and normalize image
denoise_fourier_transform	Denoise using a low-pass mask in the frequency domain
remove_radon_streaks	Remove streak artifacts using Radon transform and median filtering
sharpen_image_unsharp	Enhance edges via unsharp masking
upscale_and_deconvolve_image	Upscale and restore sharpness via Wiener deconvolution
run_image_enhancement_pipeline	Orchestrates the full pipeline; saves result

---

🧩 Pipeline Architecture

Image Enhancement Pipeline

flowchart TD
    A[Input image file] --> B[load_image]
    B --> C[denoise_fourier_transform]
    C --> D[remove_radon_streaks]
    D --> E[sharpen_image_unsharp]
    E --> F[upscale_and_deconvolve_image]
    F --> G[Save enhanced result]

Loading

---

Function Details

load_image(file_path)

• Reads an image, converts to grayscale, and normalizes to [0,1].

def load_image(file_path):
    ...

denoise_fourier_transform(input_img)

• Applies a low-pass circular mask in the frequency domain to suppress high-frequency noise.
• Uses FFT for fast computation.

def denoise_fourier_transform(input_img):
    ...

remove_radon_streaks(input_img)

• Computes the Radon transform (sinogram).
• Median-filters the sinogram to suppress streak artifacts.
• Reconstructs the image using the inverse Radon transform.

def remove_radon_streaks(input_img):
    ...

sharpen_image_unsharp(original_img)

• Enhances edges using unsharp masking (subtracts a Gaussian-blurred version).
• Increases detail contrast.

def sharpen_image_unsharp(original_img):
    ...

upscale_and_deconvolve_image(low_res_img, scale_factor=2)

• Upscales the image using interpolation.
• Simulates a point spread function (PSF) and applies Wiener deconvolution to restore detail.

def upscale_and_deconvolve_image(low_res_img, scale_factor=2):
    ...

run_image_enhancement_pipeline()

• Executes the above steps in sequence.
• Saves the enhanced output image.

---

Typical Usage

1. Set in_file to your input image path in run_image_enhancement_pipeline().
2. Optionally, adjust out_file for the output location.
3. Run the script.

---

🧑‍🔬 Algorithm Summary Table

Stage	Purpose	Techniques Used
Denoising	Reduce high-frequency noise	Fourier transform, masking
Radon streak removal	Remove linear artifacts, stripes	Radon transform, median filter
Sharpening	Enhance edges	Unsharp masking, Gaussian blur
Upscale & Deconvolution	Restore lost detail after upscaling	Wiener filter, PSF modeling

---

📈 Data Flow Example

sequenceDiagram
    participant User
    participant Pipeline as EnhancementPipeline
    participant Disk

    User->>Pipeline: Start with image file path
    Pipeline->>Disk: Load image
    Pipeline->>Pipeline: Denoise (Fourier mask)
    Pipeline->>Pipeline: Remove Radon streaks
    Pipeline->>Pipeline: Sharpen (Unsharp mask)
    Pipeline->>Pipeline: Upscale & Deconvolve
    Pipeline->>Disk: Save enhanced image

Loading

---