Image-Processing-Toolbox

(ES 203: Digital Systems Project, Prof. Joycee Mekie, IITGN)

Table of Contents

• Overview/Introduction
• Features
• Project Implementations
  • PC-based (Read/Write from System)
  • FPGA-based (Live on Basys3 with VGA Display)
• Prerequisites
• Getting Started / Setup Instructions
• Usage Instructions
  • Python Scripts for .coe Generation
  • PC-based Implementation
  • FPGA-based Implementation
• Directory Structure Overview
• Technical Details
  • Block Memory (BRAM)
  • VGA Interface
  • COE (Coefficient) File Format
• Contributing
• License

Overview/Introduction

This Image Processing Toolbox is a project developed for the Basys3 FPGA, primarily using Verilog for hardware description and Python for image-to-binary conversion tasks. It enables users to perform various image processing operations, particularly convolution-based techniques. The workflow involves converting input images to a binary format (.coe files) via Python scripts, loading these files into the FPGA's Block RAM, processing the image based on user selection, and then either writing the results back to a PC or displaying them in real-time on a VGA monitor. Development and FPGA implementation are carried out using the Xilinx Vivado software suite.

Features

The toolbox offers a range of image processing functions, selectable via the sel_module in the Verilog code. Some functions also allow parameter adjustments (e.g., brightness) using the val input. The available operations include:

Operation	sel_module	Image
RGB2Gray	0000
Increase brightness	0001
Decrease brightness	0010
Color Inversion	0011
Red Filter	0100
Blue Filter	0101
Green Filter	0110
Original Image	0111
Average Blurring	1000
Sobel Edge Detection	1001
Edge Detection	1010
Motion Blurring xy	1011

Project Implementations

This project offers two distinct methods for performing image processing operations:

PC-based (Read/Write from System)

• Workflow:
  1. An image file resides on your PC.
  2. A Python script (e.g., scripts/coe_generator.py) converts this image into a .coe (Coefficient) file format.
  3. A Verilog module reads this .coe file from the system.
  4. The selected image processing operation is performed in Verilog.
  5. The processed image data is written back to the PC as a new image file (e.g., .bmp).
• Details: Verilog's built-in file I/O functions are used for reading the .coe file and writing the output image. It's crucial to understand the .bmp image structure for correct output.
• Location: Relevant Verilog files and examples for this approach can be found in the BIPT/ and Blurring/ folders.

FPGA-based (Live on Basys3 with VGA Display)

• Workflow:
  1. An image file resides on your PC.
  2. Python scripts (e.g., scripts/parallel_image_generator.py and scripts/kernel_coe_generator.py) are used to prepare and convert this image into a .coe file.
  3. The generated .coe file is manually loaded into the Block RAM (BRAM) of the Basys3 FPGA (typically using Vivado).
  4. The Verilog design on the FPGA performs the selected image processing operation in real-time. Operations are often selected using physical switches or buttons on the Basys3 board.
  5. The processed image is displayed live on a VGA monitor connected to the Basys3 board.
• Details: This implementation provides a more interactive experience but is also more complex. It requires a Basys3 FPGA board, a VGA monitor, and is constrained by the available BRAM size on the FPGA, limiting the dimensions of processable images.
• Location: Relevant Verilog files and examples for this approach are located in the Final Project/VGA_1/ folder.

Prerequisites

To use this toolbox, you will need:

• Software:
  • Xilinx Vivado Design Suite (used for Verilog development and FPGA implementation)
  • Python (for .coe file generation from images)
    • Common Python image processing libraries (e.g., Pillow/PIL) might be required for the Python scripts.
• Hardware (for FPGA-based implementation):
  • Basys3 FPGA board
  • VGA Monitor

Getting Started / Setup Instructions

1. Clone the repository:

   git clone https://github.com/Gowtham1729/Image-Processing-Toolbox.git
   cd Image-Processing-Toolbox
   

2. Software Setup:
   • Xilinx Vivado: Ensure Vivado is installed and properly configured for the Basys3 board if you plan to use the FPGA-based implementation.
   • Python: Ensure Python is installed. Install necessary libraries like Pillow for image manipulation:

     pip install opencv-python numpy
     

Usage Instructions

Python Utility Scripts

This project includes several Python scripts located in the scripts/ directory. They are used for generating Xilinx COE files from images and for preparing images for processing. All scripts are now run from the command line with arguments specifying input and output paths.

• scripts/parallel_image_generator.py:

  • Purpose: Prepares an input image for kernel-based operations by generating multiple versions of it: the original image (named gray.bmp for compatibility, though it can be color) and 8 images shifted by one pixel in cardinal and diagonal directions (e.g., up.bmp, left.bmp, leftup.bmp, etc.). These 9 images are saved as BMP files.
  • Usage: This script is a prerequisite for scripts/kernel_coe_generator.py.
  • Command-line example:

    python scripts/parallel_image_generator.py <input_image_path.bmp_or_jpg> <output_directory_for_shifted_images>
    

    (Example: python scripts/parallel_image_generator.py path/to/your/image.jpg path/to/shifted_output/)

• scripts/coe_generator.py:

  • Purpose: Converts a single input image into a Xilinx COE (.coe) file where each pixel is represented as a 96-bit binary string. This format includes 72 bits of zero-padding followed by the 24-bit BGR (Blue, Green, Red) color data for the pixel. It's suitable for operations processing pixels individually.
  • Command-line example:

    python scripts/coe_generator.py <input_image_path.bmp_or_jpg> <output_coe_path.coe>
    

    (Example: python scripts/coe_generator.py path/to/your/image.jpg path/to/output/image.coe)

• scripts/kernel_coe_generator.py:

  • Purpose: Generates a Xilinx COE file specifically for convolution-based operations. It combines data from the 9 images generated by scripts/parallel_image_generator.py. Each line in the output COE file is a 96-bit binary string:
    • The first 72 bits are derived from the blue channels of the corresponding pixels in the 9 shifted/grayscale images.
    • The next 24 bits are the BGR color data from the original (unshifted) color image.
  • Prerequisite: Requires the 9 shifted/grayscale images previously generated by scripts/parallel_image_generator.py.
  • Command-line example:

    python scripts/kernel_coe_generator.py <main_color_image_path.bmp_or_jpg> <directory_containing_shifted_images> <output_coe_path.coe>
    

    (Example: )

General Python Script Usage: All scripts are executed from the terminal using Python 3. They now accept command-line arguments for input and output paths, providing flexibility in specifying files and locations. Ensure you have the necessary dependencies installed (e.g., opencv-python, numpy) as listed in the Prerequisites and Getting Started / Setup Instructions sections.

PC-based (Read/Write from System)

Process images using Verilog simulations on your PC.

1. Prepare Images (if needed for kernel operations):
   • If you are testing convolution-based operations, first run scripts/parallel_image_generator.py to create the 9 shifted/grayscale versions of your test image.

     python scripts/parallel_image_generator.py path/to/your/test_image.bmp path/to/shifted_images_output_dir/
     

2. Generate .coe file: Use the appropriate Python script from the scripts/ directory.
   • For basic operations (e.g., brightness, color inversion):

     python scripts/coe_generator.py path/to/your/test_image.bmp path/to/output.coe
     

   • For convolution-based operations (e.g., blur, edge detection), use the output from scripts/parallel_image_generator.py:

     python scripts/kernel_coe_generator.py path/to/your/test_image.bmp path/to/shifted_images_output_dir/ path/to/kernel_output.coe
     

3. Configure Verilog: Update file paths or parameters in the Verilog source files within the BIPT/ (basic operations) or Blurring/ (convolution operations) folders to point to your generated .coe file.
4. Simulate: Use Xilinx Vivado (or another Verilog simulator) to run the design. Testbenches will read the .coe file, process data, and write the output to a new image file (often .bmp) on your PC.

FPGA-based Implementation

Run image processing operations directly on the Basys3 FPGA, with output to a VGA monitor.

1. Prepare images for kernel operations:
   • Run scripts/parallel_image_generator.py to generate the 9 shifted/grayscale versions of your input image.

     # Example:
     python scripts/parallel_image_generator.py path/to/your_input_image.png path/to/output_shifted_images_directory/
     

2. Generate the final .coe file for FPGA:
   • Run scripts/kernel_coe_generator.py. This script takes your original color input image and the directory of shifted images (generated in the previous step) to produce the .coe file.

     # Example:
     python scripts/kernel_coe_generator.py path/to/your_input_image.png path/to/output_shifted_images_directory/ path/to/final_fpga_image.coe
     

   • The output .coe file from this step is what you'll load into the FPGA's Block RAM.
3. Load .coe into Block RAM: In your Vivado project for the Final Project/VGA_1/ design, manually initialize the Block RAM (BRAM) with the generated .coe file's contents. This is typically done via the Vivado GUI by associating the .coe file with the BRAM IP core.
4. Synthesize, Implement, and Program:
   • Execute synthesis and implementation for the Verilog design in Vivado.
   • Generate the bitstream file.
   • Program the Basys3 FPGA with this bitstream.
5. Operate: Use the physical switches/keys on the Basys3 board to choose the desired image processing function.
6. View Output: The processed image is displayed in real-time on the connected VGA monitor.

Directory Structure Overview

• BIPT/: Verilog modules for PC-based simulation of basic image processing operations (e.g., brightness, grayscale).
• Blurring/: Verilog modules for PC-based simulation of convolution-based operations (e.g., blur filters).
• Final Project/: The complete Verilog project for Basys3 FPGA implementation, including VGA output and real-time operation selection.
  • Final Project/VGA_1/: Contains the specific VGA controller and top-level module for FPGA.
• images/: Sample input images and example output images referenced in this README.
• scripts/: Contains Python utility scripts:
  • coe_generator.py: Converts a single image to a standard COE file.
  • kernel_coe_generator.py: Generates a COE file for kernel-based operations using a main image and 9 pre-shifted neighbor images.
  • parallel_image_generator.py: Generates the 9 pre-shifted neighbor images required by kernel_coe_generator.py from a single input image.
• tests/: Contains unit tests for the Python scripts, ensuring their functionality and correctness.

Technical Details

This section briefly overviews key technical components.

Block Memory (BRAM)

The FPGA's Block RAM stores image data in binary format (from a .coe file). Python scripts convert images into this format, where each line typically represents pixel data. For convolution operations needing simultaneous access to multiple pixels (e.g., a 3x3 kernel), the Python scripts structure the .coe file to group necessary neighboring pixel data. This allows Verilog modules to efficiently access required BRAM data.

VGA Interface

A Verilog-implemented VGA controller module displays the processed image on a standard VGA monitor. It generates horizontal sync (hsync) and vertical sync (vsync) signals, plus RGB color data, typically for a 480p (640x480 pixels) resolution at 60Hz. The controller reads pixel data from BRAM and translates it into analog signals for the VGA standard.

COE (Coefficient) File Format

A .coe file is a Xilinx-specific ASCII text file for initializing Block RAMs or ROMs in FPGA designs. It usually includes a header specifying the data's radix (binary or hexadecimal), followed by a vector of data values. In this project, these values represent image pixel data. For more details, see the COE File Syntax in the AMD/Xilinx documentation.

Contributing

Contributions are welcome! If you have improvements or bug fixes, please fork the repository, make your changes, and submit a pull request.

License

This project is licensed under the Apache License 2.0. See the LICENSE file for full details.

---

If you find this repository helpful, please consider giving it a star on GitHub! ⭐