Image Processing from Scratch

Introduction

This is my course project for Image Processing at HIT.

In this project, I implemented the following common image-processing procedures.

All procedures except image io are implemented from scratch, without using any external library.

The IO module that involves image read/write/show uses OpenCV.

Maybe I will write those image-reading functions by myself if I have extra time. :)

Color Converters

• RGB<->YUV color space converter
• RGB<->HSI color space converter
• RGB<->HSV color space converter

Histogram-Based Image Enhancing

• Histogram Equalization
• Histogram Specification

Noise Filters

• Median Filter
• Fast Median Filter
• Mean Filter
• Fast Mean Filter
• Gradient Filters: Sobel, Roberts, Prewitt

Results

The outputs of all test cases in main.cc are located in output directory. Those file names are self-explanatory.

Getting Started

This project uses CMake and C++ 14 standard. Here are some options for you to get it compiled.

Compiling With IO Module io.cc & io.h

You need to have OpenCV installed in your system first, then follow these steps in project directory.

Create build directory

mkdir build && cd build

Build the project

cmake ..
make

Run the demo

./improclab

Compiling Without IO Module io.cc & io.h

You cannot run the demo if no IO module presents. But you may still integrate this project in your own projects.

Just remove main.cc, io.cc, io.h, and have fun. At this time, no OpenCV required.

Compiling With C++ Standard < 14

The utility function IDX(...) in util.h is a black magic that uses variadic template pack. It converts multi-demensional indices to 1D(linear) indice in a general way. To understand its use case, you can see My answer at StackOverflow.

Its definition requires some features of C++ 14. If C++ 14 is not available in your project, you can just remove it and use the macros IDX_2D() and IDX_3D() instead, which are defined at the top of util.h.

IDX(x,y,w) can be replaced by IDX_2D(x,y,w)

IDX(x,y,z,w,d) can be replaced by IDX_3D(x,y,z,w,d)

Now this project is C++ 11 compatible!