Image 2 Sketch

📃 Report | 📓 Explanation Notebook with Code | ⭐ Results

📖 Overview

1. Introduction
2. Current Method
3. My Method (Novel Idea)
4. A Few Results
5. Running it On Your Machine
6. Bibliography

Introduction

There are several methods to extract edges from images. There are also several other methods for object detection and region segmentation. In this project I propose a novel method for Feature extraction from Images and use this method to create sketch composite for the given Image.

Current Method

The current State of the art method is the Gaussian Blur-Blend Method which has the following algorithm:

def gaussian_blend_blur(I: image, k: Kernel Size):
    J = grayscale_of_image(I)
    K = gaussian_kernel_of_size(k, k)
    B = gaussian_blur_of_image_with_kernel(J, K)
    result = J / B
    return result

Novel Method

We introduce a new method for feature extraction, the Orthogonal Gaussian Lattice Method which takes 3 Gaussian and computes teh inverse of the grayscale of the image with these 3 Gaussian and then uses a connectivity bound parameter alpha to create 3 Simple Graphs from these inverse gaussian. These graphs are then used to extract Lattices from the image and these lattices are then used as features and for image sketching.

The algorithm is as follows and for full details please reefr to the Jupyter Notebook with Complete Explanation.

A Few Results

Running it On Your Machine

The program requires the following dependencies:

1. Python 3
2. pip
3. Git

Clone and install all dependencies:

git clone https://github.com/anishLearnsToCode/image2sketch.git
cd image2sketch
mkdir results
mkdir data
pip install -r requirements.txt

To test this program on some sample images see src/test.py. Create a directory /data under the project root and add some sample images there for testing purposes. Preferably add PNG Images as they store full data without compression. Inside src/test.py set the path to yor image and you can also further change Hyper-parameters from the file src/control_parameters.py and run it as

cd src
python test.py 

Results will be displayed and also saved under the directory results/{image_name} where image_name comes from src/test.py.

Bibliography

1. How to create a beautiful pencil sketch effect with OpenCV and Python ~Ask A Swiss
2. Open CV
3. Python 3
4. Git
5. Graph Theory and Its Applications to Image Segmentation ~IEEE
6. Image Sketch Online
7. Photofunia Sketch Effect
8. convertimage.net
9. Convert Image to GrayScale ~KD Nuggets
10. pip