Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/aretrosen/image-foreground-extraction-and-affine-rectification

This project successfully extracts the foreground of an image, and then applies Affine rectification, much like Adobe Scan.
https://github.com/aretrosen/image-foreground-extraction-and-affine-rectification

Last synced: 10 days ago
JSON representation

This project successfully extracts the foreground of an image, and then applies Affine rectification, much like Adobe Scan.

Awesome Lists containing this project

README 

        
# ADIP & CV (CS60052) Assignment 2



## Installation:

Follow this link to download [Miniconda](https://docs.conda.io/en/latest/miniconda.html "Miniconda Page"). Then, to install required packages, and create environment, run

```bash

conda env create -f environment.yml

```



inside the project directory. Activate `conda` environment by running



```bash

conda activate adip-assignment-2

```



Assuming the OS you run is Linux based, run the `run.sh` script for effortless

viewing of the results. Each `slide` will show after 4 seconds. If the OS is

**not** Linux based, then you can run each of the lines in the `run.sh` file

successively.



All functions are heavily doucumented, and the operations are all

automatic. Where it is not, like the selection of the rectangular area, the best

result by experimentation is presented. Chance is that you won't have to, but

feel free to change any parameter.



## Experimental Results:

Read function docstrings for a overview of all the functions, and what they

do. There are many parameters involved, but I have tried to make it as automatic

as possible, inspired by recent Scan apps. Image wise results:



* **PataChitraPuri_1.jpg** - Runs completely automatic.

* **PataChitraPuri2.jpg** - It **can** run completely automatic, but since the

main painting is above many other paintings, in automatic run, it can't

distinguish the borderline (neither can I at first glance, frankly). I have

provided an approximate bounding rectangle, so that the results are better.



The algorithm for tranformations work on an approximate quadrilateral

calculation from the (estimated) four corner points in the image. Then all

results from all operations are stored asnd shown as a slideshow, with the title

depicting which "operated" image is shown. For further information regarding any

function or OpenCV method, please visit [OpenCV Documentation Page](https://docs.opencv.org/4.5.5/index.html "OpenCV Documentation").



## Usage:

```bash

# Creating and activating conda environment

conda env create -f environment.yml

conda activate adip-assignment-2



# See all options; accessing the help menu

python assignment.py --help



# Without any option, it runs with the PataChitraPuri_1,jpg image, if present

python assignment.py



# Set time interval in seconds between image "slides", here it is 2.5s

python assignment.py --interval 2.5



# Everything automatic, except if corners are hard to detect

python assignment.py --image myimage.jpg



# Manually select four corners, overriding automatic mechanism

python assignment.py --image myimage.jpg --manual



# NOTE: Provide either mask or rectangular area. Providing both as input is

# just counter-intuitive, and the program will run with --rect flag only.



# Manually provide the rectangular part of the image whose foreground we

# need to extract, for example, from (x=1,y=2) of width 600 and height 400

python assignment.py --image myimage.jpg --rect 1 2 600 400



# Manually provide mask for the image

python assignment.py --image myimage.jpg --mask mymask.jpg

```



### Author Information:

	• Name : Aritra Sen.

	• Roll no. : 19ME10101.

	• Department : Mechanical Engineering.

	• Subject : Advanced Digital Image Processing & Computer Vision (CS60052).