Remove Moving Objects from Images

Overview

Remove Moving Objects is a Python script that removes moving objects from a set of images taken from the same location. The script aligns the images and computes a static background by applying a median filter. This is particularly useful for scenarios such as:

• Capturing a clean view of a tourist attraction without people.
• Creating a background image of a street scene without cars.
• Enhancing time-lapse photography by removing transient objects.

Features

• Uses ORB (Oriented FAST and Rotated BRIEF) for feature detection and FLANN for image alignment.
• Applies a median filter to remove moving objects and extract the static background.
• Processes images in parallel for improved performance.
• POSIX-compliant command-line interface.

Installation

Prerequisites

Ensure you have Python installed along with the following dependencies:

pip install opencv-python numpy imageio psutil tqdm

Usage

Run the script with the following command:

python script.py /path/to/images output.jpg

Parameters

• /path/to/images - Directory containing the images (.jpg format).
• output.jpg - Filename for the resulting background image.

If incorrect parameters are provided, the script will display usage instructions.

Algorithm Explanation

1. Image Alignment

   • ORB (Oriented FAST and Rotated BRIEF) detects keypoints.
   • FLANN (Fast Library for Approximate Nearest Neighbors) matches keypoints.
   • Homography transformation is applied to align images.

2. Background Extraction

   • Images are stacked and split into color channels.
   • The median of each pixel value across images is computed.
   • The channels are merged to produce a clean background image.

Example Use Case

Suppose you have a set of images taken from a busy square. Running:

python script.py images/ cleaned_background.jpg

Will generate cleaned_background.jpg with moving objects removed.

License

This project is open-source and available under the MIT License.