multimedia-systems

Video Compression and Object Removal using Motion Compensation

Overview

This repository contains a set of Python implementations for video compression and object removal using motion compensation techniques. The work was part of a project for the "Multimedia Systems" course at the University of Piraeus (2022-2023). The project was divided into three tasks, each focusing on a different aspect of video processing: Huffman compression, motion compensation, and object removal.

Project Structure

This project contains the following tasks:

1. Task 1: Video Compression Using Huffman Encoding
2. Task 2: Motion Compensation for Video Compression
3. Task 3: Object Removal from Video Using Motion Compensation

Task 1: Video Compression Using Huffman Encoding

In this task, a video is compressed using Huffman encoding applied to the difference between consecutive video frames. The encoded video is then decoded, resulting in a noisy reconstruction due to the error accumulation in the decoding process.

Key Steps:

• Frame difference calculation.
• Huffman encoding of the difference.
• Decoding process with accumulated errors.

Requirements:

• Python 3
• OpenCV: pip install opencv-python

Task 2: Motion Compensation for Video Compression

This task focuses on using motion compensation to compress a video. The video is split into macroblocks, and motion vectors are calculated to represent the movement between frames. A hierarchical search approach is applied to speed up the process.

Key Steps:

• Convert frames to grayscale.
• Macroblock division and motion vector calculation.
• Image prediction and error image calculation.
• Motion compensation for encoding and decoding.

Requirements:

• Python 3
• OpenCV: pip install opencv-python
• tqdm: pip install tqdm

Task 3: Object Removal from Video Using Motion Compensation

In this task, an object is algorithmically removed from a video by utilizing motion compensation. The frames are split into macroblocks, and motion vectors are adjusted to exclude the moving object.

Key Steps:

• Select the first frame with minimal motion.
• Apply motion compensation across frames to exclude the moving object.
• Reconstruct the video with the object removed.

Requirements:

• Python 3
• OpenCV: pip install opencv-python
• Numpy: pip install numpy

File Descriptions

• encoder1ii.py: Encoder for motion compensation video compression (Task 2).
• decoder1ii.py: Decoder for motion compensation video compression (Task 2).
• function.py: Utility functions for motion compensation (Task 2).
• Thema1i.py: Encoder for Huffman compression (Task 1).
• Thema2.py: Encoder for object removal (Task 3).

Installation

1. Clone the repository:

   git clone <repository-url>
   cd <repository-directory>

2. Install required libraries:

   pip install opencv-python numpy tqdm

3. Run the respective Python scripts for each task:

   • Task 1: python3 Thema1i.py
   • Task 2: python3 encode1_2.py and python3 decode1_2.py
   • Task 3: python3 Thema2.py

Examples

Task 1: Video Compression Using Huffman Encoding

Example of an input video and its encoded/decoded versions:

Original Video:

Encoded Video:

Decoded Video:

Task 2: Motion Compensation for Video Compression

Example of an input video and its encoded/decoded versions:

Original Video:

Encoded Video:

Decoded Video:

Task 3: Object Removal from Video Using Motion Compensation

Example of an input video with an object and the resulting video after object removal:

Before Object Removal:

After Object Removal:

Bibliography

• Multimedia Systems - Algorithms, Standards, and Applications by Parag Havaldar, Gerard Medioni
• OpenCV Documentation: https://docs.opencv.org/3.4/d6/d00/tutorial_py_root.html
• Numpy Documentation: https://numpy.org/doc
• tqdm Documentation: https://tqdm.github.io/

License

This project is licensed under the MIT License.