Video Shot Change Detection

This project develops an algorithm to detect shot changes in three videos (news, climate, ngc) and assesses its performance. Notably, the climate and ngc videos involve gradual transitions.

Visual Features Used

1. Grayscale Image
2. Histogram (64-bin gray-level)

Shot Change Detection Algorithm

The following algorithm is designed to detect shot changes in the videos:

1. Convert each frame to a grayscale image and compute its histogram.
2. Calculate the Bhattacharyya distance between histograms of consecutive frames.
3. Detect a shot change if the distance exceeds a threshold t_b.
4. Detect a shot change if cumulative distances between t_s and t_b exceed t_b

Usage

1. Clone or download this project.
2. Install the required dependencies:

   pip install -r requirements.txt
   

3. Run the main script:

   python main.py
   

Results and Analysis

1. News
   Precision: 0.8571428571428571
   Recall: 0.8571428571428571
2. Climate
   
3. NGC