Anime Frame Optimizer (AFOptimizer)

Introduction

Anime Frame Optimizer (AFOptimizer) is a Python-based tool that enhances anime viewing experiences by automatically removing static or 'dead' frames from videos. Using OpenCV, AFOptimizer employs three sophisticated frame analysis methods: Optical Flow, Frame Difference, and Structural Similarity Index (SSIM), each tailored to specific video processing requirements.

Use this project on Google Colab

Features

• Tri-Method Analysis: Utilizes Optical Flow, Frame Difference, and SSIM for comprehensive frame analysis.
• Efficient Frame Removal: Automatically detects and removes static frames, streamlining the viewing experience.
• Customizable Sensitivity: Adjusts movement detection sensitivity for different video types and preferences.
• Command-Line Interface: Easy-to-use CLI for processing videos with specified methods and settings.
• Performance Variation: Each method varies in processing time and efficiency, offering flexibility based on user needs:
  • SSIM: ~2.13 frames/s
  • Frame Difference: ~37.12 frames/s
  • Optical Flow: ~1.18 frames/s
  • Note: Performance metrics tested on a 4vCPU, 8GB RAM replit core environment.

Setup and Installation

To use AFOptimizer, ensure Python and all the dependencies from requirements.txt are installed

Usage

Run main.py with the desired method and video file as arguments:

• python FrameEnhancer.py -of --video=path/to/video for Optical Flow.
• python FrameEnhancer.py -fd --video=path/to/video for Frame Difference.
• python FrameEnhancer.py -ss --video=path/to/video --ssim_threshold=0.9587 for SSIM (the --ssim_threshold flag is optional; if not used, the value will default to 0.9587).

Methods Explained

Optical Flow Method

• How It Works: Calculates motion between frames based on pixel changes.
• Technique: Uses the Farneback algorithm for dense optical flow, providing a motion vector for each pixel.
• Ideal Use: Best for videos where detailed movement detection is crucial.
• Output: Generates a video with smoother transitions by excluding frames with minimal pixel motion.
• Performance: More processing-intensive due to complex vector calculations.

Frame Difference Method

• How It Works: Assesses the difference in pixel values between consecutive frames.
• Technique: Compares grayscale versions of frames and measures the change in pixel intensity.
• Ideal Use: Effective for videos with varying lighting conditions or minimal background movement.
• Output: Focuses on significant movements by discarding frames with minimal changes.
• Performance: Faster than Optical Flow, offering a balance between speed and precision.

SSIM Method

• How It Works: Uses the Structural Similarity Index to evaluate frame similarity.
• Technique: Measures changes in luminance, contrast, and structure between frames.
• Ideal Use: Suited for videos where perceptual frame similarity is essential.
• Output: Removes frames that are perceptually similar to their preceding frames.
• Performance: Quicker than Optical Flow, prioritizing perceptual quality over pixel-level changes.

Development and Contributions

AFOptimizer is in active development. Contributions to enhance functionality, especially in algorithm refinement and feature additions, are welcome.

Contact and Support

For support, contributions, or inquiries, please contact hello@karanprasad.com.