Stereo 3D Viewer

This project provides a GUI-based 3D viewer for displaying a stereo point cloud. The main purpose is to rectify stereo images, compute the disparity map, and then reproject the disparity into a 3D point cloud that can be visualized interactively. The GUI includes sliders for adjusting stereo matching and depth parameters to observe their impact on the 3D reconstruction in real-time.

Features

• Stereo rectification and disparity computation.
• Interactive GUI using Open3D for visualization.
• Adjustable parameters for stereo matching, such as block size, number of disparities, uniqueness ratio, etc.
• Filter points based on adjustable minimum and maximum depth.
• Toggleable axes display for orientation.

Dependencies

You can install the required dependencies using:

pip install numpy opencv-python open3d

Running the Code

To run the project, you need stereo images named left_0.png and right_0.png in an images/ directory. You can execute the following command to start the viewer:

python3 points_cloud_using_StereoSGBM.py

This will launch an interactive window where you can modify parameters and see their effect on the generated 3D point cloud.

GUI Elements

• 3D Viewer: Displays the generated point cloud.
• Sliders: Adjust various parameters for disparity calculation and depth filtering:
  • Min Disparity: Minimum disparity used for SGBM.
  • Num Disparities: Controls the range of disparities used for the matching process.
  • Block Size: Size of the block used for matching.
  • Uniqueness Ratio: Uniqueness check ratio for validating disparity matches.
  • Speckle Window Size and Speckle Range: Used for filtering out small speckles in the disparity map.
  • Min Depth and Max Depth: Control the range of depth values displayed in the point cloud.

How It Works

1. Stereo Rectification: The left and right images are rectified to align corresponding rows, which makes the disparity computation more straightforward.
2. Disparity Computation: Disparity is computed using the Semi-Global Block Matching (SGBM) algorithm.
3. Point Cloud Generation: The disparity map is used to reproject points into 3D space. Color information from the left image is used to colorize the point cloud.
4. Interactive Visualization: The generated point cloud is displayed using Open3D, and you can adjust various parameters interactively.