Match Anything (ELOFTR)

This function block performs keypoint matching between two images to find corresponding points and visualize matches. Use it when you need to compare two views of the same scene, verify alignment, or track correspondences between image pairs.

📥 Inputs (sockets)

Image A The first image to match (can be color or grayscale).

Image B The second image to match (can be color or grayscale).

📤 Outputs (sockets)

Visualization Annotated image that shows the two inputs side-by-side with matched keypoints and optional connecting lines.

Keypoints A List of matched keypoints in Image A (coordinates).

Keypoints B List of matched keypoints in Image B (coordinates).

Scores Matching confidence scores for each matched pair.

🕹️ Controls

Show Matches Toggle that enables or disables drawing lines between matched keypoints on the visualization.

Threshold Slider to set the minimum matching confidence (higher values keep only stronger matches).

🎯 Features

• Matches keypoints between two images and returns matched point coordinates and confidence scores.

• Optional visualization that places the two images side-by-side and marks matched points (and lines if Show Matches is enabled).

• Works with color or grayscale inputs.

• Adjustable confidence threshold to filter weaker matches.

⚙️ Running mechanism

When the block runs it accepts two image inputs and computes corresponding keypoints between them. The block filters matches by the configured Threshold and, if requested, prepares a visualization image that shows the matched keypoints and (optionally) connecting lines. The matched coordinates and scores are provided through the corresponding outputs.

📝 Usage instructions

1. Provide two related images to Image A and Image B (for example, two views of the same scene).

2. Adjust Threshold to control how strict matching should be.

3. Toggle Show Matches to turn on/off lines between matched points in the visualization.

4. Use the Visualization output to inspect matches and the Keypoints A, Keypoints B and Scores outputs for downstream processing.

💡 Tips and Tricks

• If input images have very different sizes, normalize them first using Image Resize to improve matching stability.

• To focus matching on a region of interest, crop images with Image ROI or Image ROI Select before feeding them into this block.

• Visualize results quickly by connecting the Visualization output to Show Image.

• Combine with Feature Detector when you want to compare classical feature detections with model-based keypoint matches.

• If you later need to draw detection boxes or overlays based on matches, use Draw Detections or Draw Point with the keypoint coordinates.

🛠️ Troubleshooting

• No matches or very few matches: lower the Threshold slightly, ensure the input images contain overlapping content, or try resizing/preprocessing.

• Many low-confidence matches: increase the Threshold to filter out weak correspondences.

• Visualization not appearing: ensure the Visualization output is connected to a viewer block like Show Image.

• Images with little texture or repetitive patterns may produce ambiguous matches—try focusing on a more distinctive ROI or improving image quality with Contrast Optimization.