Visual Features and Representations: Edge, Blobs, Corner Detection

Understanding visual features such as edges, blobs, and corners is crucial for many computer vision and image processing applications. These features help algorithms extract meaningful information for tasks like segmentation, object detection, tracking, and recognition.

1. Edge Detection

Definition:
Edge detection identifies sharp intensity changes in an image, often corresponding to object boundaries or regions with significant texture change.

Key Points:

• Edges are detected where there is a discontinuity in brightness or color.

• Provides essential structural information about scene objects.

Common Algorithms:

• Sobel Edge Detection: Computes gradients in horizontal and vertical directions to highlight edge regions.

• Canny Edge Detection: A multi-step algorithm involving smoothing, finding gradients, non-maximum suppression, and hysteresis thresholding to deliver thin and strong edge maps.

• Prewitt, Laplacian, Roberts, Scharr Operators: Various filters targeting different orientations or properties of edges.

Applications:

• Image segmentation.

• Object detection and recognition.

• Motion analysis.

• Measuring and counting objects in industrial systems.

Summary Table: Edge Detection Algorithms

2. Blob Detection

Definition:
Blobs are image regions differing in properties (e.g., brightness, color) from their surrounding area. Detecting blobs involves finding and localizing these uniform areas.

Key Techniques:

• Laplacian of Gaussian (LoG): Smooths image with a Gaussian filter, then applies the Laplacian to find regions (blobs) where intensity rapidly changes.

• Difference of Gaussian (DoG): Fast approximation of LoG, subtracts two blurred images (with different degrees of smoothing).

• Determinant of Hessian (DoH): Uses second-order derivatives to localize blobs by analyzing local curvature.

Applications:

• Image segmentation.

• Object tracking in videos.

• Detecting abnormalities (tumors, defects) in medical and industrial images.

• Counting objects or identifying keypoints in scene analysis.

Blob Detection Workflow:

1. Threshold image to separate potential blobs from the background.

2. Group connected pixels to form candidate blobs.

3. Analyze blobs' properties (area, center, radius) for further processing.

3. Corner Detection

Definition:
Corners are points where edges intersect or where there is a significant directional change in intensity. They are highly distinctive, making them ideal for matching features across images.

Popular Algorithms:

• Harris Corner Detector: Uses gradients to determine if a pixel has significant changes in all directions, identifying it as a corner when true for two principal directions.

• FAST (Features from Accelerated Segment Test): Compares a pixel's intensity with its neighbors to efficiently detect corners, suitable for real-time applications.

• Moravec Detector: Measures the variation of image intensity in various directions around each pixel.

• Shi-Tomasi, SUSAN: Variants for specific performance and robustness trade-offs.

Applications:

• Feature matching (e.g., panorama stitching).

• 3D reconstruction.

• Tracking—identifying persistent, recognizable points over time in video.

Corner Detection Steps:

1. Compute local image gradients.

2. Analyze local structure via mathematical measures (e.g., eigenvalues).

3. Apply non-maximum suppression to extract distinctive corner points.

Summary Table: Feature Types

Robust detection and representation of edges, blobs, and corners form the foundation for most vision systems, providing stable features for higher-level analysis, recognition, and decision-making.