Boundary Length Measures : It describe the length of the boundary (or perimeter) of an
object in an image. These measures are fundamental for understanding the shape and size
of objects.
Common Measures
1. Perimeter: The total length of the boundary enclosing the object.
o Calculated by summing the distances between consecutive points along the
boundary.
Where (xi,yi)(x_i, y_i) are the coordinates of boundary points.
2. Pixel-Based Perimeter: For digital images, the perimeter is often measured as the
number of pixels that make up the boundary.
3. Approximation Methods: If the boundary is irregular, techniques like the chain code
method or polygon approximation can estimate the perimeter.
Normalized Measures
To compare boundaries of different sizes or shapes:
1. Normalized Perimeter:
o Perimeter scaled relative to the area or bounding box of the object.
2. Compactness:
o A measure of how "round" the shape is, derived from the perimeter.
Applications
• Identifying object shapes (e.g., distinguishing circles, squares, and irregular objects).
• Measuring object size in medical imaging (e.g., tumor boundaries).
• Analyzing shapes in geographic data (e.g., land boundaries).
Boundary Descriptors - Boundary descriptors are ways to describe the properties of the
outline (boundary) of an object in an image. They help analyze and compare shapes by
focusing on the object's edge rather than its entire region.
Types of Boundary Descriptors
1. Length: Measures the total length of the boundary (perimeter).
2. Shape: Curvature: Measures how sharply the boundary bends at different points.
o Aspect Ratio: Compares the height and width of the shape.
3. Compactness: Describes how "circular" the boundary is.
4. Fourier Descriptors: Uses the Fourier Transform to represent the boundary in a
compact way, capturing its overall shape.
5. Chain Codes: Encodes the boundary as a sequence of directions between
consecutive pixels (e.g., up, down, left, right).
6. Convexity: Compares the actual boundary to its convex hull (tightest shape that can
enclose it).
Applications- Same
Example For a star-shaped object:
• Length: The total boundary length is long.
• Curvature: High at the points of the star.
• Compactness: Less circular compared to a circle.
Chain Codes are a way to represent the boundary or contour of an object in a digital image. They
describe the shape of an object by encoding the sequence of directions between consecutive
boundary pixels.
How Chain Codes Work
1. Choose a starting point (usually a boundary pixel).
2. Trace the boundary, recording the direction of movement from one boundary pixel to the
next.
3. Encode the direction into the chain code.
4. Repeat the process until the boundary is fully traced.
Common Direction Schemes
1. 4-Connectivity: Uses 4 directions (up, down, left, right).
o Encoded as: 0: Right --- 1: Up --- 2: Left--- 3: Down
2. 8-Connectivity: Uses 8 directions (diagonal movements included).
o Encoded as: 0: Right -- 1: Top-right--2: Up---3: Top-left---4: Left---5: Bottom-left----
o 6: Down---7: Bottom-right
Advantages of Chain Codes
• Compact representation of an object's boundary.
• Easy to calculate and interpret.
Applications : Same as Moments
Fourier Descriptors - Fourier Descriptors are a method for representing the shape of an
object's boundary in a compact and efficient way using the Fourier Transform. They
transform the spatial representation of a boundary into the frequency domain, allowing for
analysis, comparison, and shape reconstruction.
Applications : Same as Moments
Region Descriptors : They are ways to describe the properties of a specific area or object in
an image. These properties help to understand, compare, or identify the object.
Types of Region Descriptors
1. Size and Shape: Area: How big the region is (number of pixels).
o Perimeter: Length of the boundary.
o Compactness: How "round" the shape is.
o Bounding Box: The smallest rectangle that fits around the object.
2. Position: Centroid: The centre point of the region.
o Orientation: The tilt or angle of the object.
3. Aspect Ratio: The ratio of width to height (e.g., tall vs wide).
4. Convex Hull:
o The "tightest" shape around the object without any dents.
5. Texture:
o Measures how smooth or rough the region looks.
Applications
• Identify objects (e.g., recognizing a leaf by its shape).
• Measure properties (e.g., size of a tumor in medical images).
• Compare regions (e.g., distinguishing between circles and squares).
Example
For a round apple in an image:
• Area: Large.
• Compactness: Close to 1 (almost a circle).
• Centroid: The middle of the apple.
• Texture: Smooth.
Moments are numerical values that capture the shape, size, and orientation of objects in an image.
Moments are widely used in feature extraction, pattern recognition, and image analysis.
3. Normalized Central Moments:
o Central moments normalized to make them scale-invariant.
o Useful for comparing objects of different sizes.
4. Hu Moments: Derived from normalized central moments.
a. Invariant to translation, scale, and rotation.
b. Widely used for shape recognition.
Applications of Moments
1. Object Detection: Identify objects in an image based on shape properties.
2. Shape Analysis: Compare shapes for classification and recognition.
3. Pattern Recognition: Recognize digits, characters, or other features.
4. Pose Estimation: Determine an object's orientation in an image.
5. Medical Imaging: Analyze regions of interest in X-rays, MRIs, etc.
