Key Terms in Computer Vision
Core Concepts
- Image Processing: Manipulating digital images to enhance or extract information.
- Computer Vision: Teaching computers to interpret and understand visual information.
- Feature Extraction: Identifying essential characteristics of an image.
- Feature Matching: Comparing features between images.
- Image Segmentation: Dividing an image into meaningful regions.
- Pixel: Smallest unit of a digital image.
- Image Resolution: Number of pixels in an image.
- Color Space: Model describing colors and how they relate.
- Image Noise: Random variation of brightness or color in an image.
- Image Enhancement: Improving image quality for better perception.
- Edge Detection: Finding boundaries between regions in an image.
- Corner Detection: Identifying points of significant curvature.
- Blob Detection: Identifying regions with similar properties.
- Texture Analysis: Describing image patterns.
- Shape Analysis: Analyzing the geometric properties of objects.
Object Recognition and Analysis
- Object Recognition: Identifying and classifying objects in an image.
- Object Detection: Locating and identifying objects within an image.
- Object Tracking: Following objects over time in a video sequence.
- Image Classification: Assigning an image to a predefined category.
- Image Retrieval: Finding images similar to a query image.
- Image Segmentation: Dividing an image into meaningful regions.
- Image Registration: Aligning multiple images to a common coordinate system.
- Stereo Vision: Creating depth information from two images.
- Optical Flow: Estimating motion between image frames.
- Motion Estimation: Determining the movement of objects in a scene.
Deep Learning
- Convolutional Neural Network (CNN): Neural network architecture for image analysis.
- Feature Maps: Output of a convolutional layer.
- Pooling: Reducing the dimensionality of feature maps.
- Fully Connected Layer: Layer connecting all neurons in one layer to all neurons in the
next. - Overfitting: Model performs well on training data but poorly on new data.
- Activation Function: Introduces non-linearity to neural networks.
- Backpropagation: Algorithm to adjust weights in a neural network.
- Batch Normalization: Normalizes inputs to a layer for faster training.
- Dropout: Regularization technique to prevent overfitting.
- Hyperparameters: Parameters set before training, not learned.
Image Processing Techniques
- Histogram Equalization: Adjusting image contrast.
- Morphological Operations: Processing images based on shape.
- Canny Edge Detector: Edge detection algorithm.
- Harris Corner Detector: Corner detection algorithm.
- SIFT (Scale-Invariant Feature Transform): Feature detection and description
algorithm. - SURF (Speeded-Up Robust Features): Feature detection and description algorithm.
- HOG (Histogram of Oriented Gradients): Feature descriptor for object detection.
- LBP (Local Binary Patterns): Texture descriptor.
- Gabor Filters: Texture analysis.
- Fourier Transform: Frequency domain representation of an image.
Evaluation Metrics
- Precision: Ratio of correct positive predictions to total positive predictions.
- Recall: Ratio of correct positive predictions to actual positive cases.
- Accuracy: Overall correctness of a model.
- F1-Score: Harmonic mean of precision and recall.
- Mean Average Precision (mAP): Evaluation metric for object detection.
- Intersection over Union (IoU): Overlap between predicted and ground truth bounding
boxes. - False Positive Rate (FPR): Ratio of false positives to total negatives.
- False Negative Rate (FNR): Ratio of false negatives to total positives.
- True Positive Rate (TPR): Ratio of true positives to total positives.
- True Negative Rate (TNR): Ratio of true negatives to total negatives.