AL-701 Computer Vision - Nov 2023 Question Paper

Under each question, How to answer lists the points an answer should cover, outlined by CampusPrep from the syllabus. Check the details against your textbook.

1. 7 Marks
   aWhat is computer vision? And how does it differ from traditional image processing?
   How to answer

   • Definition of computer vision as automated visual understanding
   • Goal: high-level interpretation vs pixel manipulation
   • Differences: objectives, output, AI/ML use vs signal processing
   • Examples of each
2. 7 Marks
   bExplain the difference between gray scale and colour image.
   How to answer

   • Grayscale: single intensity channel, 8-bit
   • Colour: 3 channels e.g. RGB, 24-bit
   • Storage and processing differences
   • Use cases for each
3. 7 Marks
   aDescribe the concept of data type conversion in image processing. How does it impact image quality and storage?
   How to answer

   • Need for data type conversion: uint8, double, float
   • Normalization and scaling effects
   • Impact on quality: precision, dynamic range
   • Impact on storage and computation
4. 7 Marks
   bDescribe the operations of erosion and dilation in binary image processing. How can these operations be used for noise removal and object size modification in binary images?
   How to answer

   • Definition of erosion: shrinking foreground with structuring element
   • Definition of dilation: expanding foreground
   • Opening/closing for noise removal
   • Object size modification examples
5. 7 Marks
   aExplain the concept of thresholding in binary image processing. How can you choose an optimal threshold value for a given image? Provide an example.
   How to answer

   • Thresholding definition: $T$ to create binary image
   • Global vs adaptive thresholding
   • Methods to choose optimal threshold: histogram, Otsu
   • Example with histogram
6. 7 Marks
   bProvide an algorithm or method for performing connected component analysis on a binary image. Explain the steps involved.
   How to answer

   • Definition of connected components and connectivity (4/8)
   • Two-pass labeling algorithm steps
   • Equivalence resolution and relabeling
   • Applications: counting, measurement
7. 7 Marks
   aExplain the concept of image enhancement and its importance in image processing. Provide examples of scenarios where image enhancement is necessary.
   How to answer

   • Definition of image enhancement: improve visual quality
   • Importance: contrast, visibility for analysis
   • Spatial and frequency domain examples
   • Scenarios: medical, satellite, low-light
8. 7 Marks
   bDefine colour transforms in image processing. Provide an example of a colour transform and explain how it alters the appearance of an image?
   How to answer

   • Definition of colour transforms between color spaces
   • Example: RGB to HSV / grayscale
   • Mathematical mapping and effect on appearance
   • Use in enhancement and segmentation
9. 7 Marks
   aDiscuss scenarios where precise colour adjustment is essential, and how curves can be employed to achieve the desired result.
   How to answer

   • Scenarios needing precise colour: medical, printing, film
   • Curves for tonal and colour correction
   • Adjusting RGB channels independently
   • Achieving desired look with control points
10. 7 Marks
    bDescribe a real-world application where image smoothing is necessary, and specify which type of blur filter would be most appropriate.
    How to answer

    • Real-world need: denoising, preprocessing for vision
    • Blur filters: mean, Gaussian, median, bilateral
    • Selection criteria based on noise type
    • Recommended filter with justification
11. 7 Marks
    aExplain the role of convolution in image filtering. How is a filter (kernel) applied to an image using convolution?
    How to answer

    • Convolution as weighted sum with kernel
    • Kernel sliding, multiply-accumulate operation
    • Examples: blur, sharpen, edge kernels
    • Borders, stride and padding handling
12. 7 Marks
    bDefine image smoothing and its role in image processing. Why is it essential to reduce noise or enhance an image's appearance?
    How to answer

    • Definition of smoothing: suppress noise/high frequencies
    • Role in preprocessing pipeline
    • Why noise reduction improves analysis
    • Trade-off: detail loss vs appearance
13. 7 Marks
    aWhat is image segmentation and how is it related to image recognition? Describe the steps involved in segmenting objects within an image.
    How to answer

    • Definition of segmentation: partition into regions
    • Relation as precursor to recognition
    • Steps: preprocessing, segmentation, feature extraction
    • Example methods: threshold, clustering
14. 7 Marks
    bExplain the Canny Edge detection algorithm, including its key steps and the role of various parameters.
    How to answer

    • Canny pipeline: smooth, gradient, NMS, double threshold
    • Hysteresis edge tracking
    • Role of sigma and low/high thresholds
    • Advantages: thin, low error edges
15. 7 Marks
    aWhat is OpenCV, and how is it used in deep learning for computer vision tasks?
    How to answer

    • OpenCV as open-source vision library
    • Preprocessing, DNN module, integration with frameworks
    • Use in deep learning: data prep, inference
    • Example tasks: detection, tracking
16. 7 Marks
    bDefine gesture recognition in the context of computer vision. How does it enable human-computer interaction?
    How to answer

    • Definition of gesture recognition from visual input
    • Pipeline: detection, tracking, classification
    • Enabling touchless HCI and control
    • Examples: sign language, gaming