Scribd
Upload
2 views

Computer Vision Assignment-1

Uploaded by

vijaymugen
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
2 views

Computer Vision Assignment-1

Uploaded by

vijaymugen
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Image Processing Analysis Report

1. Image Pyramids Construction and Applications

Performance Analysis (10 marks)

Image pyramids are crucial in multi-scale image analysis, enabling efficient and hierarchical processing of

images.

Steps in Constructing Image Pyramids:

- Gaussian Pyramid:

1. Start with the original image.

2. Apply Gaussian blur.

3. Downsample the image.

4. Repeat for multiple levels.

- Laplacian Pyramid:

1. Construct a Gaussian pyramid.

2. Upsample and subtract to get Laplacian levels.

Applications:

- Gaussian: Object detection at multiple scales.

- Laplacian: Image compression and blending.

Clarity of Analysis (10 marks)

Gaussian pyramids reduce image size for efficiency, Laplacian pyramids capture edge-like details.

Presentation & Structure (10 marks)

2. Normalized Cuts vs Mean-Shift Segmentation

Performance Analysis (10 marks)

Normalized Cuts: Graph-based, globally optimal, expensive.

Mean-Shift: Density-based, locally optimal, more efficient for small images.

Output Quality:
Image Processing Analysis Report

- Normalized Cuts: Better object boundaries.

- Mean-Shift: Smoother, color-consistent regions.

Efficiency:

- Normalized Cuts: High computational cost.

- Mean-Shift: Faster on simple tasks.

Representation (10 marks)

Criteria | Normalized Cuts | Mean-Shift

----------------------|--------------------------|-------------------------

Basis | Graph partitioning | Density estimation

Output | Global optimization | Local clustering

Efficiency | Computationally expensive | Faster for low dimensions

Accuracy | High boundary accuracy | Smooth segments

Clarity of Analysis (10 marks)

Each method has trade-offs. Normalized Cuts excels in structured segmentation; Mean-Shift is simple and

effective for color segmentation.

Presentation & Structure (10 marks)

3. Edge-based vs Region-based Segmentation

Performance Analysis (10 marks)

Edge-based: Detects object boundaries via gradients.

Region-based: Groups pixels by similarity.

Applications:

- Edge-based: Medical imaging, document scanning.

- Region-based: Satellite image segmentation.


Image Processing Analysis Report

Clarity of Analysis (10 marks)

Edge-based may fragment boundaries in noisy data.

Region-based more robust but needs tuning to avoid over-merge.

Presentation & Structure (10 marks)

You might also like