Sponsored Links
This presentation is the property of its rightful owner.
1 / 19

# Marr’s framework for vision PowerPoint PPT Presentation

Marr’s framework for vision. 2-1/2D sketch. Primal sketch. Object Recognition. Early processing. 3D estimation. Image. Primal sketch. Local edges Corners T-junctions Blobs Groups of features. Derivatives as edge finders. Edges are sharp changes in image intensity.

### Download Presentation

Marr’s framework for vision

An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.

- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -

2-1/2D

sketch

Primal

sketch

Object

Recognition

Early

processing

3D

estimation

Image

### Primal sketch

• Local edges

• Corners

• T-junctions

• Blobs

• Groups of features

### Derivatives as edge finders

• Edges are sharp changes in image intensity.

• 1st derivative of the image intensity peaks at an edge

• 2nd derivative is zero at edges

• Edges are at zero-crossings of the second derivative

### Multi-scale filtering

• Find zero-crossings at multiple scales

• Filter with Laplacian of Gaussian filters that have different sizes

• Edges = zero-crossing sat all scales

• Find spatial coincidence of zero-crossings across scales

### Mirage

• Filter with three Laplacian of Gaussians (different sizes)

• Seperately sum negative and positive parts

• Mark zero (Z) regions, positive response regions (R+) and negative response regions (R-)

• Rules

• Z region = luminance plateau

• R region with only one Z on a side = edge

• R region with Z on both sides = bar

### Ways to improve edge detection

• Take advantage of the spatial structure of edges

• Edges are oriented

• Use directional derivatives

• Simple cells as directional derivatives

• Compute local oriented contrast “energy”