Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Depth image recognition using isomorphic graph theory


Published on

registration of depth-images to each other using graph isomorphism

Published in: Technology, Education
  • Be the first to comment

Depth image recognition using isomorphic graph theory

  1. 1. Depth image recognition using isomorphic graph theory 1
  2. 2. Our project... ● ● ● ● We try to construct a mesh of the world using the data from a depth sensing camera. Having the depth images Transforming to normal images Apply billboard concept ( each region / plane in different color) 2
  3. 3. Image == sub-graph? ● ● Lets assume that we can perfectly recognize objects in an image The image graph is a sub-graph of the real world 3
  4. 4. Image analysis RGB Images Normal Maps Region Growing Images 4
  5. 5. Graph-based object recognition ● ● Objects can be matched using region growing Each object in image → Node ● ● ● RGB histogram is the “color” of the node Average color pixel of each Node(region) Edges are assigned to neighbouring regions 5
  6. 6. Graph-based object recognition (2) ● ● ● In image sequences, nodes represent objects. The nodes of the graph can be used for pose estimation Saved they can be used for global positioning 6
  7. 7. One approach ● ● ● ● ● Recursive color mapping to detect each region One random pixel is given, a contour is obtained by color flooding and a new node is created The process is repeated for every border pixel An edge is created when two regions are neighbors 7
  8. 8. The sequence Nodes Delaunay Voronoi 8
  9. 9. Finding the isomorphisms 9
  10. 10. Finding the isomorphisms 10
  11. 11. In the end ● ● When the largest isomorphic sub-graph is found we can see if the images have matched. Then its just looking at the sizes.. ? 11
  12. 12. Demo-time 12