Top Banner

Click here to load reader

Object Based Video Coding - A Multimedia Communication Perspective Muhammad Hassan Khan 2004-03-0020

Dec 21, 2015

ReportDownload

Documents

  • Slide 1
  • Object Based Video Coding - A Multimedia Communication Perspective Muhammad Hassan Khan 2004-03-0020
  • Slide 2
  • Recap Motivation for Video Coding Todays Video Coding Problems with todays video coding Desirable Features Solution to get desirable features Object Based Video Coding MPEG-4 Support Model Based Coding Major Problem: Segmentation Segmentation by Graph Cuts
  • Slide 3
  • Overview of Todays Presentation Details of the Segmentation Process Segmentation using Graph Cuts Results What can we do once we have the segmented regions Block-based Vs Parametric Motion Representation Compatibility with MPEG-4
  • Slide 4
  • Segmentation using Graph Cuts Lets quickly see what is a graph cut!
  • Slide 5
  • Segmentation using Graph Cuts Lets now see what is Max Flow Min Cut 3 4 7 2 2 2 9 9 2
  • Slide 6
  • Segmentation using Graph Cuts How does it relate to segmentation of images? It is primarily a pixel labeling problem Consider we want to label a pixel D = Distance Function (Depends on the current pixel) S = Smoothness Function (Depends on the neighborhood) To be minimized = D + (1- ) S serves as a prior! Hence graph based segmentation answers the question: What is the best segmentation, given this function? We still havent answered how the two relate
  • Slide 7
  • Segmentation using Graph Cuts Let us construct a simple graph to see how the two (graph cuts and segmentation of images) relate D() D() S
  • Slide 8
  • Segmentation using Graph Cuts Start with an initial labeling Find the Min-Cut Adjust the labels Iterate until a good minimization of the function is reached
  • Slide 9
  • Results
  • Slide 10
  • Slide 11
  • Slide 12
  • What can we do once we have the segmented regions? Shape Description Generalized Hough Transform R-Table based representation We need to know a few things Centroid of a shape Texture Model Not explored in detail yet!
  • Slide 13
  • Example for Centroid 00000000 00000000 00011000 00100100 01000010 01000010 00100100 0001100 00000000 00000000
  • Slide 14
  • Shape Description-Finding Centroid For each boundary point Find r = (x, y) x c = x + x y c = y + y is the angle which the tangent at (x, y) makes with the x-axis x y (x c, y c ) r x y
  • Slide 15
  • Shape Description-Creating R-Table R-Table x y (x c, y c ) r x y
  • Slide 16
  • x y =0 =45 =90 =135 =180 =225 =270 =315 Shape Description-Creating R-Table
  • Slide 17
  • Encoding The R-Table This can heavily exploit the redundancy between the magnitudes and directions of R- Table entries We might as well go for DPCM Heaven Knows Benefits Objects encoded independently and can hence be manipulated independently in the transform domain
  • Slide 18
  • Block-based Vs Parametric Motion Representation Block based Use variable block sizes within the segmented object based on texture model Use smaller blocks around the boundary pixels Parametric Motion We know that given that the world is planer, two images taken from a perspective camera of the same scene are related by a projective transformation We can assume each object to lie in a plane, similar to the concept of VOP, and compute the projective transformation parameters to estimate motion
  • Slide 19
  • Compatibility with MPEG-4 Hierarchical Description The scene divided into objects Our Shape/Texture Representation Goes Here
  • Slide 20
  • References Gary J. Sullivan, Pankaj Topiwala, Ajay Luthra SPIE Conference on Applications of Digital Image Processing XXVII, Special Session on Advances in the New Emerging Standard: H.264/AVC, August, 2004 Gabriel Antunes, Abrantes, Fernando Pereira, MPEG-4 Facial Animation Technology : Survey, Implementation and Results, IEEE Transactions on Circuits and Systems for Video Technology, Vol. 9, No. 2, March 1999 Roger H Clarke, Image and Video Compression: A Survey Department of Computing and Electrical Engineering, Heriot-Watt University, Riccarton, Edinburgh EH14 4 AS, Scotland. Noel Brady, MPEG-4 Standardized Methods for the Compression of Arbitrarily Shaped Video Objects, IEEE Transactions on Circuits and Systems for Video Technology, Vol. 9, No. 8, December 1999 Boykov, Y.; Veksler, O.; Zabih, R.; Fast approximate energy minimization via graph cuts, Pattern Analysis and Machine Intelligence, IEEE Transactions on Volume 23, Issue 11, Nov. 2001 Page(s):1222 - 1239
  • Slide 21
  • References P. Gerken, Object-based analysis-synthesis coding of image sequences at very low bit rates, IEEE Circuits System. Video Technology., vol. 4, pp. 228235, June 1994. T. Kaneko and M. Okudaira, Encoding of arbitrary curves based on the chain code representation, IEEE Trans. Communications., vol. 33, July 1985. P. Nunes, F. Marques, F. Pereira, and A. Gasull, A contour-based approach to binary shape coding using a multiple grid chain code, Signal Process. Image Communications., to be published. Moving Picture Experts Group. [Online]. Available www:http://www.cselt.it/mpeg G. Abrantes and F. Pereira, Interactive analysis for MPEG-4 facial models configuration, in EUROGRAPHICS98Short Presentations, Lisbon, Portugal, Sept. 1998, pp. 1.6.11.6.4.