	US 7,489,726 B2
	Resource-constrained sampling of multiple compressed videos
Anthony Vetro, Cambridge, Mass. (US); and Huifang Sun, Cambridge, Mass. (US)
Assigned to Mitsubishi Electric Research Laboratories, Inc., Cambridge, Mass. (US)
Filed on Sep. 11, 2003, as Appl. No. 10/659,904.
Application 10/659904 is a continuation in part of application No. 10/639951, filed on Aug. 13, 2003.
Prior Publication US 2005/0036547 A1, Feb. 17, 2005
Int. Cl. H04N 7/12 (2006.01); H04N 11/02 (2006.01); H04N 11/04 (2006.01); H04B 1/66 (2006.01)

U.S. Cl. 375—240.01

16 Claims

1. A method for processing a plurality of videos, comprising:

acquiring, in parallel, a plurality of compressed videos, in which compressed frames of each input video are acquired at a fixed sampling rate;

applying, concurrently and in parallel, joint analysis to the plurality of compressed videos to determine a variable and non-uniform temporal sampling rate for each compressed video, so that a combined distortion is minimized while a combined frame rate constraint is satisfied for the plurality of compressed videos, in which the combined distortion includes a temporal distortion, and in which the compressed frames are inter-frames; and

sampling compressed frames of each compressed video at the associated variable and non-uniform temporal sampling rate to produce a plurality of compressed output videos having variable temporal resolutions, in which the temporal distortion E{Δ²Z_i,k} between a frame i and frame k is estimated by

where (σ²_x_{_i},σ²_y_{_i}) represent a variances for x and y spatial gradients in frame i, and (σ²_Δx_{_i,k}, σ²_Δy_{_i,k}) represent variances for motion vectors between the frame i and frame k in x and y direction.