| 摘要 |
The method is intended to estimate movement in a sequence of animated images by executing a gradient algorithm which minimises the quadratic deviation in the local luminance variations of the current point of the image with the point which corresponds to it in the preceding image frame. It consists in making two different displacement estimates in each filtered frame according to two different resolutions: - a first estimate 6 takes place between the current frame and the frame which precedes it, de-interlaced and filtered at a quarter of their band; - a second estimate 9 takes place between the current frame and the preceding frame, de-interlaced and filtered to a half of their band; and in initialising the gradient algorithm used in each estimate by two initial values of displacement originating from two adjacent causal points which are closest to the current point situated on the preceding line, by inserting the value of displacement of the current point estimated in the first estimate 6 as a value for calculating 9 the second estimate. Each initial displacement value necessitates the acquisition of only 4 points in the memory containing the preceding frame. Application: data rate compression in digital television. <IMAGE> |
