Facsimile and static presentation processing – Facsimile – Specific signal processing circuitry
Patent
1988-01-19
1989-12-26
Peng, John K.
Facsimile and static presentation processing
Facsimile
Specific signal processing circuitry
358135, 358136, H04N 718, H04N 712
Patent
active
048901601
DESCRIPTION:
BRIEF SUMMARY
1.0 FIELD OF THE INVENTION
There are many picture processing applications in which knowledge of the speed and direction of movement of all parts of the TV picture would be very useful. These applications include standards converters, noise reducers, bandwidth reduction schemes, and others where any sort of temporal interpolation is required. Even information concerning pans and the movement of the larger objects in the scene would be very useful.
1.1 PRIOR ART
Various techniques have been used to measure motion in TV pictures. The most promising of these appeared to be a method based on phase correlation, which has been shown to be capable of measuring large movements of the complete scene to sub-pixel accuracy. (PEARSON, J. J. HINES, D. C., GOLOSMAN, S., KUGLIN, C. D. 1977 Video-rate Image Correlation Processor. S.P.I.E. Vol. 119. Application of Digital Image Processing (IOCC 1977)).
1.2 THE INVENTION
The object of the present invention is to extend the known technique to measure the motion vectors in a scene containing many objects moving in different directions and speeds.
The invention is defined with particularity in the appended claims. The preferred practice of the invention is explained in section 2 below.
2.0 BASIC PHASE CORRELATION TECHNIQUE
The process of calculating motion vectors for every pixel in the picture is broken down into two stages. The first stage involves correlating two successive pictures (or fields, depending on the exact application) to determine the principal motion vectors present in the scene. The second stage attempts to assign one of these vectors to every pixel. For some pixels it may be impossible to assign a motion vector, for example if the pixel corresponds to a very small object or to uncovered background.
2.1 STAGE ONE--VECTOR MEASUREMENT
In the first stage of the process, 2-dimensional Fast Fourier Transforms (FFTs) of the luminance components of two successive pictures are calculated. Then for each spatial frequency component in the transforms, a unit length vector is calculated whose phase angle is equal to the difference in the phases of this frequency in the two pictures. A reverse FFT is performed on the resulting complex array, which produces an array of real numbers giving the correlation between the two pictures. Mathematically, if G.sub.1 and G.sub.2 are the discrete 2-dimensional Fourier transforms of the two successive images, then the complex array Z is calculated at every spatial frequency (m,n) using ##EQU1## and the phase correlation is given by the inverse Fourier transform of Z, which will only have real components.
The resulting phase correlation array can be thought of as a surface whose height at a particular point (x,y) is proportional to how well the two images correlate when the relative displacement between them is (x,y). In the case of a simple shift between the two pictures, the correlation surface would be a delta function centered on the shift vector. The idea is that there will be a peak in this surfaced for each dominant motion vector in the scene. Measuring these motion vectors involves hunting for large peaks in the surface. The relative heights of the peaks will reflect the relative sizes of the moving objects. The main novel feature of this method is to look for several peaks rather than just one, thereby allowing the detection of many velocities in one operation.
To measure the motion vectors to sub-pixel accuracy it is necessary to perform some interpolation on the correlation surface.
In order to measure as many of the velocities present in a scene as possible, it helps to divide the picture up into blocks, rather than to perform correlations on whole pictures. This is because the number of individual peaks that can be detected accurately is limited by noise to about 3 peaks per block. In addition it is only possible to resolve velocity peaks if they are separated by a shift vector greater than about one pixel per field period. The block size would be large compared with the largest shifts that are expected, as the t
REFERENCES:
patent: 4460923 (1984-07-01), Hirano et al.
patent: 4689672 (1987-08-01), Furukawa et al.
patent: 4689673 (1987-08-01), Ohki et al.
patent: 4727422 (1988-02-01), Hinman
patent: 4733298 (1988-03-01), Koga
Martin, W. N., and J. K. Aggarwal, "Survey Dynamic Scene Analysis", Computer Graphics & Image Processing, vol. 7, No. 3, (1978), pp. 356-374.
Pratt, William K., "Correlation Techniques of Image Registration", IEEE Transactions on Aerospace and Electronic Systems, May 1974, pp. 553-558.
Anandan, P. and Richard Weiss, "Introducing a Smoothness Constraint in a Matching Approach of the Computation of Optical Flow Fields", IEEE 1985, pp. 186-194.
Pearson, J. J. et al., "Video-Rate Image Correlation Processor", SPIE, vol. 119, Application of Digitial Image Processing, (1977), pp. 197-205.
British Broadcasting Corporation
O'Connell Robert F.
Peng John K.
LandOfFree
TV picture motion vector measurement by correlation of pictures does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with TV picture motion vector measurement by correlation of pictures, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and TV picture motion vector measurement by correlation of pictures will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-1579832