Image analysis – Histogram processing – For setting a threshold
Patent
1992-02-04
1994-01-11
Boudreau, Leo H.
Image analysis
Histogram processing
For setting a threshold
358105, 382 49, 382 56, G06K 900
Patent
active
052789152
DESCRIPTION:
BRIEF SUMMARY
The present invention relates to a method for hierarchical estimation of the movement in a sequence of images based on the decomposition of the spectrum of the images into sub-bands.
The conventional algorithms for estimation of movement working directly on the images and based on techniques of correlation or of gradient do not give satisfactory results in the case of large movements and in the uniform areas of the images. Moreover, the movement of the fine details of the images is generally very badly rendered.
The aim of the invention is to remedy the abovementioned drawbacks.
To this end, the subject of the invention is a method for hierarchical estimation of the movement in a sequence of images, characterized in that it consists in effecting a hierarchical decomposition of the image by levels of resolutions, in estimating an overall movement at the coarsest level of resolution and in refining the estimate in each level as the resolution increases.
The invention has the principal advantage that it permits calculation of a displacement field of vectors between two successive images of a sequence. This field can be defined for each pixel of the image, with a precision of displacement less than 1 pixel. Moreover, with the addition of two pyramids of high-frequency images with growing resolutions it permits the movement of fine details of the image to be restored, which would disappear in a simple single-pyramid low frequency estimate.
The method according to the invention also has the advantage of being able to be used in all the image processing devices which necessitate a field of movement close to the real field such as the devices for coding by compensation for movement, in the context, for example, of the applications for high-definition television, or also the devices for 50HZ/60HZ standard conversion, for de-interlacing of frames or for 50HZ/100HZ conversion.
Other characteristics and advantages of the invention will appear below with the aid of the description which follows given with reference to the attached drawings wherein
FIG. 1 is a multi-resolution pyramid;
FIG. 2 is a decomposition of an image into sub-bands;
FIG. 3 is a mode of embodiment of an elementary module for decomposition of an image into sub-bands;
FIG. 4 is a mode of embodiment of an elementary module for reconstitution of an image from its sub-bands.
FIG. 5 is a division of the spectrum of the image following three successive iterations.
FIG. 6 is a tree for decomposition of the image obtained from a cascade of three elementary cells.
FIG. 7 is a three-pyramid structure obtained for four levels of resolution.
FIG. 8 is a movement vector represented between two successive images.
FIGS. 9 and 10 is two examples of projection of displacement vectors of an image of coarse resolution towards an image of finer resolution.
FIG. 11 is an illustration of the possible choices between proposed displacement vectors.
FIG. 12 is an example of fusion of estimates obtained on the high-frequency pyramids for the calculation of the estimate on the low-frequency pyramid.
FIG. 13 is an illustration of the processing algorithm employed by the method according to the invention.
FIG. 14 is an interpolation curve permitting the fields of movement to be estimated with a precision less than one pixel.
According to the invention the analysis of the movement in a sequence of images takes place in a hierarchical fashion. It consists in first of all estimating an overall movement at a coarse level of resolution of the image, then in refining this estimate as the resolution of the images processed increases.
FIG. 1 describes a multi-resolution pyramid of images, composed of an assembly of images of resolution 2.sup.-j where j belongs in the interval [o,j], the fullresolution image corresponding to j=o. Thus, the estimate made at a level of resolution 2.sup.-j serves for initialization of the estimate made at the higher level of resolution 2.sup.-(j-1). The invention uses 3 such multi-resolution pyramids, one pyramid of low-frequencies images and 2 pyramids of
REFERENCES:
patent: 4809347 (1989-02-01), Nash et al.
patent: 5027201 (1991-06-01), Bernard
patent: 5050230 (1991-09-01), Jones et al.
Proceedings of the Conference on Computer Vision and Pattern Recognition Jun., 1983, Washington, pp. 246-252; P. J. Burt et al.: "Multi-resolution flow-through motion analysis".
Proceedings of the Conference on Computer Vision and Pattern Recognition Jun., 1986, Miami Beach pp. 219-226; L. R. Williams et al: "A coarse-to-fine control strategy for stereo and motion on a mesh-connected computer".
Proceedings of the Conference on Computer Vision and Pattern Recognition, Jun., 1983, Washington, pp. 432-441; F. Glazer: "Scene matching by hierarchical correlation".
Proceedings of the Eusipco-88, Fourth European Signal Processing Conference Sep., 1988, Grenoble pp. 246-250; J. A. M. Stuifbergen et al.: "Hierarchical estimation of spatio-temporal intensity changes in images".
Chupeau Bertrand
Pecot Michel
"Thomson-CSF"
Boudreau Leo H.
LandOfFree
Method for hierarchical estimation of the movement in a sequence does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method for hierarchical estimation of the movement in a sequence, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for hierarchical estimation of the movement in a sequence will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-1637044