Facsimile and static presentation processing – Facsimile – Specific signal processing circuitry
Patent
1991-02-26
1993-03-09
Kostak, Victor R.
Facsimile and static presentation processing
Facsimile
Specific signal processing circuitry
358136, H04N 718
Patent
active
051930013
DESCRIPTION:
BRIEF SUMMARY
BACKGROUND OF THE INVENTION
The invention relates to the domain of the encoding/decoding of video images, in particular with the aim of transmitting high-definition images via a transmission channel which allows the addition of decoding-auxiliary digital data with a throughput equal to 1 Mbit/s. The subject of the invention is in particular a method of segmenting a pointwise field of motion of images with a view to encoding it to the limit of 1 Mbit/s, and its application to the encoding of video images.
DISCUSSION OF FIELD OF THE INVENTION
The most powerful currently known methods for transmitting digital video images implement methods of processing prior to transmission which enable reduction of the line flow rate transmitted. Thus for example, instead of transmitting all the information (chrominance, brightness) relating to all the pixels, the processing makes it possible to seek the differences between images and in particular the displacements of elements of images or of pixels and to transmit only the information for updating the image, namely the vectors characterising the displacement of the zones or of the pixels from one image to the next. A powerful method of processing consists in establishing one vector of motion per point and in transmitting this field of motion.
The flow rate of 1 Mbit/s is too low to enable the transmission of a pointwise field of motion for the video images in full; consequently, it is important to carefully choose the information, relating to the field of motion, which will be transmitted so as to obtain the best possible restoration of the images, the encoding of the transmitted information being established from the preceding image and from the field of motion associated with the current image.
A first arrangement adopted for the processing of the field of motion arising from a pointwise estimator of motion is to transmit only one motion vector per block of n.times.n pixels (n always being a power of 2). On the basis of a high-definition grid of 1440 points per line and 1152 lines, the resulting flow rate is 1.3 Mblocks for blocks of 8.8.
In order to optimise the encoding whilst still keeping to a flow rate lower than the prescribed flow rate (1 Mbit/s), the encoding method referred to as "quad-tree" has proven to be very interesting. This method in fact makes it possible to employ blocks of different sizes and therefore to reduce the block flow rate. To this end, the problem is to return the most uniform possible pointwise field of motion in sets of blocks.
It would then be possible to envisage constructing the histogram of the pointwise field of motion in its entirety. From the outset this procedure is global, this giving a large volume of data to be processed. Moreover, in order to next assign a vector to each of the n.times.n blocks, each vector acknowledged as dominant from the histogram must be compared with all the motion vectors of an n.times.n block so as to choose the best vector for the block. This method of segmenting is therefore weighty and possibly lengthy.
SUMMARY OF THE INVENTION
The subject of the invention is a method of segmenting the field of motion obtained by an estimator of motion enabling extraction, in a manner optimal for a subsequent processing by "quad-tree" encoding, of the "dominant" vectors which will be transmitted after encoding.
To this end, the segmenting method according to the invention consists firstly in carrying out a local processing on the field of motion in order to assign to each n.times.n block the "best" vector of motion from the set of vectors of motion which are available for the block; it is only then that a global processing enables extraction of a set of dominant vectors from the n.times.n blockwise field of vectors, which already comprises a reduced number of vectors and which can be considered as the best block field of motion which may be obtained from a pointwise field of motion.
The phase of reassigning an n.times.n block vector after extraction of the dominant vectors is thus facilitated.
Moreover, it woul
REFERENCES:
patent: 4661849 (1987-04-01), Hinman
patent: 4901149 (1990-02-01), Fernando et al.
patent: 4965666 (1990-10-01), Haghiri
patent: 5027203 (1991-06-01), Samad et al.
Kostak Victor R.
Thomson Consumer Electronics
LandOfFree
Method of segmenting the field of motion of an image and its app 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 of segmenting the field of motion of an image and its app, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method of segmenting the field of motion of an image and its app will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-213696