 |
Claims  |
|
|
We claim:
1. In an apparatus for generating signals in real time representative of
visual information, in particular visual information relating to movements
of the human body, the apparatus being of the type comprising: a camera
suitable for forming successive images and for converting each such image
into a sequence of corresponding electrical signals; binary conversion
means for converting the signal representative of each point in the
sequence into binary form; and encoding means for encoding the resulting
sequences of binary signals, the improvement which includes means for
compressing said signals to a data rate compatible with transmission over
an ordinary telephone line by virtue of said binary conversion means which
link the camera to the encoding means, said means for compressing
comprising means responsive to changes in brightness between neighboring
points within each image to extract contours present in the original
image, said binary signals then being representative of said extracted
contours.
2. Apparatus according to claim 1, including sampling means connected to
receive the output signal from the contour extracting device, said
sampling means being connected to select signals relating to certain
points only in such a manner as to reduce the contour image by a
predetermined factor.
3. Apparatus according to claim 1, wherein the data rate at the output from
the encoding means is not greater than 4,800 bauds.
4. Apparatus according to claim 1, further including compression means
linking the output from the contour extractor device to the encoder means,
said compression means comprising movement selecting means for selecting
information representative of contour movement.
5. Apparatus according to claim 4, wherein said movement selecting means
includes means for performing point-by-point comparison of each successive
contour image with the preceding contour image.
6. Apparatus according to claim 1, further including filter means for
filtering out signals representing isolated points in said contour images.
7. Apparatus according to claim 4, further including filter means connected
to receive the output signal from said movement selecting means filter out
signals from each image sequence representative of modifications of
information from one image to the next and relating to points which are
isolated in the image.
8. Apparatus according to claim 7, wherein said filter means operates to
filter out information concerning isolated points only when the density of
points surrounding said isolated points is below a predetermined
threshold.
9. Apparatus according to claim 7, wherein said movement selecting means
comprises a memory for storing information relating to points in the image
preceding the current image, and wherein means are provided for correcting
the information stored in said memory to correspond with the image signal
applied to the encoding means after rejection of isolated points.
10. The separation defined in claim 1 in a transmitter/receiver system for
producing signals representative of visual information at a rate
compatible with transmission over telephone type lines, said visual
information being in particular representative of movement of the human
body, and said receiver being suitable for receiving such signals and for
displaying animated cartoon type images representative of said movement,
further comprising means for connecting said encoder to an interface for
connection to a telecommunications line, together with a receiver also
suitable for connection to said interface and including means for decoding
signals received serially via said interface from a telecommunications
line and for feeding signals to a device for displaying successive images.
11. A system according to claim 10, wherein the transmitted digitally
encoded signals convey binary information relating to image points which
have changed since the preceding image to be transmitted, and wherein the
receiver includes a memory for storing a sequence of signals
representative of the preceding image, and a device for reconstituting the
current image on the basis of the received signals and the stored image
signals, said reconstituting device being connected to feed image data
into said memory and the display means being connected to read data from
said memory.
12. A system according to claim 10 or 11, wherein the receiver includes a
digital memory having a capacity corresponding to a plurality of
successive contour images in addition to the image being reconstituted,
thereby enabling successive images to be displayed at a substantially
uniform rate in spite of differences in the time taken to transmit
successive images.
13. A method for transmitting information based on movements of the human
body, between a first and a second station, remote from each other, said
method comprising the steps of:
(a) acquiring at said first station video signals representing successive
images of said human body;
(b) converting said video signals into first digital signals, each
corresponding to a point in one of said successive images;
(c) processing said digitial siganls, within each image, to determine for
each point of the image, a binary signal indicating whether that point
belongs to a contour of said image;
(d) encoding such binary signals indicating a contour point into second
digital signals;
(e) transmitting said second digital signals to said second station through
a telecommunication network;
(f) receiving said second digital signals at said second station;
(g) reconstituting the contours of the successive images, at said second
station, from said second digital signals, and
(h) displaying the contour images at the second station.
14. The method according to claim 13, wherein step (e) further comprises
transmitting alphanumeric information, at least during the initial
transmission period, while step (g) is reconstituting a first full image
to be displayed, and step (h) comprises displaying said alphanumeric
information.
15. The method according to claim 13, further comprising also implementing
said steps (a) to (h) from the second station to the first station.
16. The method according to claim 13, wherein said step (d) of encoding
comprises:
sampling said binary signals of the image by blocks of adjacent points, and
forming a reduced binary signal from the binary signals corresponding to
the points in each block.
17. The method according to claim 16, wherein step (d) further comprises:
encoding the binary signals indicating a contour point in a second portion
of the current image into second digital signals for said second portion,
which is disjointed from said first portion;
said step (g) comprises:
reconstituting the second portion of the current image from the
correspondingly received binary signals; and
storing these as an updating of the second portion of the preceding image;
said first and second portions being varied from one image transmission to
another, so as to have the stored preceding image entirely refreshed over
a predetermined period of time.
18. The method according to claim 16 wherein said step (d) of encoding
comprises:
storing the binary signals corresponding to at least one preceding image,
comparing the binary signals in a first portion of the current image, for
each respective point in said portions;
forming those of said second digital signals corresponding to the first
image portion only for binary signals of the current image being different
from the corresponding ones in the preceding image; and
updating said stored preceding image with such different binary signals in
the first portion of said stored image; and step (g) of reconstituting
comprises:
storing the binary signals received corresponding to at least one preceding
image; and
reconstituting said first portion of the current image by updating the same
portion of the preceding image with the binary signals received for the
first portion of the current image.
19. The method according to claim 18, wherein step (d) of encoding
comprises:
filtering out those of said different binary signals which correspond to
substantially isolated points;
forming said second digital signals from the unfiltered ones of said
different binary signals; and
inverting in said stored preceding image of the first station, the binary
signals corresponding to said filtered out binary signals of the current
image. |
|
 |
|
 |
Claims |
|
|
|
Description |
|
|
FIELD OF THE INVENTION
The present invention relates to telecommunications, and in particular to
telecommunications by telephone or telematics lines. The invention aims
more particularly at transmitting images over such links, where it is
understood that telephone links are not limited to cables or electric
lines but also include communication links such as microwaves, optical
fibers, etc.
BACKGROUND OF THE INVENTION
Telephone lines have traditionally been developed and installed in networks
for transmitting voice communications. More recently, they have found a
use in transmitting digital type data, by means of modulator-demodulators
generally known as modems. Such telephone links have transmission
characteristics which are well adapted to transmitting analog signals at
frequencies of not more than about 4 kHz. When they are used for
transmitting digital signals, they tend to be useable at data rates of up
to about 4800 bauds. Such performance limits the volume of information
which may be transmitted in given time.
Thus, while telephone lines are well adapted to transmitting human speech
or to transmitting computer messages in the form of relatively low data
rate binary signals, they are not suitable for real time transmission of
images which are liable to change quickly in time. Image transmission
requires a relatively large amount of information to be transferred in
comparison to the volume of information needed for a voice message. The
image must be divided into point-like zones or "pixels" which are smaller
with increasing definition to be transmitted. A signal representative of
the brightness of each pixel must be produced. The set of such signals
must then be transmitted, generally in series in the form of a sequence.
Each sequence corresponds to an image, and the level of the signal at a
given instant in the sequence corresponds to the brightness of a
corresponding point in the transmitted image.
Such sequences are transmitted, as is well known, in television systems. In
order to show moving scenes in such systems, the transmitted images must
be renewed at a relatively high rate. It is necessary to have links
capable of transmitting a frequency band whose width is orders of
magnitude greater than the bandwidth of single telephone links.
Attempts have already been made to improve the comfort of telephone
communications by transmitting the parties' images to each other by means
of such a TV-like system which is generally called a visiphone. The
development of such systems has been held back by their cost which is
particularly related to the fact that the information transmission
capacity required for transmitting adequate images is twenty to forty
times greater than the capacity needed for a speech only telephone line.
When the images to be transmitted are relatively static, proposals have
already been made for reducing the data transmission rate. One such system
is described in a paper entitled "Video teleconferencing at 9600 bauds" by
Robert H. Wallis and William K. Pratt in the "IEEE Picture Coding
Symposium, Montreal 1981. An opto-electronic cine-camera is used to
produce a transmissible image of a subject such as the head and shoulders
of a speaker in a remote conference. The resulting electrical signals are
converted into binary in such a manner that each "pixel" is represented by
a signal capable of taking only two values. The resulting sequence
obtained for the image is then subjected to a coding process capable of
compressing the information content in the sequence so that it may be
transmitted over a telematics line at a rate of about 9,600 bauds. Images
obtained by this procedure are of relatively poor quality, even when the
image renewal rate is limited to only a few images per second. This
technique is thus not suitable for real time re-transmission of scenes in
which movement must be produced to some degree of accuracy.
OBJECTS OF THE INVENTION
It is the object of the present invention to provide a method and apparatus
for real time transmission of varying visual information over a telephone
line or a telematics line having equivalent data rate performance. The
invention is aimed in particular at transmitting information corresponding
to movements of the human body, eg. to gestures or to lip movements. In
this respect, one application of the invention consists in supplying a
telecommunications system for people suffering from deafness but who are
capable of communicating by sign language or of understanding speech by
lip reading.
SUMMARY OF THE INVENTION
To this end, the present invention provides a method of transmitting visual
information over a telecommunications network, the method being of the
type in which the information is observed as a stationary image, digital
signals corresponding to each image point are generated, and the digital
signals are transmitted over a network line to be received by a receiver
connected to the network and then decoded to enable the corresponding
image to be displayed. The improvement in the method lies in converting
the signals representative of each image into binary signals by
discriminating on the basis of differences in brightness between
neighboring points in image in order to obtain a sequence of electric
signals representative of the contours in the original image, said
contour-representing signals then being digitally encoded and transmitted
over the network.
In a preferred implementation, the contour-representing signals are
compressed before being encoded for transmission. The compression may take
various forms, which may be applied in a cumulative manner.
One preferred form of compression consists in detecting variations between
one image and the next, said variations being representative of movement
by the subject whose image is to be transmitted. The detection may take
place by a point-by-point comparison of the successive images in such a
manner as to forward only signals representative of those parts of the
contour which change position from one image to the next. At the receiver
end, an image reconstruction process is then used which receives said
movement-representing signals together with signals representative of one
or more previously reconstructed images to reconstruct the current image.
Thus the invention is based in the observation that there are applications
in which the richness of image provided by a television system is not
strictly necessary, and in which the subject matter to be communicated can
usefully be represented by a contour image analogous to a line drawing.
The drawing changes over time as the subject in the original scene moves.
The transmitted image can thus be compared to an animated cartoon. It has
been observed in particular that transmission of such a cartoon-like image
is quite acceptable for lip-reading or for communicating sign language
such as used by the deaf and dumb. For sign language in particular, the
drawing-like nature of the transmitted image can enhance understanding.
One particularly advantageous implementation of the method consists in
extracting contours from a relatively high definition image, and then in
sampling the high-definition contour image in blocks of four or nine (ie.
reducing groups of four or nine adjacent high definition points to a
single low-definition point) thereby reducing processing and transmission
capacity requirements downstream from the sampler by a factor of four or
nine. It may also be advantageous to filter signals prior to their
application to a telephone line interface in such a manner as to eliminate
isolated points in the image or drawing to be transmitted.
The invention also provides a device for real-time generation of electrical
signals representative of visual information and which may be used to
implement the above-defined method. For this purpose, such a device
comprises, in particular, a camera, a contour extractor device, and means
for compressing the data in each image-representing data sequence for
transmission over a telephone line or for intermediate storage at a data
rate comparable to the data capacity of a telephone line. In addition to
an encoding device, the compression means may include means sensitive to
changes between successive "drawings" from the contour extractor. The
compression means may also include filter means for eliminating isolated
points and possibly also for distinguishing points whose position does not
vary significantly from one image to the next.
The invention also provides a transmitter/receiver system for producing
signals representative of visual information using the principles outlined
above and for displaying an image on the basis of signals generated in
that manner. The receiver portion includes, in particular, display means
suitable for reproducing successive images of the animated cartoon
transmitted by such signals. When the encoding takes changes from one
image to the next into account, the receiver should include storage means
for successive reconstituted images.
BRIEF DESCRIPTION OF THE DRAWING
An embodiment of the invention is described by way of example with
reference to the accompanying drawings, in which:
FIG. 1 is an overall block diagram of a transmitter/receiver system
suitable for connection to a telephone line;
FIGS. 2A to 2D are diagrams showing the effect of a sampling device on
contour images;
FIGS. 3A to 3C show two successive animated cartoon contour images and a
third image showing the difference between the first two due to movement
of the subject;
FIGS. 4 and 5 show "windows" used in a filtering stage; and
FIGS. 6A and 6B show two different situations for a FIG. 5 window.
SPECIFIC DESCRIPTION
In addition to enabling speakers to see each other, devices for
transmitting images over telephone links also enable them to make gestures
and to point to things. This amounts to an improvement in the quality of
communication which is taking place essentially verbally. Such devices
could in theory enable communication with a person who is totally deaf but
who is capable of lip-reading, or enable communication between two deaf
and dumb people communicating by sign language. However, in so far as
visiphone devices are spreading extremely slowly because of the very large
quantities of information they need to transmit and the corresponding
complexity of the transmission networks that need to be installed, the
possibilities of telecommunications open to deaf people are, in practice,
very limited.
The studies that lead to the present invention have shown that accurate
transmission of lip movements, or of gestures for sign language, must take
place at not less than about 12 images per second. This rate is necessary
both because of the rapidity of certain gestures and because of the need
to avoid eye-strain in the observer by too jerky an image.
Ordinary telephone lines can be used to transmit at up to about 4800 bauds.
Such a rate is quite inadequate when using ordinary television type image
transmission procedures for transmitting an image such as the head or the
head, body and arms of a speaker.
FIG. 1 shows an example of a transmitter/receiver station in accordance
with the invention for transmitting visual information such as the
movements of the human body. The station comprises a transmitter
sub-assembly 10 and a receiver sub-assembly 12 both of which are connected
via a conventional interface or modem 14 to a telephone or telematics line
15.
The transmission sub-assembly comprises a camera 20 having a lens 22
suitable for making an image of a subject. The camera is mounted for this
purpose on a stand, not shown, in such a manner as to obtain a framed
image of a subject facing the transmitter/receiver station 10, 12. The
camera 20 is an electronic camera suitable for transforming each image of
a subject or a scene formed on its screen into a sequence of electronic
signals which appear at an output 23. The signals at the output 23 are
supplied to a contour extracting device 25. The contour extracting device
may be of a known type such as is described below, and serves to transform
the signals of each sequence (each signal representing the brightness
level of a corresponding pixel in a half-tone image), into signals capable
of taking two levels only (eg. corresponding to black or to white). The
black pixels in the resulting image lie for the most part on contours in
the original image. Thus the signals which appear at the output 26 from
the contour extracting device 25 are representative of a kind of cartoon
image animated at a rate of, say, twelve images per second. The data in
the contoured signal is then compressed before being applied to the
interface 14.
The data is compressed by a series of juxtaposed modules including a
sampler 28 which serves to divide the total volume of information to be
transmitted in each image by a predetermined factor, and a motion
extractor circuit 30 having an output 31 providing signals which
correspond to points which have changed brightness level between two
successive contoured images. These signals are then applied to a filter
device 34 serving to eliminate isolated points and having a coder 36
connected to its output for further compressing data as a function of the
data structure as analysed by the coder and then encoded according to a
pretermined set of rules. The output 37 from the coder 36 is connected to
the send input 38 of the interface 14.
The receive output 39 of the interface 14 is connected to the input 41 of a
decoder 40 which applies inverse rules corresponding to the rules used by
the coder 36, thereby reestablishing a sequence of electrical signals at
its output 42 corresponding to an animated cartoon of the type produced by
a second transmitter identical to the device 10, but at the other end of
the line 15. The output signals at 42 are processed by an image
information restoring device 44 before being applied to the input 45 of a
display screen 46 which displays the information transmitted on the line
15 in the form of a sequence of visible images at the appropriate rate.
The camera 20 is preferably of the charge transfer type including a
photosensitive array of 256.times.256 elements such as are now
commercially available. Naturally the choice of camera technology is not a
limiting factor in the invention, and other opto-electronic conversion
devices could also be used. One advantage of the charge transfer type
camera is that it does not require a high tension power supply. Each
photosensitive element in the camera produces an analog signal whose level
is proportional to the brightness received by that element. The pixel
brightness levels thus received for a single image are read sequentially
and may be digitized by an analog to digital converter so as to produce a
frame of 256.times.256 six-bit words at the output 23 of the camera. The
position of each word in the sequence corresponds to the position of a
pixel in the image.
The contour extracting device 25 may be made, for example, in accordance
with the technique described in published French Pat. No. 2 163 815 filed
July 2, 1973 by Nadler, Adamoff and Oisel. Another form of contour
extracting device is described in European patent application No. 81 402
085.5 dated Dec. 28, 1981 which readily transforms an image of a natural
scene into a contour image. The invention is naturally not limited to
either of the above-mentioned contour extracting devices.
Whatever method is used to extract contours, the result is conventionally
interpreted as dark lines on a light background, with the lines
corresponding to boudaries between different parts of the natural image.
An example is shown in FIG. 2A. The contour extracting device reduces the
image to binary levels in that each pixel represented in the signal at its
output 26 is represented by a single bit of value 0 or 1 depending on
whether the pixel is black or white. Contour extracting devices are
sensitive to changes in brightness level between neighboring pixels as
represented in the signal from the camera 20, ie. between words
representative of groups of neighboring pixels. In this respect they
differ from clipping devices for reducing an image signal to 0s and 1s
depending on whether the signal exceeds a threshold level. This difference
remains even if the threshold in question varies according to other
criteria.
One of the important underlying principles of the present invention is the
observation that sufficient visual information could be transmitted on the
basis of the output from a contour extracting device for communication
between the deaf and dumb, and further that this communication could be
achieved at high enough rates for real time transmission of the
information while remaining within the capacity of a telephone line.
The sampler 28 simply serves to compress the image. In practice, the
electrical image at the output 23 from the camera 20 is divided into scan
"lines" each of which corresponds to one line of photosensitive elements
in the array. In the electrical signal each line is separated by an
interline separator symbol. The sampler 28 can thus reduce image width by
extracting one point in two or one point in three (ie. one bit in two or
three as the case may be) along selected line signals, and can also reduce
image height by selecting one line in two or three. Arrows 9 link FIG. 2A
to FIG. 2B showing the compression which results when selecting one in
two. It is clear that FIG. 2B is a compressed version of FIG. 2A.
The sampling may be performed by conventional means, eg. by bit and line
counters, or by a suitably programmed microprocessor.
The resulting compression by a factor of four in image area is shown in
FIG. 2B. In both FIGS. 2A and 2B, the black and white squares correspond
respectively to 1 and 0 level bits in the sequence of signals for each
line. Successive lines are progressively further down the image. FIG. 2C
shows the effect of area compression by a factor of three (one point in
three along one line in three). It has been observed that better results
are obtained by starting from a relatively high definition image at the
output from the contour extracting device 25 (ie. based on a signal from a
camera using a 256.times.256 matrix) and then reducing the image after the
contours have been extracted, than by attempting to extract contours
directly from a camera of lower definition. FIG. 2D shows the result of
extracting contours from the same subject as was used for FIG. 2A, but
using a 128.times.128 matrix camera. It is clear that FIG. 2D is
considerably poorer in information, than is FIG. 2B, or even FIG. 2C.
The data compression begun by the sampler 28 is continued by a series of
compression stages aiming at reducing the sampler output data rate by a
factor of about twenty before the signal is applied to the input 38 of the
interface 14.
The movement information extractor 30 includes a memory 50 capable of
storing an entire contour image as applied to its input 51 by the output
29 from the sampler 28. Suitable memory addressing means are provided. The
output 52 from the memory 50 is connected to one input 55 of a difference
calculator 56 having a second input 57 which is directly connected to the
output 29 from the sampler 28. The difference calculator thus performs a
point-by-point comparison of successive contour images that appear at the
output 29 from the sampler 28. The memory 50 is suitably addressed to
produce corresponding points from the previous image at the input 55 in
time with the arrival of new points at the input 57. The difference
calculator 56 performs an exclusive-OR (XOR) function to provide bits at
its output 31 only when the corresponding image bits at its two inputs are
different.
More precisely, if a given point in a contour image is at level 1 whereas
the corresponding point in the previous image was at level 0, then a level
1 bit appears at the output 31. Similarly, if a given point is currently
at level 0 whereas in the previous image it was at level 1, then a level 1
bit also appears at the output 31. However, in the other possible cases
(both input bits the same, either both at level 1 or both at level 0) a
level 0 signal appears at the output 31. The addressing device for the
memory 50 is arranged to read from the memory slightly before writing new
information to it to ensure that the previous values are used before being
overwritten, this sometimes known as read modified write addressing.
The movement extractor device 30 may be controlled by suitably programmed
microprocessors or by other conventional integrated circuits as is well
known in the electronics art.
FIG. 3A shows a contour image of the face and body of a subject who is
gesturing with the right hand. This figure corresponds to the type of
signal which appears at the output from the contour extractor 25.
FIG. 3B shows the next image in a sequence, and it can be seen that the arm
has moved between images. FIG. 3C is an image showing the differences
between the two preceding images. Sequences of signals at the output 31
from the difference calculator 56 correspond to images like that shown in
FIG. 3C. Such sequences thus correspond to movement by the subject to be
displayed. This thus represents a reduction in the volume of information
that needs to be transmitted when compared to the static contour image.
In this respect, other image compression techniques may be used depending
on the modifications which are found to occur from one image to the next
and on the experience accumulated while transmitting preceding images. In
particular, more than one previous image may be taken into account. Known
techniques for analysing movement can be used to take several preceding
images to predict the current image. In such cases, the information
transmitted concerns only the difference between the actual image and the
predicted image. Clearly such techniques require more complicated means to
replace the XOR function 56, but have the advantage of providing greater
compression for use with a telephone line. Suitable means are described in
an article by H. M. Nagel entitled "Analysis Techniques for Image
Sequences" in Proc. Int. Joint Conf. on Pattern Recognition Kyoto, Japan,
1978, and in a book edited by T. S. Huang, entitled "Image Sequence
Analysis" published by Springer, Berlin, 1981.
The sequence of signals at the output 31 includes two kinds of information:
firstly there are signals corresponding to definite movement of various
parts of the contour between successive images, these signals contain the
information to be tansmitted; and secondly there are numerous "isolated
point" signals due to small and inevitable contrast differences in a
system using a camera 20 of the type described. FIG. 3C shows such
isolated points in addition to the definite movements. Definite movement
is essentially restricted to the hands and arms. Some isolated points are
scattered at relatively low density over the entire body and background.
In contrast there is a high density of "isolated" points in the subject's
face. This is due to the mobility of the face when speaking, and such
mobility is itself a significant feature of signal language as well as
being the essential feature of communication by lip reading.
The filter 34 thus has two functions: firstly it must eliminate isolated
points in areas of low density where the points correspond to little
movement, if any; and secondly it must retain points in high density
regions, which in practise means in the face.
The filter detects isolated points by means of a window function as shown
in FIG. 4. The window shows the values of an image variable in the
neighborhood of a current point of interest. In FIG. 4 a 3.times.3 point
window is used. Given a point of interest i,j in the matrix of image
points being applied serially to the input of the filter, points which
fall in the intersection of columns i-1 to i+1 by lines j-1 to j+1 are
taken into consideration. Let X be the property of being a difference
point, ie. a point corresponding to a change from a 0 to 1 or from a 1 to
0 in the sequence of signals present on the output 31, as indicated above.
X.sub.is is defined as the property of being an isolated difference point,
corresponding to the configuration shown in FIG. 4 where the value of X is
different for the point i,j than for all the surrounding points in the
window under consideration. In Boolean algebra:
X.sub.is =X.sub.i-1,j-1 .multidot.X.sub.i-1,j .multidot.X.sub.i-1,j+1
.multidot.X.sub.i,j-1 .multidot.X.sub.i,j .multidot.X.sub.i,j+1
.multidot.X.sub.i+1,j-1 .multidot.X.sub.i+1,j .multidot.X.sub.i+1,j+1
where X indicates absence of the property.
Thus the signals applied to the input of the filter 34 may thus be
processed as defined above, eg. by means of a memory and by selective
interrogation of the signals thus stored as a function of the selected
window. An integrated circuit or a suitably programmed microprocessor is
used to to control the operation.
Having thus detected isolated difference points in a sequence, the device
analyses the density of such points over the image as a whole to reject
only those isolated points which are in regions having less than a
predetermined threshold density. This is done by using a larger window as
shown in FIG. 5 which uses a 7.times.7 matrix. The window is only used for
observing "isolated" points as determined by a window 60 identical to the
window of FIG. 4. The window 60 is at the center of the FIG. 5 window 62,
and for density measurements, only the 40 points in the window 62 which
surround the window 65 are taken into consideration.
Density is calculated by counting the number of difference points in said
40 points and then comparing the total count with a threshold. Supposing a
threshold of 30% is used, an isolated point 65 will not be retained if
fewer than 12 difference points are counted in the surrounding 40 points
of the window under test. Such a situation is shown in FIG. 6A. Otherwise,
and as shown in FIG. 6B, there are at least 12 difference points in the
surrounding 40, in which case the isolated point 65 is retained and a
corresponding signal is transmitted to the output 57 of the filter 34
towards the input 58 of the coder 36.
Thus, the function for detecting isolated points X.sub.is is applied
successively to all the difference points of the image. However, the
density threshold function is applied only to points which have already
been detected as possible isolated points.
The filter 34 removes genuinely isolated points from the image by applying
a signal to an output 59 which is fed via a line 70 to an input 72 of the
memory 50, to cause the memory to invert the content of the corresponding
location. The reason for this degree of feedback is explained more fully
below.
The coder 36 encodes the sequence of signals appearing at its input 58 for
each image. Several different methods exist for compressing sequential
information representative of a black and white image. The sequential data
present on the input 58 is converted into a series of digital words on an
output 37 for transmission over the line 15 by the interface 14. One
method is "run length" encoding in which each white to black transition (0
to 1) gives rise to a special code followed by a code for the immediately
following black to white transition which code includes the number of
points between said transitions. Other encoding systems can give even
greater compression, eg. block or quad-tree type encoding. Several such
encoding methods are described in the July 1980 issue of the journal
"Proceedings of the IEEE" which was a special graphic image encoding
issue.
At the receiver, the sequences of signals appearing at the receiver output
39 from the interface 14 are of the same kind as the sequences transmitted
by the coder 36, but they come from a coder 36 in another instrument
connected to some other line in the telephone network to which the line 15
is connected. The received signals are decoded by a decoder 40 which
applies the inverse function to that used by the coder 36. When the
decoded signals at the output 42 correspond to difference points between
consecutive images, they are applied to the input 75 of a device 76 for
reconstituting the image inside the image information restoring device 44.
Said restoring device 44 also includes a memory 78 having an input 77
connected to the output 79 from the reconstituting device 76. The output
81 from the memory is connected to the input 45 of a display screen 46
which may be constituted by a cathode ray tube or the like as used in
computer terminals.
The memory 78 is preferably not limited in capacity to a single image. It
should be capable of storing several complete successive images suitable
for being applied one after the other to the output 81 to feed the
display. In particular it should include a portion of memory which is
directly addressable from the input 77 for reconstituting each image
during reception.
The image may be reconstituted in a manner symmetrical to the decomposition
performed by the memory extractor 30. Thus for each sequence at the input
75 corresponding to a current image, the reconstituting device 76 receives
a second signal on a second input 82 connected to deliver the
corresponding signal from the previously reconstituted image as stored in
the memory 78 and output via an output 83. The signals applied to the
inputs 75 and 82 of the device 76 are synchronised in such a manner as to
enable a complete contour image to be reconstituted point by point, with
each corresponding bit in the previous image being updated by the bit
present at the input 75. A sequence of reconstituted images is thus stored
in the memory 78. It should be observed that by virtue of some of the
isolated points being eliminated by the filter 34, the reconstituted image
would not be an exact reproduction of the image in the memory 50 of the
transmitter if the corresponding isolated point had not been inverted in
the memory 50.
The data compression techniques mentioned above for use by the coder 36
introduce a variable delay in image transmission time. Further, decoding
by the decoder 40 and reconstitution by the device 44 which requires
knowledge of a complete prior image in addition to reception of
information concerning the current image both add a certain delay between
the time when information begins to be received at the interface output 36
and the appearance of a complete image on the screen 46. If several images
are stored digitally in the memory 78, at least the image redrawing rate
can be kept substantially uniform on the screen in spite of the variable
overall transmission rate and necessarily with some delay. The redrawing
rate at the screen 46 can then be synchronised with the rate at which
images are generated by the camera 20, even though some individual images
take longer to transmit than others.
Further, because of the overall delay inherent to filling all the stages of
buffer memory in the chain, there is a definite dead period after a call
begins and the appearance of a complete image on the screen. This dead
period can be filled-in by an alphnumeric message which may identify the
caller by name or by number or both. Thus the called party can receive
some information concerning the visual communication which is about to
begin.
A permanent re-initialization device is also provided for re-initializing
the displayed image during transmission. There will inevitably be
transmission errors or interference, and this could lead to the image
being progressively degraded. To avoid this phenomenon, a small portion of
each image from the contour extractor 25 or from the sampler 28 is
transmitted in full to the interface input 38, by a line not shown, to
update the corresponding receiver image. The updating information may
concern about 1% of each image, say one complete line thereof, and
replaces a portion of the compressed data from the coder 37. On reception,
the corresponding portions transmitted in full are forwarded, again by
means not shown, to an appropriate point downstream from the
reconstituting device 76 to reinitialize the corresponding image line.
Suitable signals are naturally tansmitted to announce the beginning and
the end of each portion transmitted "in the clear" to enable the portions
to be properly located. The "clear" information is thus used to update the
entire transmitted image over a period of 100 sequences (supposing a 1%
clear rate as outlined above).
FIG. 1 shows a dashed box 90 inserted between the memory 78 and the screen
46. The dashed box represents an optional image enhancement device
suitable for smoothing between points, thereb | | |
