CA1292057C - Apparatus for encoding/transmitting an image - Google Patents

Abstract

ABSTRACT
This invention relates to an apparatus for encoding and transmitting an image. The apparatus includes a vector quantizer. A quantization preprocessing circuit is used for subsampling an image signal, for thinning out pixels from the image signal according to a pixel ratio, and for generating from the pixel signal an n-dimensional vector. A group of input buffers are used to store each input vector in a sequence for quantization processing and for repetitiously reading the input vector at an interval of the processing period. A vector output circuit is used to prepare the output vector corresponding to a probability distribution of the input vector and for selecting and outputting the output vector based on an index input. The distance between the input vector and the output vector is calculated by the distortion computation circuit. An index output generating means is used to generate an index for a processing period based on the calculated distance value from the encoding section. Finally, a quantized index output circuit will generate an index of a quantizer from the generated index and will output the index.

Description

APPARATUS FOR ENcoDI~7G/TR~ysMITTING AN I~IAG~

.
This is a division of copending Canadian Patent Application Serial No. 537,929 which was filed on May 25, 1987.
~LZ9~

BACKGROUND OF THE INVEMTION

Field of the Invention _ .
The present invention relates to a technology for transmitting image signals obtained by encoding image information which utilizes the so-called vector quantization technique and which is applicable to the fields such as the television (TV) conference system and the TV telephone system.

:
Description of the Prior Art As a result of the remarkable advance of the image processing technology in recent years, there have been .
made various attempts to put, for example, the TV
conference system and the TV telephone system to the practical use by mutually and bidirectionally -~ 15 transmitting the image information. In such a technolo-; gical ~lel--, the quant~a.icn technique has been used in ~; which the image signals as the analog quantity are classified into a finite number of levels changing in a discrete fashion~within a fi~ed wi~th and a unique value ; 20~ is assigned to each oL t;nese levels. Particularly, there has been a considerable advance in the vector quantization technique in which~a plurality of samples of the image signals are grouped in blocks and each block thereof is mapped onto a pattern most similar thereto in a ~ 1 ~t : ~:

-multidimensional signal space; thereby accomplishing ~ne quantization.
The study of the vector quantization technolo~ has been described in the following reference materials, for example.
(1) "An Algorithm for Vector Quantizer Design" by Y. Linde, A. Buzo, and R. M. Gray ( IEEE TRANSACTION ON
COMMUNICATIONS, Vol. COM.28, No. l, Januar~ 1980, pp.
84 - 95)

(2) "On the Structure of Vector Quantizers" by A. Gersho (IEEE TRANSACTION ON INFORMATION THEORY, Vol. IT28, No. 2, March 1982, pp. 157 - 166)

(3) "Speech Coding Based Upon Vector Quantization" by A. Buzo, A. H. Gray Jr., R.~M. Gray and J. D. Markel (IEEE TRANSACTION ON ACOUSTICS, SPEECH, AND SIGNAL
PROCES5ING, Vol. ASSP28, No. 5, October 1980, pp. 562 -574) Moreover, the following U.S. Patents have been obtained by the assignee of the present invention.

20 (4) U.S.P.N. 4,558,350 " VECTOR QUANTIZER", Murakami (5) U.S.P.N. 4,560,977 "VECTOR QUANTIZER", ~.urakami et al.
:: ~
An image encoding apparatuses to whlch the vector quantization technology described in the reference : ~
25 mater als a~ove is applied include the following device.

; (6) ~U.S. patent No. 4,769,826 "VIDEO ENCODING APPARATUSn, Kubo et al, issued Septem~er 6, 1988.

-- ~292i~57 A BRIEF DESCRIPTION OF THE DRA~JINGS
-To enable the prior art to be described with the aid of diagrams, the figures of the drawings will first be listed.
The present invention taken in conjunction with the invention disclosed in copending Canadian Patent Application Serial ~o.
537,929 which was filed on May 25, 1987 will be described hereinbelow with the aid of the accompanying drawings, in ~,thich:
FIG. 1 is a block diagram showing the configuration on the transmission side of the prior art image encoding/transmitting apparatus utilizing the DPCM system;
FIG. 2A is a block diagram illustrating a detailed configuration of the movement detecting circuit of FIG. 1, FIG. 2B
is a block diagram illustrating a detailed configuration of the variable-length encode circuit 4, FIG. 2C is a schematic diagram lllustrating an example of the multiplexing of the circuit of FIG.
1 ;
~ FIG. 3A is a block configuration diagram depicting the reception side of the prior art image encoding/transmitting apparatus, FIG. 3B is a block diagram showing the details of the variable-length decode circuit of FIG. 3A;
FIG. 4A is a block configuration diagram showing the configuration of the apparatus on the transmission side of the first embodiment of the image encoding/transmitting apparatus using the DPCM system according to the present invention, FIG. 4B is a block configuration diagram showing in detail the variable-length encode ~clrcuit 4, FIG. 4C is a schematic diagram illustrating an example ;~ 3 of the encoding format of FIG. 4A;
FIG. SA is a schematic bloc~ diagram illustrating the details of the threshold compensation circuit Oc FIG. 4A, FIG. 5B is a block diagram.illustrating in detail the polarity judge circuit l9 of FIG. 5A, FIG. 5C
is a block diagram showing in detail the polarity adder circuit of FIG. 5A;
FIG. 6A is a block configuration diagram showing the construction of the apparatus on the recept~on side according to the first embodiment of FIG. 4~, FIG. 6B is a block diagram showing the configuration of the variable-length decoder as an apparatus on the reception side;
FIG. ?A is a schematic ~lock diagram depicting the configuration of the threshold compensation regeneration circuit of FIG. 6A, FIGS. 7B - 7C are a table of threshold values with polarity and the characteristic diagram, respectively for explaining the operation of the threshold compensation regenerator circuit;
FIG. 8A is a block configuration circuit illustrating ~ne transmission c~rcuit section of a second embodiment of the image encoding/transmitting apparatus using the DPCM system according to the present invention, FIG. 8B
is a block circuit showing in detail the threshold value 25~ generating circuit 3 of FIG. 8A, FIG. 8C is a characteristic graph depicting a conversion example;
FIG. 9A is a block configuration diagram showing the details of the adaptation quantization circuit of ~ - 4 -:

--- lZ9Z057 FIG. 8A, FIG. 9B is a bloc~ configuration diagram - illustrating the details of the adaptation local decoding circuit of FIG. 8A, FIG. 9C is a block circuit diagram showing in detail the characteristic selecting circuit 28 of FIG. 9A, FIG. 9D is a block circuit diagram illustrating in detail the characteristic selecting circuit 29 of FIG. 9B;
FIG. 10 is a block configuration diagram illustrating the circuit section on ~he reception side in the transmission apparatus c. the second embodiment of FIG. 8;
FIGS. ll(a) - ll(b) are characteristic diagrams showing examples of the adaptive quantization and encoding characteristics according to the present 1~ invention;
FIG. 12 is a bloc~ configuration diagram depicting the encoding section in each stage of the general vector quantizer as a basis of a third embodiment of the : present invention;
FIG. 13 is a schematic block diagram depicting the overall constitution of the vector quantizer of FIG. 12;
FIG. 14A is a schematic block diagram depicting the :~ ~ overall consti ution of the vector quantizer in the :~ image encoding/transmitting apparatus of the third embodiment associated with the vector quantizer of FIGS. 12 - 13 according to the present invention, FIG. 14B is a block diagram illustrating in detail the history circuit 314 of FIG. 14A;

: - 5 -.

lZ92057 FIG. 15 is an explanatory diagram showing blocks of the vector quantizer of FIG. 14A;
FIGS. 16(a) - 16(g) are explanatory diagrams ilLustrating the indices in the third embodiment;
FIGS. 17A(a) - 17A(d) are explai~atory diagrams showing a state in which pixels are thinned out from the results of the subsampling on the image signals, FIGS.
17B(a) - 17B(e) are explanatory diagrams showing a method for reading an r.pu~ vector from each buffer, FIG. 17C is an explanatory diagram for explaining the details of the additional bufer;
: FIG. 18 is a timing chart illustrating output waveforms of the respective sections of the history circuit;

- FIG. 19 is a block configuration diagram depicting a general example as a basis of a fourth embodiment of the image encoding/transmitting apparatus according to the present invention;
FIG. 20A is a block constxuction diagram depicting the overall constitution of the image encoding/
transmitting apparatus of the fourth embodiment according to the present invention, FIG. 20B is a detailed ~loc.
dLagram showing the threshol2 value control circuit, FIG. 20C is a schematic diagram illustrating positions of pixels in a reference block, FIG. 20D is a table depLcting an example of the map data of the reference block;
- ~ _ ~ ~32057 - FIG. 21 is an explanatory diagram conceptuall~
illustratinq the disadvantage of the general transmitting apparatus of FIG. 19;
FIG. 22 is a block construction diagram showing the image encoding/transmitting apparatus as a general example and as a basis of a fifth em~odiment of the present invention;
FIG. 23 is a block diagram of the transmitting apparatus o the fifth embodiment according to the present invention;
: FIG. 24 is an explanatory diagram depicting an example of extraction of datum undergone the pseudo encoding;

FIG. 25(a) is a graph showing the encoding lS processing time associated with the image encoding/
transmitting apparatus in the general example of FIG. 22;
FIG. 25(b) is a graph illustrating the processinq time of the fifth em~odiment; and FIG. 26 is an explanatory diagram showing the time-lapse control.

: ~ ~

::

12~32~57 Referring now to FIGS. 1 - 3B, the prior art technolog~ of the present invention ~will ~e described.
The conventional image encoding/transmitting apparatus, as shown in FIG. 1, includes a su~tractor 1 for ootaining a difference bet~een an input signal Sl such as an imag~
signal and an estimation signal Sg and for outputting an estimated error signal S2, a movement detecting circuit 2 for comparing a threshold value T with the estimated error signal S2 to detect a movement or a change and for generating and outputting a movement or change detect signal 53 and a differential signal S4, a quantization circuit 3 for quantizing the movement or change detect signal S3 and the differential signal S4 to output a ~ quantization signal S;, a variable length encoder 4 for ; ~ 15 generating from the quantization signal 55 an encoded signal S6 with a variable length and for outputting the encoded signal S6, a transmission data buffer circuit 5 : for temporarily storing the ~ncoded signal S6 and for outputting t~e encoded signal S6 to the transmission side, a local decoding circuit 6 for generating a reproduced differential signal S7 from the quantization signal S5 del Ye_ec from the quantization circuit 3 and out~utting t:ne reproduced or regenerated differential signal S7, an adder 7 .o. ach_eving an addition on the reproduced : 25 differential signal S7 and the estimation signal Sg and : : for outputting a reproduced input signal S8, an estimating : circuit 8 for outputting an estimation signal Sg based on ~:

, ., . ~ , , .

S~

the reproduced input signal S8, and a threshold generating circuit 9 for monitoring the amount of the encoded signal 56 accumulated in the transmission data buffer circuit 5 and for generating an appropriate threshold ~alue T.
The movement detecting circuit 2 comprises, as shown in FIG. 2, an absolute value circuit 10 for calculating the absolute value ¦S2¦ of the estimated error signal S2, a comparing circuit 11 for effecting a comparison between the absolute value ¦S2¦ of the estimated error signal 52 and the threshold value T and for outputting the movement : or change detect signal S3, and a zero allocator 12 for allotting 0 and outputting 0 as the differential signal -S4 when the movement or change is not detected as a : 15 result of the comparison in the comparing circuit 11.
~ The movement detect signal S3 is converted into a running : ~ ~ record R by use of the running léngth encode table 4a to generate serial data. In addition, only when the movement ; detect signal S3 lS indicating the validness, the ; 20 ~uantization signal 55 is converted into a variable-length record through the variable-length encode table 4b to : senerate serial data (FIGS. 23 ~ 2C). Reference numeral 4c lDdicates a multLple~ operation control section.
In contrast to the configuration on the transmission side of FIGS. 1 - 2B, the configuration on the reception side is shown in FIGS. 3a - 3B. In FIG. 3A, the equipment on the reception side includes a receivinq data ~ ~92~)S7 buffer circuit 13 for receiving and for temporaril~
storing the encoded signal S6 delivered from ~he transmission data buffer circuit 5 on the transmission side, a variable length decoder 14 for decoding the S encoded signal S6 stored in the receiving data bu~fer 13 to output a reproduced quantization signal SlL, a local decoding circuit 15 for outputting a reproduced differential signal S12 based on the reproduced qu ntization signal Sll, an adder circuit 16 for obtaining the di~ference between the reproduced differential signal.
S12 and the reproduced estimation signaL S13 and for reproducing the input signal S14 which corresponds ~o the reproduced input signal S8 on the transmission side, and an estimating circuit 17 for outputting the reproduced estimation signal S13.
Aftex the encoded signal S6 undergone the multiplexing in the variable-length encode circuit 4 i5 received by the receive buffer circuit 13, the data is dlst-_buted to the res ective decode tables of var~a'ole codes under control of the multiplex separation control circuit 14a. As a result of the decoding, the movement cetect sisnal anc the quantlzation signal are attained.
oreover, when the decoded movement detect signal indicates the invalidness (= "0"), the quantization . 25 signal is reset to "~" by the flip-flop 14e, thereby outputting the output Sll (FIG. 3B~.

, ~

1~2~57 Next, the operation on the transmission side willbe described with reference to FIGS. l - 2.
Assuming first the non-e~fective error in the movement detecting circuit 2 to be d, the estimation coefficient to be applied to the reproduced input signal S~ in the estimating circuit 8 to be A, and the delay of the time t to be z t, the following relationships are : satisfied.

: S4 = S2 + d S7 = S4 + Q
S8 = S7 + Sg = Sl + Q
: ~ ~ Sg = A- S8, z t ~he subtractor l calculates the e~stimated error signal S2 representing the difference between the input signal S
and the estimated signal Sg, whereas the movement detectinq circuit 2 outputs the movement or change :
: detection signal S3 and the differential signal S4 based .
on ~he es lmated er~o. slgnal S2 calculated bv the subtractor l.
A detai1ed~descriptLon will be given of the operation of the mo~ement detectins ci_c~i;t 2 by ref~rring to FIG. 2. The allotting abso1ute vaLue circuit lO obtains : the absolute value of the est1mated error signal S2 and : 25 then the comparison circuit ll achieves a comparison between the absolute:value ¦S2¦ of the estimated error .
signal S2 and the threshold value T generated by the threshold value generating circuit 9.

:~ ' The movement detection signal S3 is output in conformity with the following conditions.
S3 = 0 (invalid) 1S21 T
S3 = l tvalid) 1S21 ~ T
When the movement or change is not detected, namel~, for "S3 = 0", zero allocator 12 outputs "0" for the differential signal S4.
On the other hand, the quantization circuit 3 converts the inputted differential signal S4 according to an arbitraxy characteristic. The variable encoding ~circuit 4 receives the quantization signal S5 only when : the movement detection signal S3 is valid, namely, for . 1153 - l" and, for example, conducts a run-length encoding : on the movement detection signal S3. For the quantization : 15 signal S5, a code having a smaller code length is assigned .
to a ~alue in the neighborhood of "0" for which the generation frequency is high and then the code is stored : : in the transmission data buffer circuit 5. The transmission data buffer circuit 5 outputs the : :20 accumulated datum as the encoded signal S6 to a :~ : transmission line. The threshold generating circuit 9 ; mon-tors the accumulated amount:of the transmission data bu~^fer circu-t S and further controls the generation : amount of the encoded data by generating an a~propriate ~ threshold value.
Next, the operation on the reception side will be ~: ~ : described with reference to FIG. 3. The receiving data , . ., . ~ , .

buffer circuit 13 first receives the encoded signal 56 undergone the varlable length encoding on the transmission signal and outputs the signal S6 to tne variable length decoder 14. Only when the movement detection signal S3 undergone the decoding operation indicates the validness, the variable length decoder 14 outputs the reproduced quantization signal Sll. If the movement detection signal 53 indicates the invalidness, the varlable length decoder 14 outputs ~-on. Next, the local decoding circuit lS decodes the reproduced quantization signal Sll and outputs the reproduced differential signal S12 to the adder l6. The adder 16 adds the reproduced differential signal S12 to the reproduced estimation signal S13 from the estimation circuit 17 thereby to reproduce the input signal S14.
The operation to effect the data compression and transmission hy use of the dif~erential signal is referred to as the differential pulse code modulation (to be abbrevia~ed as DPC.~ herebelow) system.
: ~20 However, in the image encoding/transmittinq apparatus : using the DPCM systam, the variable length encoding is :achieved cn the datu~ which is judged to be effective at the step of the~variab~le length encoding; consequently, as the threshold value increases, the code having a short ~ ;25 :code length to be assigned in the neighborhood of "0"
:~: : cannot be generated and hence the efficiency of the encoding is deteriorated; moreover, there has been a ~%~7 problem that as the threshold value becomes greater, the precision of the quantization is not improved for the quantization characteristic of the quantization circult in the circuitry on the transmission side even when the dynamic range of the effective datum is narrowed.
SUMMARY OF THE INVENTION
In accordance with one aspect of the invention there is provided an image encoding/transmitting apparatus including a vector quantizer wherein said vector quantizer comprises: a quantization preprocessing circuit for subsampling an image signal, for thinning out pixels from the image signal according to a pixel ratio, and for generating from the pixel signal an n-dimensional vector; a group of input buffers for respectively storing each said input vector in a sequence of a :15 quantization processing and for repetitiously reading the input vector at an interval of the processing period; a vector output circuit for beforehand preparing an output vector corresponding to a probability distribution of the input vector and for selecting and outputting the output vector ~ased on an index input; a distortion computation circuit for computing a distance between the input vector and the output vector se1ectsd and output; a plurality of encoding sections disposed corresponding to the huffers; index output generating means for generating an index for a period of a subsequent 25~ prooessing based on the calculated distance value outputted ~: ~ from said each encoding section and the input vector and for send1ng the index to each said vector output circuit; and a: guantized index output circuit for generating an index of a : ~ ~ : quantizer from said generated index and for outputting the index.

: 14 2~5~

A description will be now gi~Jen of the e~cdiments suitable for ~he image encodins/transmitting apparatus according to the present invention.
In FIG. 4~, reference numeral 18 indicates a threshold compensation circuit for assigning a polarity to the threshold value T output from the threshol~
generating circuit 9 according to the di~ferential signaL S4 supplied f~om the movement detecting circuit 2, or generatins the compensated differential signal 515 by effecting a subtraction bet~een the threshold value T to which the polarity is assiqned and the differential signal 54, and~for outputting the compensated di'ferential signal Sl5~
As shown in FIG. 4B, t~.e movement detect signal S3 lS is converted into a run length code R by use of the run length encode t~ble 4a to generate the serial data. On when the movement detect sisnal S3 is indicating the validr.ess, the threshold Yalue T and the quantization signal 55 are respectively converted into variable-length : : : ~ :

~ , J ~2~)57 records by use of the variable-length encode tables 4b - 4c; thereafter, the records are converted into serial data. The encode tables each are constituted from a storage device such as an ROM. Reference numer~l 4d indicates a multiplexing control section. In this case, the multiplexing is accomplished as shown in FIG. 4C.
In FIG. 5A, reference numeral 19 denotes a polarit~
judgment circuit for judging the polarity of the difference signal S4 and for outputting the polarity signal S16, reference numeral Z0 is a polarity adding circuit for assigning the polarity to the threshold value T output from the threshold generating circuit 9 according to the polarity signal S16, and reference numeral 21 is a subtractor for achLeving a subtraction between the threshold value Tl with the polarity and the differential signal S4 and for outputting the compensated differential signal S15. The polarity judgment circuit 19 includes a comparator l9a as shown in FIG. 5B.
Moreover, the subtractor 20 comprises a flip-flop 20a of FIG. SC. In FIG. 5B, the following condition is satisfied.
B ( 0 ) then ~ S16 = "l"
else Sl6 = '~o~l ~ In FIG. 5C, the polarity signal Sl6 is used as the sign bit of the threshold value T so as to be fetched by the FF. This FF is reset to "0" when the movement detect signal S3 is indicating the invalidness (= "0").

1~32057 In FIG. 6A, reference numeral 22 denotes athreshold compensation regeneration circuit for generating the regenerated differential signal S21 from the detection signal S17 output from the variable length decoding circuit 14, the threshold value T, and the regenerated, compensated difference signal S20 output from the local decoding circuit 15 and for outputting the regenerated differential signal S21. The threshold value regeneration circuit 14, as shown in FIG. 6B, receives by means of the receive buffer circuit 13 the encoded signal S6 undergone the multiplexing in the varlable-length encode circuit 4 on the transmission side, distributes the data to the respective decode tables of the variable-length code under control of the lS multiplexing separation control circuit, and obtains as the results of the variable-length decoding operation the movement detect signal S17, the threshold value T, and : the quantization signal Slg. When the movement detect signal indicates the invalidness, the threshold value T
and the quantization signal S19 are not delivered.
Incidentally, the decode tables each are constituted from ~: a memory device such as an ROM.
: In addition, in FIG. 7A, reference numeral 23 is a polarity judgment circuit for judging the polarity of the reqenerated, compensated differential signal S20 and for outputting the polarity signal S23, reference numeral 24 designates a polarity adding circuit for assigning the ~20S7 polarity to t~e threshold value T according to the - polarity signal S23 and for outputting the threshold value Tl with the polarity, and reference numeral 25 indicates an adder for outputting the regenerated differential signal S21 by adding the threshold value T
with the polarity and the regenerated, compensated differential signal S20.
Next, the operation on the transmission side of the present invention will be described with reference to FIGS. 4A - 5C.
First, assuming the non-effective error in the movement detecting circuit 2 to be d, the estimation coefficient to be applied:to the regenerated input signal S8 in the estimating circuit 8 to be A, and the delay of the time t to be Z ~, the following relationships are satisfied.
S2, Sl S9 : S4 = S2 + _ S15 = S4 - T
: 20 S7 = S7 + ~
S6 = Sg + S7 + T Sl Q
Sg = A- S8- z t The subtractor 1 here calculates the estimated error slgnal S2 representing the difference between the input signal 51 and the estimated signal Sg, whereas the movement detecting circuit 2 outputs the movement ; detection signal S3 and the differential signal S4 based : - 18 -1~

on the estimated error signal S2 calculated by the subtractor l.
The operation of the movement detecting circuit 18 in this case is the same as that of the conventional system.
On the other hand, the threshold compensation circuit 18 outputs the threshold value Tl with the polarity and the compensated differential signal S15 based on the movement detection signal 53 and the differential signal S4.
The operation of the threshold compensation circui' 18 in this case will be described in details with reference to FIG. 5.
In FIG. 5A, the polarity judgment circuit l9 judges :~ 15 the polarity of the differential signal S4 and outputs .
: the polarity signal S16 to the~polarity adding circuit ~: 200 On receiving the polarity signal Sl6, the polarity ~ adding circuit 20 assigns the polarity to the threshold .

: ~ : value ~ f_om the threshold generating circuit 9 and : : 20 outputs the resultant signal as the threshold value T

wlth the polarity. However, in the case where the : movement detection signal S3 indicates the invalidness, :the threshold value Tl with the polarity is set to 0.

In this case, there exists also a method in which the O

lS judged in the polarity judgment circuit 19 and the :
~ : : poIarity adding circuit 20 sets the threshold value Sl ::
~ with the polarity to O according to the judgment. By : :: ~ : - 1 9 : :

- ~2~32~S7 subtracting the threshold value Sl ~,Jith the polarity from the differential signal S4 in the subtractor 21, the threshold value is compensated and hence the compensated differential signal S15 is outputted. The compensated differential signal S15 is converted into the quantization signal S5 in the quantization circuit 3 and is delivered to the variable length encoding circuit

4 and the local decoding circuit 6. When the movement detection signal S3 is valid, the variable encoding circuit 4 receives the quantization signal S5 and the : threshold value T, and for example, effects a run length : encoding on the movement detection signal 53. For the ~: quantization signal 55, the variable length encoding : circuit 4 assigns a code with a short code length to a value in the vicinity of 0 for which the generation frequency is high and accumulates the code in the transmission data buffer circuit 5; thereby outputting : ~ the signal as the encoded signal S6 to the transmission line. Moreover, in the threshold generating circuit 9, the accumulated amount of data in the transmission data buffer 5 is monitored so as to generate an appropriate threshold value T~to control the generation amount of the encoded datum. On the other hand, the local decoding circuit 6 decodes the quantization signal S5 into the regenerated, compensated differential signal S7, which is : then fed to the adder 7. The adder 7 adds the threshold value:T, the estimated slgnal 58~ and the regenerated, .

2~5~

compensated differential signal S7 to obtain the regenerated input signal S8. The estimating circuit delays the regenerated input signal S8 by a period of time t beforehand set, multiplies the signal by the coefficient A, and outputs the resultant signal as the estimated signal Sg. If the time t is set ta the period of time of a frame when the input signal Sl is an image signal, a frame-to-frame DPCM transmitting apparatus is implemented; whéreas, if the time t is set to the period of time of a field, a field-to-field DPCM transmitting apparatus is materialized.
In addition, in the embodiment above, the same effect can be attained by reducing the threshold value for the transmission by means of a control in which the : ~ ~15 generation and update of the threshold value are conducted at an interval of an arbitrary period of time Ta -: ; Next, the operation on the reception side will be - described with reference to FIGS. 6A - 6B. The encoded signal S6 undergone the variable length encoding on the : ~ transmission side is received by the receiving data buffer circuit 13.~ Only when~the decoded movement :~ detection signal S17 is valid, the variable length decoding circuit 14 outputs the threshold value T and : 25 the regenerated quantization signal Slg. When the signal , : ~ S17 is invalid, 0 is outputted. The local decoding circuit 15 decodes the regenerated quantization signal Slg into the regenerated, compensated differential signal; furthermore, the regenerated differential signal S21 is regenerated in the threshold compensation regeneration circuit 22. The operation will be described with reference to FIG. 7A. The polarity of the regenerated, compensated differential signal S20 is judged by the polarity judgment circuit 23, the polarity is assigned to the threshold value T depending on the polarity signal S23 representing the positïve or negative polarity, and the resultant signal is delivered as the threshold value Tl with the polarity. However, when the movement detection signal S17 lndicates the invalidness, the t~hreshold value Tl with the polarity is set to Q.
Moreover, the adder 25 adds the regenerated, compensated differential signal S20 to the threshold value Tl with the polarity, thereby attaining the regenerated dIfferential signal S21. The adder 16 adds the estimated signal S22 from the estimating circuit 17 to the ; regenerated differential signal S21 so as to regenerate ~20 ~ the objective input signal Sl on the transmission side.
The operation of the threshold compensation circuit can be represented by numeric values as shown in FIG 7B
This example shows a case of T = 5. FIG. 7C is an inputjoutput characterlstic graph corresponding thereto.

~ , The broken lines indic~ate the conventional example, whereas the solid lines represent the characteristic of the present invention.

~ .
, :
~ : . ' l~Z057 As described above, even if the invalid data period (-T-O-T) occurs due to the movement detect circuit, the invalid data period can be canceled by effecting the threshold value compensation according to the present invention, which enables to increase the quantization precision as compared with the conventional apparatus As described above, according to the embodiment of the present invention, the polarity of the differential signal is judged, a polarity is assigned to the threshold value depending on the judgment, the compensation of the differential signal is achieved according to the threshold value to which the polarity is assigned, and the compensated differential signal undergone the quantization is used for the transmission and reception;
consequently, the generation of the code with a short code length allocated in the neighborhood of O is suppressed and hence the efficiency of the communication : is improved.
Next, the second embodiment of the present invention : 20 will be described.
The datum compressing/transmitting apparatus utilizing the differential pulse modulation system ~: : according to the second embodiment includes a transmis-; sion circuit section for effecting the data compression ~ 25 and transmission on the digital input signal by use of : the differential pulse modulation and a reception circuit section associated with the transmission circuit section.
::

1~320S7 FIG. 8A shows the block construction diagram of the transmission circuit section of the embodiment.
In the second embodiment, the quantization cirsuit comprises an adaptive quantization circuit 3 having a plurality of quantization characteristics. The adaptive quantization circuit 3 selects a characteristic depending on the threshold value T, conducts an adaptive quantization on the differential signal S4, and outputs the adapted quantization signal S5.
The adaptive local decoding circuit 6 is disposed corresponding-to the adaptive quantization circuit 6, decodes the adapted quantization signal S5, and outputs the regenerated differential signal S7. Incidentally, the same reference numerals are assigned to the same components as those of the prior art example and the description thereof will be omitted.
FIG. 9A is a block diagram for explaining the adaptive quantization circuit 3 in details. The character selecting circuit 28 selects a quantization character-:
istic depending on the threshold value T and thenquantizes the differential signal S4.
FIG. 9B is a block diagram for explaining the adaptation local decoding circuit 6 in details. The character selecting circuit Z9 selects a decoding characteristic depending on the threshold value T and then decodes the quantlzatlon signal S5; furthermore, the~obtained signal passes the zero allocation circuit 25 so as to be subjected to the zero allocation depending on the movement detection signal S3.
Next, the flow of the signal will be described. As shown in FIG. 8A, assuming the digitalized input signal such as an image signal to be Sl, the estimated signal calculated by the estimating circuit 8 (to be described later? to be Sg, the estimation error signal to be 52' the differential signal to be S4, the regenerated differential signal to be S7, and the regenerated input signal to be S8, these exists the following relationships among the respective signal values.

S2 = Sl - Sg S4 = S2 + _ S7 S4 + Q
S8 = Sg + S7 Sl + Q
Sg = A. S8- z t where, d is the non-effective error in the movement detection circuit, Q stands for the quantization error, A indicates the estimation coeff1cient, and z t denotes zo a delay of the time t.
: The estimating circuit 8 delays the regenerated nput signal S8 by a period of time t beforehand set, multiplies the obtained signal by the coefficient A, and outputs the estimation signal Sg.
: 25 Like FIG. 2, in the movement detecting circuit 2, assuming the absolute value of the estimated error signal : S2 calculated by the absolute value circuit lO to be ¦S2¦

and the threshold value to be T, the value of the movement detection signal V calculated by the comparing circuit 11 is determined as follows.
If ¦S2¦ ~ T, then V = 0 (invalid) else V = 1 (valid).
Furthermore, if the predetermined amount of movement is not detected (within the invalid data range), the zero allocation circuit 12 allocates "0" to the estimated error signal S2 and hence the differential signal S4 is outputted as "0" to the adaptive quantization circuit 3.
As shown in FIG. 9A, the differential signal S4 inputted to the adaptive quantization circuit 3 is converted into the quantization signal S5 according to the quantization characteristic specified by the character selecting circuit 29 which changes over the characteristic depending on the threshold value T and the quantization signal S5 is then outputted to the variable length encoding circuit 4.
A description wilI be here given of the quantization characteristic example in reference to FIG. ll(a).
The dot-and-dash line indicates a characteristic :~ example in the case of T = 0, whereas the solid line denotes a characteristic example in the case of T = 3.
: As shown in this diagram, when the datum of the differential signal S4 is in the invalid data range T < data < +T), the quantization is not accomplished, namely, the quantization is effected only for the datum ~z~

in the valid data range (data < -T or fT < data). As can be seen from FIG. 9A, the quantization characteristic selected by the character selecting circuit 28 is such that the quantization output on the differential signal within the invalid data range (within the specified range of the corresponding threshold value) becomes to be "0", which improves the quantization accuracy, The characteristic selecting circuit 28, as shown in detail in FIG. 9C, comprises a fiip-flop 28a (FF? to ~ which the decoded quantization signal selected by the characteristic selecting circuit 28 is delivered as an input signal and is reset to "0" when the movement detect slgnal VD indicates the invalidness (V = 0).
: ~ The variable length encoding circuit 4 receives lS only the quantization signal S5 for which the movement : detection signal S3 is valid (V = 1) and effects the run length encoding on the movement detection signal S3.
For the quantization signal S5, a code with a short code : length is assigned to the value in the neighborhood of 0 for which the generation frequency is high, and the resultant code is transmitted to the transmission data ~: :
: ~: buffer circuit 5. The data accumulated in the transmission data buffer circuit 5 is sent as the en~oded signal S6 to the transmission path.
;25 In addition, as shown in FIG. 8A, the quantization :
: signal S5 is also delivered to the adaptation :quantization circuit 6.

~ 27 -:: :

Moreover, as shown in FIG. 8B, the threshold value generator circuit 9 includes a conversion table 9a used to convert an input data into a threshold value based on the accumulated amount of the data. FIG. 8C shows the relationships between the accumulated data amount and the threshold value.
On the other hand, as shown in FIG. 9B, the adaptive decoding circuit 6 selects a decoding characteristic by means of the character selecting circuit 29 corresponding to the quantization characteristic selected by the adaptation quantization circuit 3. As shown here in FIG. 9D, the characteristic selecting circuit comprises a memory device such as an RO~I and the selection signal thereof is determining by use of the mapping in which the lS threshold value T is used as an address input. Moreover, the allocation zero circuit 25 allocates the invalid datum "0", which is then delivered as the regenerated differential signal S7. FIG. llB shows a decoding characteristic example. In the adder 7, the regenerated differential signal S7 is added to the estimation signal Sg to obtain the regenerated input signal S8 to the supplied to the estimating circuit 8.
The threshold generating circuit 9 monitors the data accumulation amount of the transmission data buffer circuit 5, generates an appropriate threshold value according to the data accumulation amount, and thereby achieves a smoothing operation on the data code amount.

Z~57 FIG. 10 is the block construction diagram illustrating the reception circuit section of the data compressing/transmitting apparatus utilizing the differential pulse modulation system according to the present invention. The reception data buffer circuit 13 temporarily stores the encoded signal S6, the varia~le length decoding circuit 14 decodes the encoded signal S6, the adaptive local decoding circuit 15 outputs the regenerated differential signal S12, and the estimating circuit 17 estimates the regenerated signal Sl.
Here, the detailed constitution of the adaptive local decoder circuit 15 is the same as that of the circuit 6 of FIG. 8A. In addition, the quantization characteristic and the decoding characteristic associated with the memory device such as an ROM are the same as those shown in FIGS. 9A - 9D.
The regenerated signal Sl is calculated by the adder - 16 from the estimated signal S13 and the regenerated differential signal S12.
Next, the flow of the signal in the reception circuit section will be described. The encoded signal S6 subjected to the variable length encoding in the transmission circuit section is received by the receiving data buffer circuit 13 and is then transmitted to the variable length decoding circuit 14.
Only when the movement detection signal Sllc decoded in the variable length decoding circuit 14 indicates the 1~32057 valid data range (V = 1), the quantization signal Slla and the threshold value Sllb are fed to the local decoding circuit 15. When the decoded movement detec~ion signal 53 indicates the invalid data range (V = 0), "0"
is delivered to the local decoding circuit 15.
Moreover, like the adaptive local decoding circuit 6 of the transmission circuit section, the adaptive local decoding circuit 15 selects a decoding characteristic depending on the threshold value Sllb.
:~ :10 The quantization signal Slla is then decoded into the : regenerated differential signal Sl~ and is added by : means of the adder to the estimated signal S13 from the ; ~ estimating circuit 17, thereby the encoded signal S6 : from the transmission circuit section is regenerated as 15: a signal S14.
~: In this embodiment, the same effect can be attained : by reducing the threshold value for the transmission .
: through an operation to generate (update) the threshold `
value at an interval of an arbitrar~ period of time T

20 (e.g. a frame period).
:: :
Prior to the description of the third embodiment of : the present invention, description will be glven of the , technology adopted as the basis of the third embodiment and of the outline thereof.
Z5 The third embodiment relates to a vector quantizer : for encoding image signals with a high efficiency, and n particular, to a miniturized vector quantizer.

: ~:

~æ~s~
FIG. 12 shows the technology as the basis of the third embodiment. This diagram illustrates a block diagram of the n-th stage of the vector quantizer. As shown in FIG. 13, the vector quantizer includes a multi-stage connection of a plurality of stages.
In FIG. 12, reference numeral 31 is a ~ector output circuit, reference numeral 32 indicates an input index, reference numerals 33a - 33b are distortion operating circuits, reference numeral 34 designates an input vector, reference numeral 35 indicates a comparator, ; reference numerals 36a - 36b denote distortions, reference numeral 37 is an index output circuit, reference numeral 38 indicates a delay circuit, reference ; ~ numeral 39 stands for an output vector, reference numeral 310 denotes an output index, and reference numeral 311 designates a quantizer output index. The encoding ~ ~ section 320 comprises tAe respective circuits above.
; ~ Incidentally, reference numeral 3Zl in this diagram indicates the encoder at the final stage.
Next, the operation of this configuration will be described. When the i~nput index 32 is received, the vector output circuits 31a - 31b select output vectors from a group of output vectors beforehand prepared (code . book) corresponding to the input index and then deliver 25~ the output vectors. The respective output vectors are supplied to the distortion computation circuits 33a -~ ~ 33b, which calculate the distortions (distances) with ; - 31 -, respect to the input vector sent from the input buffer 312. The distortions 36a - 36b outputted as a result of the computation from the respective distortion computation circuits 33a - 33b are fed to the comparator 35, which effects a comparison therebetween. For the distortion 36a < the distortion 36b, "0" is output;
whereas for the distortion 36a > the distortion 36b, "1l' is output. Next, the distortion compare value is output together with the index 32 from the index output circuit 37 to the encoding section 320 of the next stage so as to be used to generate a new input index. The input vector 34 to ~e subjected to the distortion computation at the next stage is sent via the delay circuit 38 to the encoding section 320 of the subsequent stage, so that lS the distortion is computated with respect to the output vector selected and output by use of the new input index.
The distortion computation is continuously repeated up to the encoding section 321 of the final stage so as to minlmize the total of the distortion through a conversion, thereby generating the final quantizer output index 311.
In FIG. 13, the input vector and the input index to each encoding section 320 are respectively the output vector 39 and the output indeY 310 from an encoding section 320 of the previous stage. For the encoding section 321 at the final stage (the 10-th stage in this case), since there does not exist the encoding section 320 of the succeeding stage as compared with the encoding ;7 sections 320 of from the first to the ninth stages, the delay circuit 38 is not required. The output index from the final stage is delivered as the quantizer output index 311.
As shown in FIGS. 12 - 13, the prior art vector quantizer is configured such that the input vector (input datum) and the input index for selecting an output vector are sequentially transmitted to the next encoding section : ~ 320 while updating the input index in each encoding section 320 so as to generate the quantizer output index in the final encoding section 321, which consequently leads to problems that the size of the circuitry is increased and that the idle time or the wait time occurs in the encoding sections and hence the overall circuitry cannot be effectively utilized.
According to the vector quantizer of the embodiment, there is provided a quantization index output circuit in which the input image signal is subsampled in the quantization preprocessing circuit to thin out pixels, :20 the respective pixel signals are used to form an : ~: ;n-dimensional input vector, and each input vector is stored in the input buffer corresponding to each pixel :-~ sig~nal; furthermore, the input vector is supplied to the : :~ :: encoding section corresponding to the input buffer at an : , : :
: 25 interval of the processing period. The encoding section : calculates the distortion between the input vector and the output vector updated at an interval of the processing :: ~

period. Based on the distortion, the input index generating means generates a new input index for selecting the output vector, and based on the input index, the quantizer output index is generated and is S output from a quantization index output circuit.
Next, the third embodiment of the present invention will be described with reference to the drawings.
FIG. 14A is a schematic block diagram depicting the entire configuration of the third embodiment in which reference numeral 316 denotes an input image, reference numer~l 317 is an quantization preprocessing circuit, reference numerals 311a - 31b designate vector output circuits for outputting vectors corresponding to the :
input index 32, reference numeraIs 33a - 33b indicate distortion computation circuits for respectively calculating the distortion between the input vector 34 and the vectors delivered from the vector output circuits 31a - 31b, reference numeral 35 denotes a comparator, reference numeral 36a stands for a distortion outputted ZO from a circuit equlvalent to the distortion computation circuit 33a in the encoding section 313a and the distortion computation circuit 33a in the respective encodlng sections 313b, 313c, and 313d, reference numeral 36b designates a distortion outputted from a circuit equivalent to the distortion computation circuit 33b in the encoding section 313a and the distortion computation circuit 33b in the respective encoding ::

1?,~0.~7 sections 313b, 313c, and 313d, reference numeral 37 indicates an index output circuit, reference numeral 311 represents a quantizer output index of the vector ~uantizer, reference numerals 312a, 312b, 312c, and 312d are input buffers for storing the blocks A, B, .., H
obtained by dividing the input vector 34 as shown in FIG~ 15, reference numerals 313a, 313b, 313c, and 313d denote encoding sections, reference numerals 314a, 314b, 314c, and 314d indicate history circuits for outputting 10 indices for the blocks of A, B, , H, and reference numeral 315 denotes a quantization index output circuit for outputting the index of the vector quantizer circuit associated with the input vector. As shown in FIG. 14B, the history circuits 314a - 314d each comprise a first flip-flop circuit 601 for receiving an output from the output index circuit 37, a second flip-flop circuit 602 for the same purpose, an output control circuit 603, a third flip-flop circuit 604 for receiving the respective outputs from the first and second flip-flop circuits 601 - 602, and a pattern comparing circuit 605 for delivering a history signal to the encoder 313a based on the output from the third flip-flop circuit 604. The patt~rn comparing circuit 605 is constituted, for example, from a code book. FIG. 15 shows a block attained by dividing the image of the third embodiment in which A - H
indicate subbloc~s. The input buffers 312a - 312d respectively store subblocks A - B, C - D, E - F, and G - H. The embodiment has been described with reference to a vector quantizer including ten stages. FIGS.
16(a) - 16tg) show the indices and quantizer output index 311 to be output from the index output circuit 37 at each stage.
Next, the operation of the embodiment will be described. The image 316 supplied to the system is thinned out in the quantization preprocessing circuit 317. (Refer to FIG. 17 for details.) Before the input vector 34 is supplied to the vector quantizer, the blocks A - B, C - D, E - F, and G - H of FIG. 15 are respectively written in the buffers 312a - 312d.
Thereafter, the image data is raad as blocks of A - B, C - D, E - F,;and G - H in parallel from the associated ~15 buffers 312a - 312d, respectively, and the vector quantization is accomplished by executing the read operation ten times for each block. (Refer to FIG. 18 for details.) However, prior to the operation above, the vectors read from the buffers 312a, 312bi 312c, and :
312d are respectivaly supplied to the distortion computation circuits 33a and 33b. The buffers 312a -:
312d are added to effect a concurrent processing by , separating the input serial data into four data groups.
The details of these b~uffers 312a - 312d are shown in FIG. 17C.
urthermore, the vectors output from the vector output circuits 31a - 31b according to the input index , ~

~ 36 -:

32 are also delivered to the distortion computation circuits 33a - 33b. The distortion computation circuit~
33a - 33b calculate the distortions between the vectors input from the input buffers 312a - 312d and the vectors output from the vector output circuits 31a - 31b, respectively. In this operation, to prevent the distortions from being delivered from the encoding sections 313a - 313d to the comparator 35 at the same time, the distortions are output therefrom in the sequence of encoding sections as 313a, 313b, 313c, and 313d. Next, the distortions are input to the comparator 35 and "0" is output from the comparator if the ; dlstortion 36a is less than the distortion 36b;
otherwise, "1" is output. Based on the output from the lS comparator 35 and the index previously supplied, the index circuit 37 outputs a new index~ From the history circuits 314a - 314d, the indices of the vectors of the respective blocks A - B, C - D, E - F, and G - H are output to the vector output circuit for the next distortion computation. In addition, the indices are also output to the index output circuit 37 to attain the next new index. This operation is repetitiously executed from the first stage to the tenth stage, and the index :
;dellvered from the last operation at the tenth stage is - 25 output as the vector quantizer index from the ~ quantization index output circuit 315. The output ::
~ ~waveforms at the respective sections described in , ~ 37 conjunction with FIG. 14B of the history circuits 314a - 314d are as shown in FIG. 18.
As described above, according to the third embodiment, the input image is subsampled, the pixels are subjected to the thinning-out operation, and the n-dimensional input vector to be shaped is minimized.
Thereafter, for the input vectors of the same image, the distortion operation is repetitiously executed by the same distortion operation circuit to converge the total of the distortions, thereb~ minimizing the resultant value of the total. For other vectors, the distortions are concurrently calculated by the respective distortion operation circuits, and based on the distortions, the quantizer index is generated for each pixel signal;
consequently, the size of the circuitry can be miniaturized and the entire circuitry can be efficiently operated without idle operations.
Prior to the description of the fourth embodiment of the present invention, description will be given of the ~20 technology adopted as the basis of the third embodiment.
FIG. 19 shows an image encoding/transmitting apparatus to which the frame-to-frame encoding method uti~lizing the vectorizing m~thod, namely, the image encoding/transmitting method as the basis of the fourth embodiment including the movement compensation is applied.
As shown in this diagram, the image encoding/
, transmitting apparatus primarily includes a preprocessing , section 41, a movement compensation section 42, and a vector quantization section 43.
The preprocessing section 41 generates from the image input signal 400 blocks each containing k pixels existing in the neighborhood of each other in the image to form a k-dimensional vector signal 401 for each block, whereas the frame memory 44 is provided to store the image signal formed in a block in advance in time by a frame of the current image signal.
The movement compensation section 42 includes a reference block generating section 42a for generating as a reference blocks a plurality of blocks each including the current vector signal 40I, a block corresponding to the current vector signal 401, and a block stored in the frame memory corresponding to the same position in the image and for calculating the block position information 402al and the vector signal 402a2 and a distortion computation section 42b for calculating the distortion (for example, the Euclid distortlon, the absolute distortion, etc.) between the current vector signal 401 and the reference vector signal 402a2 and for selecting a bl`ock having the minimum dlstortion from the reference blocks.
: ~
The subtractor 45 achieves a subtraction between the current vector signal 401 and the vector signal 402b2 of the block selected by ~th~ movement compensation section ~; - 39 -'' ';

20~57 42 and sends the differential vector signal 405 to the vector quantizing/encoding section 43.
The vector quantizing/encoding seetion 43 comprises an arith~etic section 43a for computing the average value m and the variance ~ from the differential vector signal 405, a validness/invalidness judgment eircuit 43b for judging the validness or invalidness of the selected block based on the average value m, the variance ~, and : the threshold values Tl and T2 eontrolling the compression amount of the information volume, a :: normalizing section 43c for normalizing the differential veetor signal 405, a code block 43d for storing patterns of a plurality of the normalized image vector signals, and a vector quantizing section 43e for seleeting a : ~ 15 pattern from the code book 43e whieh is the same as or similar to the normalized differential veetor signal 403e : ~ normalized by the normalizing seetion 43e and for - encoding the selected pattern number, the average m, and .
: the varianee a.
~: 20: Next, the flow of the signal will be described.
: First, the image input signal 400 is subjected to the block generation in the:preprocessing section 41 and is thereby converted lnto the veetor signal 401.
Ther after, the referenee blocks are generated in ~ the movement eompensation section. From the reference blocks, a block having the smallest distortion with respect to the vector signal 401 is selected, and then ::

~::

the selection block position information 4~2bl and ~he selection vector signal 402b2 are supplied to the subtractor 45.
In the subtractor 45, a subtraction is then accomplished between the vector signal 401 and the selection vector signal 402b2, and the differential vector signal 405 is delivered to the vector quantizing/encoding section 43.
In the vector quantizing/encoding section 43, the differential vector signal 405 is processed by the arithmetic section 43a to calculate the average value m and the variance a of the differential vector signal 405.
Thereafter, the validness/invalidness judgment circuit . 43b judges the validness or invalidness as represented by the following expressions by use of the threshold value Tl for the average value and the threshold value .
T2 for the variance.
m c Tl and a ~ T2: Invalid m > Tl or a ~ T2 Valid If the judgment results in the invalidness, the current block is assumed to be identical with the selected block and hence only the selection block :position information 402bl is encoded and the resultant signal is temporarily stored in the transmission data buffer 46.
: On the other hand, if the judgment results in the validness, the differential vector signal 405 is ,, , normalized as the datum to be transmitted in the normalizing section 43c according to the follswing formula.

Yi (Xi ~ m)/~

(where, i = 1, 2, .. , k) : i-th element of the normalized vector xi: i-th element of the differential vector Thereafter, the normalized differential vector signal 403c is quantized and encoded in the vector quantizing section 43e as follows.
First, a pattern which is most similar to the normalized differential vector signal 403c is selected from the code book 43d. As the transmission information, the pattern number, the selection position information 402bl, the average value m, and the variance ~ are ; encoded and the resultant signals are temporarily stored in the transmission data buffer 46.
The image encoded signals temporarily stored in the ~transmission data buffer 46 are transmitted in the frame-by-frame fashlon.
On the other hand, the threshold values Tl - T2 are controlled according to the amount of the image encoded slgnals stored in the transmission data buffer 46 associated with the previous frame such that the threshold .
values are set to great values for the great amount of the signals and are set to small values for the small - 4? -amount of the signals, thereby controlling tne degree othe compression for each frame.
According to the image encoding/transmitting apparatus shown in FIG. 19 as described above, as a result of the movement compensation processing step, when the differential data between the input block and the selected block is within the threshold value range in the encoding processing step, the selected block is directly reproduced as an image.
In this operation, if the selected block or the input block has a strong contour line even when the differential datum is within the threshold value range, there has been a problem that the contour line is shifted in the image as shown in FIG. 21. This phenomenon is emphasized in a case where the distance between the input block and the selected blcck is ~reat.
Such a problem takes place because the mismatching in the movement compensation cannot be fully compensated.
Next, the fourth embodiment implemented to solve the problem above will be concretely described.
FIG. 20A is a block configuration diagram showing the image encoding/transmitting apparatus to which the fourth embodiment is applied. In FIG. 20A, the same reference numerals are assigned to the same components ~25 as those of FIG. 19 and the description thereof will be :
~ omitted.
:

5'7 The characteristic item of the embodiment is a threshold control section 47 which includes a threshold generating circuit 47a controlled by the amount of t~e encoded signals temporarily stored in the transmission data buffer 46 for effecting a threshold control such that the threshold value is increased when the amount of the encoded signals ~S great and the threshold value is decreased when the amount of the encoded signals is small and a distance calculating circuit 47b for calculating the distance between the current input block and the selection block selected by the movement compensation section 42. The threshold generating circuit 47a includes, as shown in FIG. 20B, a threshold generating ; ~ section 471 and an auxiliary threshold control circuit 472 for decreasing the threshold values based on the calculated distance when the distance is great.
Next, the flow of the signal will be described.
~; The image input signal 400 is first subjected to the block generation in the preprocessing section 41 so as to be converted into the vector signal 401.
In the movement compensation section 42, reference blocks are generated and a;block having the smallest dlstortion with respect to the vector signal 401 is selected from the reference blocks. Thereafter, the selection block position information 402bl and the ~' :
selection vector signal 402b2 are supplied to the subtractor 45, while the selection block positlon information 402bl is sent to the distance calculatins section 47b of the threshold control section 47.
The details of the operation will be described ,lith reference to FIGS. 20C - 20D. FIG. 20C is a schematic diagram showing positions of pixels of a reference block, namely, pixels exist as indicated by small circles (o), 1 - 13 . From the distortion operation section 42b sends to the distance calculating section 47b the number (one of the numbers 1 - 13 in this case) of the block -having the smallest distortion among the reference blocks. The distance calculating section 47b converts the number into a distance by use of a map data beforehand set as shown in FIG. 20D and then transmits the distance to the threshold generator circuit 47a.
The threahold values Tl - T2 are controlled according to the amount of the image encoded signals stored in the transmission data buffer 46 associated with the previous frame such that the threshold values are set to great values for a large amount of the signals and are ~0 set to smail values for a small amount of the signals;
however, when the distance of the output 407b from the distance calculating circuit 47b exceeds a fixed value, the auxiliary control is conducted to reduce the threshold values.
The subtractor 45 effects a subtraction between the vector signal 401 and the selection vector signal 402b2, and then the differential vector signal 405 is delivered to the vector quantizing/encoding section 43.
-- 4~ -Thereafter, in the vector quantizing/encoding section 43, the differential vector signal 405 is processed in the arithmetic section 43a to calculate the average value m and the variance ~ of the diferential vector signal 405. In the validness/invalidness judgment circuit 43b, the validness or invalidness is judged by use of the threshold value Tl for the average and the threshold value T2 for the variance according to the following expressions.
m < Tl and a ~ T2: Invalid m _ Tl or a _ T2: Valid In this operation, if there exists a great distance between the current block and the selected block, the control is effected to reduce the threshold values by the threshold control section 47 as described above;
consequently, when the distance between the current block and the selected block is large, the possibility of the invalidness is lowered.
If the judgment results in the invalidness, the current block is assumed to be identical with the selected block and only the seLection block position in~formation 402bl is encoded and the resultant signal is temporarily stored in the transmisslon data buffer 46.
On the other hand, if the judgment results in the ~ validness, the differential vector signal 405 is normalized as the datum to be transmitted in the normalizing section 43c according to the following formula.

::
: : :

Yi (Xi ~ m)/~

(where, i = 1, 2, ..., k) ~Yi: i-th element of the normalized vector lxi: i-th element of the diferential vector Thereafter, the normalized differential vector signal 403c is quantized and encoded in the vector quantizing section 43e as follows.
First, a pattern most similar to the normalized differential vector signal 403c is selected from the code book 43d. As the transmission informatLon, the pattern number, the selection position information 402bl, the average value m, and the variance a are encoded and the resultant signals are temporarily stored in the ` transmission data buffer 46.
lS The image encoded signals temporarily stored in-the transmission data buffer 46 are transmitted in the frame-by-frame fashion.
Accordlng to the embodiment, for a selected block with a great dlstance for which the possibility of the mismatching is high as a result of the movement compensation, the control is effected to lower the threshold values and to conduct the encoding and transmission of the differential vector signal between the input block signal and the selected block signal, which leads to an effect that the occurrence of the shift of the contour of the block is minimized.
: ~
:::

;
:~

.57 In addition, according to the embodiment, although a description has been given of an example of the auxiliary threshold control in which the threshold values are set to the lower values as the distance between the distance selection block and the input block is increased, the same effect can be attained by conducting a stepwise control in which several kinds of threshold values are provided depending on the distances.
Next, the fifth embodiment of the image encoding/
transmitting apparatus will be described according to the present invention.
First, prior to the description of the concrete contents of the fifth embodiment, the general example of the technology adopted as the basis of the fifth lS embodiment will be described.
FIG. 22 shows the image encoding/transmitting ~ ~ apparatus as the basis of the flfth embodiment.
; In the analog/digital converting section (to be refe~red to as an A/D conve~ting section) 51 of FIG. 22, the pixels obtained by effecting the analog!digital conversion on the image input signals are processed to generate groups each containing k pixels being in the ;~ ~ neighborhood of each other in the image, so that for each lock, a k-dimensional vector input signal is generated.
A frame of vector signals are then stored in the frame memory 52, thereafter the vector signals are transmitted to the encoding section 53 in the block-by-block fashion.

:

:, , .5~

Incidentally, the encoding section 53 is provided with a previous frame memory (not shown) in ~1hich the image signals associated with the previous frame witn respect to the current image signal are stored in blocks.
The frame memor~ 52 is supplied with a signal from the read address generating section 57. The apparatus control section 59 outputs control signals to the respective sections.
In the encoding section 53, using as reference blocks a plurality of blocks each including the current vector signal and the block stored in the previous frame memory at a position associated with the block corresponding to the current vector signal, the vector signal and the block position information of the reference block are input-ted and then the image information is compressed and encoded as will be described later; thereafter, the encoded image signals are sent to the transmission data buffer 54.
Nex~, a brief description will be given of the operations to compress and encode the image information.
First, the distortion (such as the Euclid distortion or the absolu-te distortion) between the current vector signal and the reference vector signal is computed to ~- select a block having the minimum distortion from the reference blocks and then the selected block position information is stored.

A subtraction is then effected between the current vector signal and the vector signal of the selected reference block to calculate the differential vector signal.
The average m and the variance a of the differential vector signal are thereafter obtained and the validness/
invalidness is judged according to the following expressions by use of the threshold value Tl for the average and the threshold value T2 for the variance controlling the compression amount of the information volume.
m < Tl and a < T2: Invalid m > Tl or ~ > T2: Valid If the judgment results in the invalidness, the current block is assumed to be identical with the selected block and only the selection block position information and the information indicating the invalid block are encoded, thereby temporarily storing the resultant signals in the transmission data buffer 54.
On the other hand, if the judgment results in the validness, the differential vector signal as the datum to be transmitted is normalized according to the following formula. Assuming the differential vector signal to be = ~1~ 2~ Ek and the vector signal after the 25 normalization to be x = xl, x2, , Xk, then x = O ( - m) 1 E 1 k where, m = k i~

Next, the normalized vector signal is subjected to the vector quantization to output the index code I. For details about the vector quantization, refer to the "Image Dynamic Multistage Vector Quantization", the journal of the institute of electronics and communication engineer, IE84-l8~ The average value and the variance are output by effecting the quantizing and encoding operations on the m and a. If the result indicates the validness, the information notifying the valid bloc~, the index code, and the average and the variance undergone the quantizing and encoding operations are temporarily stored in the transmission data buffer 54.
From the transmission data buffer 54, the image encoded signals thus encoded are transmitted in the frame-by-frame fashion.
In the encode controlling parameter control section : -55, the threshold values are controlled according to the amount of the image encoded signals stored in the transfer data buffer 54 associated with the previous frame. If the amount of the signals is great, the threshold values are set to the greater values; whereas, if the signal amount is small, the threshold values are set to the smaller values, thereby controlling the degree of the compressiOn for each frame.

: ~ :

~ ~ ' Moreover, the compression of the image input signal in the conventional image encoding/transmitting method has been achieved by applying.the bloc~ generation of the image, the differential modulation, and the threshold value control as described above; however, when transmitting an image including a considerable change therein, namely, when transmitting an image having many blocks for which there exists a great difference between the current frame image and the preceding frame image and hence the threshold judgment results in the validness (the transmission information is to be required), the compression of the encoded information volume cannot be satisfactorily effected in same cases.
:~ In such a case, a long period of time is necessary to transmit a frame due to the great amount of the information; consequently, the time-lapse control is effected to thin out the input of the image signals for each frame, thereby controlling the information volume.
FIG. 26 shows an e~ample of the time-lapse control.
As shown in FIG. 26, the input image frames are ~ input in the sequence of A, B, C, D, E, and F. If the : frame A contains a great amount of encoded information, ~ the transmission of the encode information takes a long :: :
:period of time, and hence in the actual transmission of ~ 25 the: encoded information, the frames B and C are omitted, :~ namely, the encoded information is transmitted in the ~ sequence of A, D, E, and F.
: - :
~ 52 -:

Accarding to the image encoding/transmitting method of FIG. 22 as described above, the control of the encoding control parameter such as the threshold values for achieving the smoothing operation of the information generation volume is effected depending on the volume of the encoded signals contained in the preceding frame;
consequently, in a case where an abrupt change takes place between frames, the optimum encoding control parameter cannot be attained, which leads to a problem that the appropriate value cannot be obtained for the information generation, To overcome this difficulty, the fifth embodiment is implemented to solve the problem that the image transmission is attended with the visual unfamiliarity due to the delay of the image transmission time and the time-lapse control associated with the increase of the information generation volume.
Next, the image encodingjtransmitting apparatus of the fifth embodiment according to the present invention will be described with reference to FIG. 23.
FIG. 23 is a block configuration diagram llustrating the image encoding/transmitting apparatus of the fifth e~bodiment in whlch the same reference numerals : : ~ are assigned to the same components as those of the prior art:technolcgy shown in FIG. ~2 and the description thereof will be omitted.

- ~3 -.r;7 In this apparatus, there is provided a pseudo encoding/actual encoding control circuit 56 for accomplishing the control of the pseudo encoding/actual encoding. This system further includes a read address counter 57 for extracting the input datum for pseudo encoding operation from the digital signals stored in the frame memory 52 and a pseudo encoded signal volume counter 58 for counting the pseudo encoded signal volume.
Next, the flow of the signal will be described.
First, the image input signal 500 is converted into a digital signal 501 by the A/D converting section 51 and the obtained digital signal 501 is stored in the frame memory 52 in the frame-by-frame fashion.
: :
Next, the pseudo encoding is achieved as follows.
First, prior to the actual encoding, the pseudo encoding control signal 506 indicating an execution of a pseudo encoding is read from the pseudo/actual encoding control circuit 56 and is output to the read address counter 57. The read address counter 57 generates a read 20 address for the pseudo encoding and then the pseudo encoding data 502 is extracted from the frame memory 52, for example, for an image constituted by 30 block lines per frame as shown in FIG. 24, namely, the 5-th block line, the 15-th block line, the 25-th block line, etc., thereby sending the extracted signals to the encoding ::
section 53.

; ~ - 54 -.

Referring now to FIG. 23, the inside of the pseudo/
actual encode control section 56 will be described.
The change-over timing generating section 561 receives as an input (not shown) a control signal 509 delivered from the apparatus control section 59 controlling the entire apparatus, recognizes the timing to effect the pseudo or actual encoding based on the control signal 509, and outputs a ti~ing signal 661.
The change-over signal generating section 562, based on the timing signal 661, generates the change-over signal pseudo encode control signal 506 indicating whether the pseudo encoding or the actual encoding is to be effected.
Referring next to FIG. 23, the inside of the read address generating section 57 will be described.
When receiving the virtual encode control signal 506 from the pseudo/actual encode control section 56 for the pseudo encoding, the pseudo encode read address generating section 571 generates a pseudo encode read address 571 for tne pseudo encoding, whereas the actual encode address generating section 572, when indicated to effect the actual encoding by use of the pseudo encode control signal 506, generates an actual decode read address 572 for the actual decoding. The pseudo decode read address 571 or the actual decode read address 572 is selected according to the pseudo encode control signal 506 and is then outputted as a frame memory read address 507.

5~7 In the er.coding section 53, a predetermined encoding operation is conducted on the pseudo encoding datum 50Z
by using the encoding control parameter 505 determined according to the amount of the encoding signals of the preceding frame and then the pseudo encoded signal 503 is sent to the pseudo encoded signal volume counter.
Thereafter, the pseudo encoded signal volume counter 58 counts the signal volume in the pseudo encoding and then based on the output 508, namely, the pseudo encoded signal volume, the encoding cantrol parameter control section 55 corrects the encoding control parameter 505 to calculate the encoding control parameter SOS,~thereby finishing the pseudo encoding operation.
Referring here to FIG. 23, the inside of the encoding control parameter control section 55 will be described.
Based on the inputted actual encode signal volume 504, the encoding control parameter generating section 551 generates the encoding control parameter 651 for the next encoding operation. The encoding control parameter : ~ ::
651 is directly outputted through the correcting section 552 ~for the pseudo encoding; whereas for the actual encoding, the correcting section 552 corrects the 25~ encodlng control parameter 651 according to the inputted pseudo encoded signal volume 508 for the optimal encoding and then the resultant parameter is outputted.

::

.57 - Next, a pseudo encoding control signal 506 .
indicating to execute the actual encoding is deli-~ered from the pseudo/actual encoding control circuit 56, and based on the encoding control parameter 505 thus corrected, the actual encoding operation is achieved on all digital signals stored in the frame memory 52 in the optimum fashion.
According to the embodiment, the datum is extracted from the input frame, the data is once subiected to the pseudo encoding operation, the encoding control parameter is corrected, the actual encoding is executed, and the image is transmitted; however, the image transmission time is the same as that of the prior art technology.
That is, referring now to FIG. 25(a) illustrating the transmission time of the image encoding/transmitting operations in the configuration of FIG. 22, the encoding .
tlme for a frame is from 60 to 70 ms, whereas the line transmission speed is about 100 ms per frame, namely, an idle time of 30 to 40 ms exists before the encoding ~20 operation is initiated for the next frame.
Consequently, when the pseudo encoding is achieved :: :. :
during the idle time, the image encoding and transmission can be efficiently executed. ;~
; In the em~odiment, although the description has been given~ to an example in which the same encoding method applies to the pseudo encoding and the actual encoding, the s~ame effect can be developed by applying a simplified encoding method to the actual encoding.

.57 Moreover, in the example of the embodiment above, although the correction of the encoding control parameter is achieved through an execution of the pseudo encoding, the same effect or the improved effect can be attained by repetitiously effecting the pseudo encoding and the parameter correction several times.
As described above, according to the fifth embodiment, the datum is extracted from the digital signals contained in an input frame, the pseudo encoding is accomplished on the datum, and the encoding control parameter is corrected depending on the amount of the pseudo encoded signals, which enables to obtain the optlmum amount of the encoding signals and to stabilize the image transmission.
While the present invention has been described with reference to the particular illustrative embodiments, it is nbt restricted by those embodiments but only by the ~ appended claims. It is to be appreciated that those ;~ ~ skilled in the art can change and modify the embodiments without departing from the scope and spirit of the invention.

: ~ :

~ 58 -

Claims

The embodiments of the invention in which an exclusive property or privilege is claimed are defined as follows:

1. An image encoding/transmitting apparatus including a vector quantizer wherein said vector quantizer comprises:
a quantization preprocessing circuit for subsampling an image signal, for thinning out pixels from the image signal according to a pixel ratio, and for generating from the pixel signal an n-dimensional vector;
a group of input buffers for respectively storing each said input vector in a sequence of a quantization processing and for repetitiously reading the input vector at an interval of the processing period;
a vector output circuit for beforehand preparing an output vector corresponding to a probability distribution of the input vector and for selecting and outputting the output vector based on an index input;
a distortion computation circuit for computing a distance between the input vector and the output vector selected and output;
a plurality of encoding sections disposed corresponding to the buffers;
index output generating means for generating an index for a period of a subsequent processing based on the calculated distance value outputted from said each encoding section and the input vector and for sending the index to each said vector output circuit; and a quantized index output circuit for generating an index of a quantizer from said generated index and for outputting the index.