Method and device for processing image signal

[Detailed Description of the Invention]

The present invention relates to the method of generating the signal for which it was suitable by processing an image signal and changing redundant information via the medium of removal Si and a I was limited. bandwidth, and its device. The present invention relates to the method used for a video compression system, and its device in detail. Much bit reduction art used with a video compression system is known. Since it is hard to carry out noise degradation of the digital data style, it performs the time (encoding is called below) of processing the TV signal sent via a transmission channel, for example, digital coding, in many cases. When encoding a TV signal to a digital type, many pits more than t are required to give the gray scale of a range permissible to each of each pixel (pixel) of numbers of/theta 10,000 which forms one picture. Therefore, a bandwidth also with Da theta megabit Yo Si wide general per second is required for the digitized TV signal which is not processed. It is that almost all the bandwidths of a satellite will be occupied as this video signal, as for, processing Lettie is not is common supposing the communication link is carried out between the ground and a satellite. Though left behind for the of reason, the number of channels which other users can use becomes very small. although T / communication channel is generally used -- this channel -- per - bandwidth -- a /S megabit -- it is poor. Therefore, a practical and effective method which narrows the bandwidth of the digitized TV signal is required, and even if it performs Ah Si and bandwidth transmission that used only few channels for transmitting via a communication path for this reason, and was narrowed, This 7 which maintains the quality of a transmission signal is required. the [ which was transferred to the applicant for this patent / United States patent ] -- redundant information is removed in a 11.302.'7Vs item, and the improved scene conformity coding art which made the bandwidth small is indicated. That is, this United States patent is indicating the single path digital video compression system which compares a conversion factor for every block by intra-frame one, and performs a two-dimensional cosine transform, without carrying out preliminary statistics matching, i.e., preliminary treatment. Each frame of a video image is divided into the block of space Subframe, i.e., a predetermined matrix, and a system makes a space domain perform domain conversion of the pixel of each block. A system normalizes a conversion factor accommodatively so that data may be generated at the rate accommodatively determined as a function of the degree of fullness of a transmitting machine buffer. In this way, the generated conversion factor data is encoded by the transmitting machine buffer according to the zero coefficient run length Hoffman code and amplitude Hoffman code which were stored in asynchronous. In this way, the encoded data is outputted from a buffer at the synchronous rate transmitted via the medium of the limited bandwidth. It is not filled completely, without a buffer becoming empty completely, since this system is adapted for the rate which determines the degree of fullness of A buff 7 and at which data is generated and controls this. In the receiver of a system, send data is stored in the buffer of a receiver with the synchronous data rate of the medium of the limited bunt width. Next, this data is outputted to asynchronous from the buffer of a receiver, and is decoded according to the contrary of encoding of a transmitting machine. In this way, it is normalized conversely, and inverse transform of the decoded data is carried out so that the original video image may be expressed. U.S. Pat.'1, 302, and 7 Off S are reducing relative redundancy with the intra-frame coding (coding in frame) art in which intra-frame comparison of the Cosain conversion factor was used. Although other Technology Si is doing big improvement, bigger compression is required for this patent. Inter-frame coding (inter-frame coding) art is used for making low a rate required for video transmission other than intra-frame coding art until now. In this art, generally it all [ each bidet-off v -A ] holds in the memory of the both sides of a transmitting machine and a receiver, and only an inter-frame change transmits to a communication link. The quality of the coded image changes according to nature Ha of image A by whom this inter-frame coding method was coded in contrast with the intra-frame coding method which changes according to the detailed quantity in each single image frame, and an inter-frame difference. This inter-frame difference is called a motion (motion) in many cases. The inter-frame coding method is classified for whether being Arrow at a kinds. That is, it is To about space domain coding and the conversion domain coding method The. In the inter-frame space domain coding system of a certain real time, threshold processing of the space domain data is carried out, a frame difference signal is acquired, and this is stored in the buffer of a transmitting machine. This Threshold value is accommodatively determined as a function of the degree of fullness of a transmitting machine buffer. Since the picture reconstructed as a result of using the threshold of a large range was destroyed severely, this system was not satisfied completely. In order to prevent such image destruction, the both sides of the subsampling method are proposed the space Zabusan pulling method and temporarily until now. Although the these-proposed method performs a certain amount of improvement, the system which performs compression satisfied completely is not provided. The inter-frame coding method for performing conversion domain coding is not used widely. Since space-conversion domain conversion of expensive real time is performed in the system most often proposed, the conversion domain coding method is given in the impractical thing. Nevertheless, several sorts of inter-frame conversion coding systems are proposed until now. the [ for example, / digital-image-processing 5PIE ] -- a 1/9 volume t+occ iqq person -- the conditional supplement conversion video compressor is stated to Mr. 77~9thg page of A \ Lee W Jones Junior's paper 1 conditional supplement Hada Cord Video Combrettizer 1. The conversion factor of one Arene, is stored in this system, and it is conversion object i of next Arene, about this : ratio s It is carrying out. If the difference of an inter-frame Threshold coefficient exceeds a Threshold value, I will think that the picture or the sub picture changed. The difference of this coefficient is measured to the predetermined vector of the arrangement of a conversion factor. 1EEE report C0M2S about communication, /977-year l January l one-day issue, Another inter-frame conversion coding system is indicated by the paper "inter-frame cosine transform picture coding method" of the page [ 13.29~1339th ] Skeie lake, W Kay Platt, and Mr. Si- varnish Robinson. The inter-frame difference signal is used for a part of difference pulse code abnormal-conditions prediction coding art in this system. The conversion factor is predicted using a fixed-arrow alignment prediction function. In Mr. Barry and day Pid N Hain [ of the application 111sPIE Coworker theta7 volume/979 year $21. g-27? page of digital image processing ], and f 7'IJ Knee Casting Jones Junior's paper "conditional supplying method using motion prediction" Is -- another conversion domain coding system is described. Such a system is known for that (supplement) which sends only the data which changed as a conditional supplement system. Although inter-frame difference analysis is conducted in the measuring-change D space domain in this Jones's system, Jones's method of conditional supplying does not perform a data compression completely. There was nothing that the signal-processing system of a large number which include the above-mentioned art until now is proposed, and carries out data compression perfect [ each-other gap ] and sufficient. Therefore, the signal processing method and device with which it was improved for data compression systems are desired. The outline of an invention The present invention relates to the signal processing method and device using intra-frame one and the inter-frame variable prediction conversion coding method. A picture is displayed by the continuation frame of the one-dimensional arrangement of a digital signal. A digital signal is changed so that the conversion factor for each frames may be formed, and a prediction conversion factor is formed using the set of a variable prediction factor. The prediction conversion factor of each frame is compared with the correspondence real conversion factor of a frame, and a conversion factor difference signal is generated. This difference signal is processed so that the range of that value may be controlled. The difference signal processed in this way is statistically coded so that it may be expressed as code length with a short value with a high frequent number and may be displayed by code length with a low long value of a frequency count. This coded signal is stored in a buffer memory for transmission. The coding signal in a buffer memory is receiver-- Sent via the medium of a Information limited bandwidth, and receives information processing. The code which identifies variable prediction 7 actor's set currently used with the transmitting machine is included in this information processing. Also with a receiver, a prediction conversion factor is reconstructed using the same set as a variable prediction factor, and the original picture in a transmitting machine is reconstructed using this coefficient. Since the degree of correlation of an inter-frame correspondence conversion factor is generally a non-line type in variable, a variable prediction factor is used for the present invention. In a general picture, the degree of correlation of a low frequency coefficient is higher than the degree of correlation of a high frequency coefficient. A prediction factor is changed as a conversion factor changes, and it is chosen. The set in which prediction factors differ is chosen as a function of the inter-frame difference of image data. The present invention is carrying out kana Si reduction of the data A fellow which removes, codes and transmits the redundant data in an intra-frame space domain and an inter-frame time domain. Generally since [ prediction of a coefficient / use To fL To data / history ], it can predict the same conversion factor on the both sides of a transmitting machine and a receiver to it. The coded data is generated from the difference of the size of a real coefficient and a prediction coefficient, and, generally change of this difference data is smaller than change of the coefficient itself. If the inter-frame coefficient correlates highly, it will be small, and only very little data in which this difference is shown will be coded, and a - inter-frame difference will not be transmitted. However, if the degree of correlation is low or there is, the coefficient itself will be coded and transmitted rather than coding and transmitting a difference. [ no ] According to the method of carrying out scene adaptation coding of the inter-frame coefficient difference using variable prediction of the present invention, the object of providing the improved signal-processing system which can reduce A knot width required to transmit the picture of 1 high quality is achieved. The objects and the features other than the above of the present invention will become clearer than detailed explanation of the desirable example of the present invention shown in the appending drawing. DETAILED DESCRIPTION Abstract In Drawing 1, the digital signal which displays the picture which should be processed and should be transmitted to a receiver occurs on the hill of input bus 12. Generally a system of Drawing 1 is a video compression system. Sending Iha machine 24 processes a digital image signal with bus 12, looks it like [ bus 15 ], and outputs a transmitted signal. the processing performed with transmitting machine 24 is combination about intra-frame one and inter-frame scene adaptation coding -- it is a thing. The signal of bus IFM: is sent to receiver 25 via transmission bus 8 or other transmission media. Receiver 25 receives the transmitted signal on bus 17, carries out inverse transform of the received signal, and reconstructs a digital signal into output bus 21k. It is the reconstructed signal with which the signal of , displays the digital input signal of bus 12 A on bus 21. Each block of a digital image is displayed by the conversion factor of Co dimension arrangement. Generally, the degree of correlation of the correspondence coefficient under correspondence block for every frame changes, and is a non-line type. When the degree of correlation of the coefficient for every frame is high (i.e., when an inter-frame difference is small), only the slight data which actually displays a difference with a value as a predicted value is transmitted. When Brity lid P used for forming a predicted value is Clothes(ed) by the following formula, The power mean difference serves as the minimum. p=k+ (amplitude) (/-k) (mean amplitude) -- k is a constant proportional to the degree of correlation here. Since the average value of all the coefficients is a cello, the 2nd paragraph of the account equation of L is eliminable. In the case of a value smaller than unit quantity, some transmission data displays amplitude information on a coefficient absolutely, and since it is removed, or leaked and taken out in increment, the spurious error is known for the following frame as a leakage factor or a prediction factor. When there are not a noise and an inter-frame difference, the inter-frame degree of correlation becomes a uniform unit to all the coefficients. However, operation without a noise cannot be carried out under ordinary A matter. Under actual environment, a great portion of coefficient set shows few differences (the same frame). Does this prediction factor by which sectional meeting specification was carried out approach a unit value? In the low frequency wave end of spectrum, since the degree of correlation is high, even the inside of the block which shows an inter-frame big difference becomes at this. The specified anticipation factor changes over an intermediate frequency field according to the size of an inter-frame difference. A variable anticipation factor is chosen as a function of the inter-frame difference in image data. The Accumulation total value of resolution picture De and - evening difference in a block is the Toshiko Selle so that it may become a standard showing the size of the degree of data correlation in a block. If it is indicating that Prospect and a count number are small and the degree of inter-frame Knobloch correlation is large, other sets of a prediction factor are Selection(ed) three times. The size of genealogy and a count number is used for choosing which set of a prediction factor should be used. Each coefficient in a block is connected with one anticipation factor which differs from the factor of a predetermined set, and is ing. It is set between the range iJ unit value of the anticipation factor in each set, and zero, and, generally the Prediction d11] factor to a specific set changes in non-line type for every coefficient. In a certain example, selection of the set of a prediction factor is determined by carrying out resolvability of the difference between the two-dimensional space digital signals under correspondence block in continuous Off 1/- Beam which have not been changed. In other examples, selection of the set of a prediction factor is determined by analyzing the difference of an inter-frame conversion factor. The set of a prediction factor is stored in prediction day pull. In a desirable example, as a means to adjust a day krait, the object for the histories of one of these prediction factors is not carried out, but they are used as a means to optimize reduction of the data which needs a statistics value to display the various blocks which change with the quantity of an inter-frame difference. In the desirable example, rate regulation is mainly carried out by Threshold and a normalization means, and the coefficient of a size lower than the Ha Potential threshold is dealt with by the Thresholding method as the data, i.e., the data which cannot be transmitted most easily, like the minimum. In the common system described below, a conversion factor is displayed by 72 bit words. Since g quantization levels can be used if this /2 focus word is used, the number of the coefficients deleted by the Threshold method can be adjusted correctly. The transmitting machine operates so that image data may be processed, as image data is processed by the receiver. As for this double processing, a transmitting machine is the useless exact difference (it performs within a transmitting machine so that LT item can be formed.) of the prediction data of the following frame. A transmitting machine / Drawing 1 of a receiver system It is generating To shake about the coefficient which carries out y and Selection of the digital image which 3D conversion processor 2 performed three-dimensional data mapping in Drawing 1, and was given to person hippo 712L2. 3D conversion performed by processor 2 includes 2D (two dimensions) conversion. This JD conversion is 2D conversion which is generally indicated to United States patent 41..No. 30.2.773. More VC3D conversion pro Setza 2 generates the 3rd-dimensional coefficient as a result of inter-frame correlation. The conversion factor which occurred by this 3D conversion processor 2 is place JJ by coefficient processor 3 further! It is carried out. A coefficient is processed by Threshing and normalization, and this processing answers the data rate in statistics Gouda 4 to the average, i.e., the 1-pair-of-shoes data rate, passing through transmission bus 8, and is controlled by transmitting machine control device 1. Statistics Cog 4 codes the signal of coefficient processor 3 so that the required number of data bits may be made into the minimum. This coding is performed by assigning the code with little number of bits to a data value with high generating frequency, and assigning the big code of the number of bits to a data value with low generating frequency. When transmitting machine 24 performs these operations, a transmitting machine is bus 15k (since it comes to optimize Df-krait, the picture of maximum quality is transmitted to transmission bus 8 of the bandwidth which can be used.). Transmitting machine 24 performs inter-frame (inter-frame) scene adaptation coding like US Patent No. and intra-frame (inside of frame) scene adaptation coding of No. 302,773. Although full 11fl of statistical coding art and scene adaptation coding is indicated to the account United States patent of L, it can be used also by this system. In Drawing 1, the receiver in decoder 6 decodes transmission No. 16 by which bus 1T A was received, and outputs a received signal to bus 201. the received signal boiled line 20 is inputted into reverse 3D processor T, and processor T generates a digital image signal into bus 21. The signal on bus 21 is writing (it Conclusion one and displays.) about the digital image signal of input path 12 7. 3D conversion processor the 2' 3D& Furoseta 2 in transmitting machine 24 of Drawing 1 is shown in Drawing 2 in detail. This processor 2 is the cosine transform device as device 18 in US Patent No., 3 theta, and No. 2,773 with 2D conversion processor 26 same in a certain example including 2D conversion processor 26 connected to the 3rd-dimensional coefficient generator 27 via bus 2B. Generally the input signal of each pixel of the digital image of bus 12 A of Drawing 1 is displayed by ff (i, k) (m, n). When there is this input, it is an output (it becomes a cosine transform coefficient displayed by Fl (u, v).) from 1.2D conversion Brothersa 26 (m, u). At least in the procession of the block in a specific frame, subscripts m and n are here (6 is shown and subscript f shows a frame di item.). The procession position of space domain Bichsel in a block is shown in variable j. LmuHV is a conversion variable which arises from the cosine transform of the pixel block in a matrix in JX of the pixel in a block. if the 1st second figure is referred to -- this figure -- frame theta, the second, and 2 -- a plurality of frames 144 displayed by ...... and F are shown. each of a frame shows pictures, such as a television display, -- the next -- each of frame 144 -- sequence 0./, ......, (m-/) and line 0 */, and 2 -- it is arranged by the block matrix of ...... and (N-/). In a general example, since each frame has three sequences and a line of 32, the maximum of a between and the maximum of N become equal to 32, respectively. General one out of blocks 145 which have block coordinates (2, 0) is shown in Drawing 77 in detail. Block 145 of Drawing 77 comprises a JXK matrix of pixel 146. a general example -- this matrix t -- since it is the pixel arrangement of /Filtration x / 6, any maximum of J and K becomes equal to the /second. Each pixel like pixel (/, theta) 146 of Drawing 17 has a size generally displayed by the number of g bits during data processing. The number of these g bits that show the size of a pixel arises into input bus 12 only once for every sampling frequency CLK. In the Off * lock of the pixel of the 25 seconds, the total and g bit digital number of 256 pieces are inputted per seven blocks to A D conversion processor 26 of Drawing 2. The pixel of eye watch, ri (m, n); the block of eyes, and input bus 12J to the f-th frame: (1, k) The size of each signal is displayed as ff"" (m, n). To each block of the size of a pixel expressed in Drawing 77, The and 21-d 2D conversion processor 26 perform a cosine transform, and are Ff (u+v) to the block of eye watch (m, n) of the image signal f-th in 7 Rehm (.). . The conversion factor displayed is formed. A Cosain conversion sequence is carried out and U is !li A (■ is cosine transform Ryosuke.) here. r Into - cosine transform d1 of the picture arrangement of lambda dimension arrangement of f f (Jp k) published by Image trust (+n, n) which arises into bus 28 of The and 2 figure at the time of theta,/, ......, V-/-- it is expressed as follows. Ll and V = 0 +/r -- the time of ... and N-/-- Cf 1 (j, k) rr+ and ncosc (2J+/) uyr%2Ncos C (it is +/to 2) vyr/2N] -- here (u, v) It is C (u, v) =/// 2(u, v)=theta at the time of =00. The time C (u, v) and =/f Al. The ranking which processes each block of frame 144 of the 1st second figure, and carries out attitude to bus 28 is sequence m one-line n0Jllli grade. a hill -- the block uniquely identified by the count of between, N Over, and 70 A bit displayed by faucet b since it is 32, respectively in an account example -- /theta and 24' There are. If the count of these 70 A bit is divided into power 6 Tent of S bit of Task which displays the count and row number n of 5 A bit which display column index m, it is preferred. If the increment of the line count is carried out whenever it processes a line which is different in the same sequence, and all the sequences of a line are processed, the increment of the sequence count will be carried out and all the lines of the following sequence will be processed. Although processing of conversion factor F f (m (u+v), n) does not advance in order of the sequence of a uxV matrix, and a line, it advances along the Ma IJ socks square wire in a desirable example. Following this diagonal line processing, coordinates (U, V) are (0, 0)', and <0.7> (/,/). (/, theta) (2, 0) It is scanned in order of (0, 2) (0, 3), ......, (2S! and 23 A), and (moon 3 *25S). the ranking which processes procession coordinates (u, v) is displayed by the sequence of C number -- here -- C -- 0./and 2. -- it is equal to ... and 255 respectively. The relation between numbering of C and numbering of (u, v) is shown in Table 1 of the arrow. Table ■ 0, <0.0> /, 2. (0, 7), </, and 0. (3.11.3. 2, OL (/and Ha (0, 2))) 1 .9. (0, 3) (/, 2) (2,/) (3, theta), 10, /'1. (11, 0), and (3,/) -- (-- 2 and 2) (/, 31. (0, 4).), /3. and E (0, 5) (/and A) (A,/) (5, theta), (0/3) . (/,/, 2). (/4.0) (/3./(0, /A)) <0./3, and (/ and /To') (/IIL and Ha (15, 0)) (15,/) (/lLt and 21 (2, #)) (/, 15) Fu6., J'? (Hmm, 15), and (/3. / Ri) (a /person, /3) (15, Hmm), K, Scolding, 2 power (73/3), and C70/S (/A, /A). lambdar3, M (i A, /3), </S, /Da. Merit (15/3>) Cosine of Akko of bus 28 To from processor 26 is shown in the 1stg figure, and it is processed by the 3rd-dimensional coefficient generator 27 so that redundant information may be removed. Generally, the coefficient of a specific frame is compared with the coefficient of front frame f-/, and the output of bus 11t is difference signal ef (U, V) (□). - It is /] and this signal shows the difference of the coefficient for Prediction d of the frame of f@ eye and the 1st frame coefficient to a block (m, n-/). This predicted coefficient is determined as a correlation function of a correspondence block of the frame of the 1st and eye watch (f-/). If b is used for ranking being again attached as C and a coefficient variable (u, v) displaying (m, n), it is signal e f ('p V) (the difference signal of m and n -/] becomes ef (C) (b7).). Difference pulse code abnormal conditions (DPCM) are the prediction coding methods, and the difference of the present coefficient Ff (u, v) and the estimate of coefficient F ? (Ll, V) is given as follows by this method. Presumption of an ef(u, v) =Ff(u+v)-Fp (u+v) coefficient is based on inter-frame phase tut lk of the present freight and a previous frame. The coefficient of the present frame is defined by the paragraph of a previous frame as follows. a (u, v) paragraph in (u, v) Ff =a(u, v) Ff-/(u, v) and an account equation is defined as denoting a prediction factor by (u, v) (u, v) a =p /[a2(u, v)]. p (u, v) is Covariate of Ff (u, v) and F (u+v), and f-/a" (u+v) is r here : (u, v) It is having assumed that it was fixation by change of Ff (u, v) of f- / Can. Since both a receiver and a transmitting machine presume the value of a coefficient similarly, a receiver reconstructs a coefficient from the transmitted signal. Threshold selection Age of the signal of To be bus 13I: is carried out by Return to processor 2, and it becomes a difference signal which occurs by coefficient processor 3 of Drawing 1 and which is normalized. In the 3rd-dimensional coefficient generator 27, the feedback signal of bus 13E performs inverse transform and Threshold addition in the case of generating of the predicted value of a conversion factor. In this way, the predicted coefficient value is used for the comparison at the time of generating the following difference signal in bus 11 7. this 3rd-dimensional coefficient generator 27 is full -- M11 is shown in Drawing 3. The 3rd-dimensional coefficient generator - Drawing 3 In Drawing 3, the corner in conversion factor of the f-th frame and the b-th Off * lock occurs in bus 28J: in value (c) (C has value 0./,;l, ..., 255 here) Ff b. It is stored in block delay 38 in a /block cycle, and these coefficients are outputs Ff of - block delay to bus 36. (C) (b-/) It is outputted. One input is outputted to coefficient t Subtractor 45 of bus 36f:. Decrease n machine 45 receives predicted value Ff (c) and (b-/) of a coefficient from bus 49. The difference of the predicted value of bus 491: and the actual value of bus 36h is difference signal ef of output path 11f:. (c) (b /) It becomes. The difference signal of bus 11t displays the three-dimensional coefficient which decreased the inter-frame degree of correlation. Frame coefficient F of the point from which Flames Doat 41 produces the predicted value of path 49 A in bus 47J: (f-/) Since it is predicted as a function of (c) and (b-/), the degree of correlation becomes an inter-frame thing. Crossroad letter r means that the coefficient is reconstructed from the transmitted difference signal. the useless difference signal of coding passes along bus 13 -- a reverse normalization unit -- and it Heffeid back threshold addition unit 42. Coefficient F t which this signal was outputted to bus 50, was added to the prediction coefficient by bus 49 A, and was reconstructed by bus 48 when reconstructed difference signal e, (c), and (b 7) were processed by unit 42 (c) It is set to (67). Coefficient F (f-/) which the coefficient of this reconstructed bus 48L was stored by Flames Doat 41 between claim time delays, and the previous frame accomplished re-1ii2 to bus 47J: (c) and (b-/) are outputted. Flames Doat 41 also stores the coefficient of the frame of the point of previous block b, and outputs block coefficient [ of the point reconstructed by block analyzer 39 via bus 35 ] F(-7) (c) b. In Drawing 3, block analyzer 39 compares the e-th block and the cosine transform coefficient of the f-th frame, previous frame, i.e., corresponding cosine transform coefficient by which the frame of eye watch [ <t- / ] of the b-th block was reconstructed. Although block analyzer 39 generates an output signal into bus 46 according to the size of the difference signal which displays the difference during a block with the present frame block coefficient and a previous frame block coefficient, he indicates that this signal chooses one of many coefficient tables in unit 40. The prediction factor of a different set is stored in each table. A difference indicates that an example is not subtracted from the thick cosine transform coefficient which all the coefficients from unit 40 will serve as a cello, and will be supplied from block delay 38 if it kicks with subtraction machine 45, either. Under these conditions, the actual coefficient from block delay 38 is outputted to bus 11J:, without being changed. On the other hand, if an error signal is small, the coefficient taken out from one of the tables in unit 40 will become larger. Generally the output from the coefficient table in unit 40 is a non-line type. It is expressed by Drawings 7 and 3 about the details of the character of the table stored in unit 40. Off * lock analyzer 39 describes below the mode of the A frame * lock of the house left eye of f, and eye G (f-/) which carries out A frame * Rotto comparison in detail with reference to a A group figure. Drawing 7 of a block analyzer In the Hiroshi figure, block analyzer 39 of Drawing 3 is shown in detail. A block analyzer receives every one 7.2 A bit coefficient (c) Ff b of tg eye of the b-th block of - poet hippo 228 To, Bus 35J: Subtract every seven coefficient F(f-/) (c) b by which the frame of eye watch (f-/) of , was reconstructed, and generate a /2 bit difference signal in bus 35J:. Although the signal of bus 65h addresses table 61 and a factor is outputted with [ of a non-line type /: / bus 65J] g bit dignity, this factor is a function of the difference signal of bus 65 A. It is added with the Accumulation(ed) block [ by which the factor is stored in accumulator 63 with / from table 61 of bus 66t / dignity ] dignity to a factor. It has a factor with Accumulation prize to all the factors of the former [ accumulator / 63 ] in the same block b. Success-Ri from 7 key j-Mouret -Taro 3 (an output addresses table 64 and specifies one of the prediction factors of several sorts of different sets.) Drawing S shows the common data currently stored in table 61. A sign bit is disregarded including the data bit of //in which bus 65 [c Filtration 5] has table 61 and the 72nd sign bit. Drawing 6 shows the data stored in table 64. Generally, the output from table 64 is a signal of a bit, and this signal shows one of the prediction factor tables of Hurry in unit 4o of Drawing 7. C=theta,/, 2.-, 233 (in the D case (D2!;b (one each of D coefficient differences gives an input to adding machine 62 from table 61.))) Adding machine 62 adds the Astonishment-value in accumulator 63 to the value from table 61. Although accumulator 63 can be Accumulation(ed) to the value of 2'6, this value is equal to 6 g and 5J'6. For example, the value of /6 A bit of bus 68 hill generates one of the outputs of Hurry from range table 64. If the Accumulation value in accumulator 63 is less than 5'0, an output will serve as zero. If Accumulated output is between 50~/theta0 An output becomes /, and if an output is among 700~200, the output from table 64 will serve as Co. If the Accumulation value in accumulator 63 is larger than 200, the output from table 64 will be set to 3. The value from which output 0./from table 64 and 2.3, i.e., Gucci, differ is coded by Cobbit bus 46, and one of the sets of Da one of a prediction factor is specified as TO in unit 40 of Drawing 3, T/, and T2 and T3, respectively. Details of the table of Drawing 3 and prediction unit 43 is shown in Drawing 7. Drawing 7 of Predictor unit- in Drawing 7 -- memory 70 -- t table TO1TI and ■2 -- and 73 k store of is done. One of the tables of these Gucci is chosen as the Cobbit signal on bus 46 with the block analyzer of a Yo Si A group figure. one each of the tables in memory 70 -- a different /lambda bit value of 236 pieces up to -- although stored, these Gentlemen correspond to each of the coefficient in a block. Every one value from a table is addressed at a stretch by g bit address on bus 99, and outputs 72 A bit output on bus 58. The value in the table of coefficient memory 10 has a size between theta~/, and these values were reconstructed (c). Coefficient F' (1-/) ('[ here (b-/) ] ccuO,/.) ......, 2S! Leakage 7 actor Into to Gentlemen [ that it is equal to; ] is determined. IJ-cage factor to the Gentlemen of C is accessed from memory 70, is outputted to bus 58, and is multiplied by the reconstructed coefficient value on bus 47. Multiplier 11 hangs a leakage factor on a coefficient value, and is 72 Pitno (the predicted value of conversion factor Fpf (C) (b-/) is generated on A 49.). The predicted value of the conversion factor on bus 49 is inputted into adding machine 44 and subtraction machine 45 in the coefficient generator of Drawing 3. Reconstruction conversion factor r-" (C) formed from adding machine 44 of Drawing 3 in Drawing 7 It is Saif (b-/) to E CL. 72 to memory 1 father is stored in one of 12-2 for every Tal. The reconstructed conversion factor is inputted into either the 1st frame memory 72-1 or the 2nd frame memory 72-2 for a store. Memo lj 72-1 and 12-2 are 70211x2 logically, respectively!; it is together put so that A locations may be constituted, and the word of a /lambda bit is stored in each location from bus 48. While memory 12-1 or one side of 72-2 is storing the information from bus 48, lead-out of memory 72-1 or another side of 772-2 is carried out, and it outputs a conversion factor to block Toilet memory 74. The skew of coefficient memory 7G, frame memory 72-1, and the addressing of 72-2 is carried out to the order which a correction conversion factor value generates in a correction time sequence for delay peculiar to the processing loop of Drawing 3. Each addressing of these memories is under control of address control unit 69 which constitutes some transmitting machine control devices 1. Fundamental timing and addressing are determined H% of the data on bus 12 by the input data rate to the transmitting machine of 7 figure. If an input data rate is clock rate CLK, the new address of each conversion factor will be generated for every cLK period. A method desirable although the address of a coefficient is generated is the method of stopping a coefficient counter with the CLK signal [ more ] shown, for example in Drawing 72. In the 1st piece figure, as for coefficient counter 133, Ah Si and this counter output coefficient faucet Oct to address bus 136 of g bit at the counter of g bit. The carryout from coefficient counter 133 serves as coefficient synchronization pulse csync on bus 137, and the career udo from counter 133 clock-izes 70 bit-block counter which outputs block address BCt of 70 A bit on bus 138. Although the career udo from counter 13 Doo clock-izes /bit counter 135, this counter 135 outputs the signal of a /bit on Ah Si and bus 140 by odd number and even-frame 1 Indicator. address Tsuki j Pupil 69 of Drawing 7 -- an output bus -- and line 136~i+ui wins popularity and it controls memory 70.72-1 and address generating of 72-2. In Drawing 3, coefficient count Oct is 2. of block delay 38!;l, and X / 27-Doume Maury are addressed, and the conversion factor Fl (c) bk store to all the values of 231. Pieces of the b-th block is carried out. Therefore, the output from block delay 38 serves as a block of eye watch (b-/). The output from Predictor 43 corresponds to the block of eye watch (b y). In Drawing 3, since block analyzer 39 conducts analysis on the b-th block, the output from Predictor 43 also receives the frame of the b-th block and (f-/) eye watch. Therefore, block analyzer 39 analyzes the b-th block of the frame of the 1st and (f-/) eye watch. address control unit 69 is Hemorrhoidal at the memory of correction block information all figure 7 -- an address required in order to access Si and this -- generating To shake. Memory 72-1 and memo! The information stored in J72-2 is offset as a function of the number of delay in the processing loop of Drawing 3. Though natural, the cycle of what kind of number may be sufficient as this offset. Only a /block is offset in the indicated example. From memory 72-1, when considered as reading, generally a note of information is made! Information is written in 772-2. - All the /Cobbit coefficients of Co SA Pieces corresponding to each of a block of /theta2'1 piece are stored in memory 72-1 during a period. After being stored in memory 72-1 for /frames, the coefficient of the following frame is stored in memory 12-2. When the coefficient for The lambda frames is stored in memory 72-2, the memory 72-1 Yo Si successive occurrence of the coefficient from a previous frame is carried out, and frame coefficient [ of eye watch (f-/) ]"(f-/) (c) b is formed. one block address count -- Yo -- a jib -- since the skew of the lock address is carried out, this system must read and carry out the b-th memory address for a count in front only a /block cycle from Duke of 10 Toc count of eye four (D-/). This offset at the time of block address specification is controlled by address control unit 69. Though natural, when delay of an address count longer than a /block arises within the loop of Drawing 3, Yo Si between block memory addresses also has to use long offset. In Drawing 7, block delay 74 is a memory of 2S second x7.2 which stores the word of 72 A bit, and only a /block is delayed in read-out of each block from memory 12-1 or 12-2. Therefore, if the block of a fit eye arises into bus 47 (b -/), the b-th block will arise into bus 35. in 8gg figure -- the -- although the prediction factor of TO is set to /and theta when it carries out, and the details of T3 can be shown and the output from range Chief of Drawing 6 is zero, prediction factor table [ in coefficient memory 7 of 7 figure ] Ttheta, T/, and This is coefficient F"t. (C) < b / A prediction 1if means being the same as the value by which conversion factor Fft-mosquito C< b-/was reconstructed. When the output from range table 64 is /, the prediction factor of T/of a A chest figure is chosen. The prediction factor of table T/is a non-line type value which begins from 7.0, and is delayed to about 0.S as a function of the coefficient number displayed by coefficient faucet CCt which is between each block and changes between the theta~25 lefts. If table T2 is chosen, a prediction factor will change [ to / from /and 0 / near theta ] to a non-line type, and if table T3 is chosen, it will indicate that a prediction factor is set to theta and all Prediction lI Coefficient serve as zero. The /1lllJ factor of the set in which Drawings 5 differ displays the lrt nature a cosine transform coefficient is statistically predicted to be as a function of an inter-frame conversion factor comparison performed by block analyzer 39. Drawing 7 of a coefficient processor Coefficient processor 3 in transmitting machine 24 of Drawing 1 is shown in the Fly figure in detail. Difference signal e f (u, V) (b and Q are inputted into bus 11 of /2 Repose.) equal to ef(c) and (b /) Threshold value Tc is subtracted from the difference signal on bus 11 -- its difference Ha /e -- 2 bits is outputted to lotus 78. The sign bit of bus 11 bypasses Unit 70 and 77, and is re-connected to bus 13. The difference signal on bus 78 after Threshold value all subtracting within Ah Si and subtraction machine 76 with a constant, for example,/, and tide in inside Tc of an example is normalized by normalization 7 actor//N as Yo. Nf changes between A ~ / 6, and is dynamically specified with coefficient processor control device 81 on David Bath 86 by the general example. the input on bus 78 -- 1lambdathetaAg~10 -- since lambdatheta4t changes to 7, when the output/iNf on bus 13 formed into the right A child are equal to /1., it changes to 0~7.2g, and Nf is A / :1 -- the time -- theta~S/2 -- it changes in the range. The size of Nf changes with processor control device f/, and controls the rate which generates data within statistics kneader 4 of Drawing 9. By processor control device 181, Threshold Tc is dynamically controlled by another example, and is outputted on /2 bit bus 85 in it. If the value of Tc is controlled dynamically, the rate which generates data in statistics coda 4 will also change. the operation of the coefficient processor of Drawing 9 -- the -- it is shown by the graph of theta [ 1 ] figure. The range of the input number on /Cobbit bus 11 is displayed by the horizontal axis (L //), and the control range of the attitude number on 9 Bit bus 13 is a drawing straight axis (displayed by L /and 3].). In Drawing 9, coefficient processor 3 can make small the input data range from 1 20 A gruel to +20'll; to the large range to theta~s/a, and a 0~12-g small range. Dynamic pressure control of this range is carried out by the operation of statistics coda 4 of Drawing 1 and the 1st/figure. 1st/of statistics coder figure In the 1st/figure, statistics coda 4 comprises coda 82 and transmitting machine Noshifa 83. output bus 13t- receiving coda from a coefficient processor -- the output tEk store of the coefficient difference in coefficient register 102 -- it carries out. Zero detection igl way 101 detects that the signal level on bus 13 became zero, and controls selector 109. Selector 109 chooses between coefficient register 102 and the contents of run length counter 103. Whenever zero are stored in register 102, the increment of the counter 103 is carried out, and run length counter 103 is chosen by selector 109 whenever null detector 101 detects zero. If null detector 101 detects a non-zero condition, selector 109 will choose a nonzero value from coefficient register 102, and will reset run length counter 103. Register 102 or the selection value from counter 103 is inputted into the Hofmann table 111. Table 111 gives the output of 2theta A bit to the Hofmann table register 112. 20 A bit output from this table 111 -- field tM of t bit -- although carried out, this field specifies the number of bits in the code between theta~/6, and the code of theta~76 bit length. Ghit chord length is inputted into format Controller [113, and theta~76 bit code is inputted into selector 110. The Hoffmann table 111 has only the output latched to register 112, when null detector 101 indicates that the non-zero output is latched to register 102. Register 112 is made to carry out the Datake latch of the number from the maximum t'& count or register 102 from 2 Rent counter 103 to Prompt. If run length coding is used for recognizing the long string of zero, the data output from table 111 will decrease sharply. While the coefficient difference value has arisen on each CLK cycle bus 13, the nonzero value of the data on bus 13 is answered, and only the output from register 112 arises. However, an operation of coefficient processor 3 of Drawing 9 will change many small difference values, i.e., a value also with small Threshold Tcj Si, into a zero value. Since the normalization by Nf decreases the size of the coefficient difference in register 102, the chord length from table 111 also decreases. Therefore, when both coefficient processor 3 and statistics coda 4 are operated, they are 112t of registers. - The data volume outputted by passing decreases sharply. Selector 110 under control of the format control device of a /second bit gathers data in /second bit length, and sends it to format register 114. Many other data fields where the data encoded from register 112 should transmit selector 110 to a receiver in addition are chosen. Generally this data field is the /second bit synchronization code (5YNC) from synchronous generator 104, Block Code en'(EOB) from the end of block generator 105, Threshold value Tc in register 106 received from coefficient processor control device 81 of Drawing 9, Normalization factor Nf stored in register 107 from coefficient processor control device 81 of Drawing 9 and the number CT of coefficient tables in register 108 from Grinding Cove Bath 46 from the block analyzer of a Fly figure are comprised. Transmission data has the next table H and 7 Omatsu of ■. Table: l 5ync Hdr NF BOB / B-2-- -..... B(X-/) Svnc (/&) (+) (Da] (/second) -- Table ■ Bx:()bPokex, x""theta / ...... (Harvest - /)] -- generally X is /theta2 person here. C, AthetaA/ AJ A3 ROA'l R/ A5 P2 -- Rm An EOB (0 dish of -2 A-prediction factor table) Ao - Size code of intra-frame DC value Ai=i==o and n RJ ""J ""O! End format control device 113 of the run length EOB = block code of m is backed to all the data tables H and the format of 1H, and is sent to Formats register 114. When format register 114 is full, data is transmitted to buffer memory 116 of a transmitting machine. -several examples -- this buffer memory -- comprise 76 Bitso and - Do boiled 3.2 116 ri. Buffer 116 is filled with the asynchronous rate at which data is generated by control of buffer control device 115, 76 A bit are emptied by buffer memory 116 at each 7:00, and this bit is sent to average in-series register 117 by/for 76 minutes of the synchronous data rate of the output of bit serial transmission bus 15. The degree of fullness of a buffer memory is suitably determined by buffer control device 115, and displays the degree of fullness of a buffer on attitude bus Bo. To United States patent Hiroshi, 3thetalambda, and No. 775, one method of generating a buffer fullness degree signal is indicated. The buffer fullness signal on bus 8o is inputted into coefficient processor Controller @81 of Drawing 9, and it, and it changes it. [ coefficient Threshold / normalization-7-actor-Nf(ing) and/or ] If these factors are changed, the rate which outputs data from the Hochman table register 112 of the 1st/figure will be influenced directly. Decoder 6 in receiver 25 of Drawing 1 is shown in Drawing 73 in detail. This decoder contains synchronous detector 89 which detects the synchronous pattern in each frame of the data sent from a transmitting machine. The Hofmann decoder 90 carries out Hoff mandator tray Tecode of the detected synchronized signal, and it makes it possible to store this in buffer 91. the data from buffer 91 -- the inside of reverse normalization device 93 -- reverse -- normalization -- !. It reverse-Threshold within Si and unit 156. synchronous detector 89 contains direct / parallel register 151 which stores each bit synchronizing with /bit people Calah Inn 17 top -- register 151 -- as the input to comparator 152 -- the output of /6 and a throat -- all -- it generates. The input of another side to comparator 152 is a thing from 1mJ M register 153, and register 153 stores the same /6 bit-synchronization code that occurs from synchronous generator 1414 of the 1st/figure. If comparator 152 detects that the pattern stored in register 151 is the same as the synchronous code from register 153, a synchronized signal output will be generated on line 154. Synchronization No. 18 on line 154 synchronizes with the timing in timing controller 129 with a CLK input. Timing controller side 129 comprises the same counter chain as Drawing 72, and other conventional control devices which the data in a receiver Unpacking all controls. Pop 7 Monday coater 90 is a bit in-series decoder, and this acts as Anh Buck of the data from a format which was mentioned above with the format of Table m in response to the input data of the /bit from transmission line 17. It acts to buffer memory 91 as Anh Buck, and fc Dede is stored. The information output from the Hofmann decoder 90 is stored in buffer memory 91/'i, general Ku, and 32 in x? bit. The information which acted as Anh Buck from buffer 91 is outputted through bus 96 of 7 A bit, and is inputted into various registers. A sign bit is stored in register 100', and zero dish is stored in prediction register 108' in the number of tables. Normalization factor Nf is stored in register 107', and coefficient Threshold Tc is stored in register 106'. If run-length-count, t zero counter 103' is stored, register 112' stores g bit conversion factor difference and a non-zero count is stored in zero counter register 103', the output from register 112 will serve as zero. Counter 103' is Decrement(ed) for every cLK time. Whenever zero counter 103' is Decrement(ed) by 0, the gate of the coefficient from register 112' is carried out as an input to multiplier 93. Reverse normalization factor Nfi hangs multiplier 93 on a coefficient, and it generates the output of//bit into bus 155. Reverse Threshold table 156 adds Threshold Tc to all the nonzero values on bus 155, and generates a reverse processing signal from table 156 to// bit outgoing end. Although the sign bit from register ioo' is combined with// A hit output from table 156 and a code signal is generated on /2 bit bus 2o, this signal displays received coefficient difference No. 16. in the 1st left figure -- the 3rd person of Drawing 74 -- a connoisseur -- coefficient generator 51 is shown in detail. One input signal to adding machine 128 and the input signal of another side to Si and an adding machine of the 72-bit signal on bus 2o are the predicted values by which the conversion factor on bus 158 was reconstructed. The value on bus 158 occurs by acting as frame memory 159 Hesse Doat of the output value from adding machine 128. Only a /frame cycle is delayed and the value from frame memory 159 outputs the value of eye watch (f-/) of 72 A bit on bus 16G. The value of eye watch (f-/) on bus 160 is multiplied with prediction 7 actor from prediction table 162. Prediction table 162 is the same as prediction table 10 of Drawing 7, it is addressed like [ control of address control unit 69' ] Yo Si, and address control unit 69' is similar to address control unit 169 in prediction device 43 of Drawing 7. This recovery value on Ah Si and bus 53 is inputted into reverse 2D conversion processor 52 of Drawing 711 with the recovery value of a conversion domain coefficient, and the output on bus 53 generates the recovery digital image signal on output bus 21. If it is a person skilled in the art although the present invention illustrated and being explained with reference to the desirable example above, naturally a design variation can be carried out, without deviating from Religious belief and the range of the present invention.

[Brief Description of the Drawings]

From a transmitting machine and a receiver to Formation Si by whom Drawing 1 was connected to each-other Transmission line It is a figure showing the image processing system using the conversion coding method concerning the present invention, Drawing 2 is a three-dimensional conversion processor diagram used within the transmitting machine of the system of Drawing 1, Drawing 3 is an abbreviated block diagram of the 3rd-dimensional coefficient generator used within 3D conversion processor of Drawing 2, and is Ah Si, Drawing 9 is a block analyzer's abbreviated block diagram used within the 3rd-dimensional coefficient generator of Drawing 3, and is Ah Si, Drawing 6 shows the contents of the range table which uses The within the block analyzer of a figure by Drawing S showing the contents of the table used within the block analyzer of Drawing 9, and it is Drawing 7, Predictor which constitutes some 3rd-dimensional coefficient generators is shown, and Drawing S shows the contents of the coefficient table which constitutes some prediction machines of Drawing 7, The figure showing the operation of the coefficient processor with which Drawing 7 is an abbreviated diagram of statistics Cog and a coefficient processor which are some transmitting machines of the system of Drawing 1, and Drawing 70 was indicated to be all over the 9th figure, and the 1st/figure show statistics Cog shown in Drawing 7 which is with some transmitting machines in the system of Drawing 1, Drawing 1st [ the ] and 2 shows the timing controller which constitutes some transmitting control devices in the transmitting machine of the system of Drawing 1, Drawing 13 is an abbreviated block diagram of the decoder used within the receiver of the system of Drawing 1, and is Ah Si, The 1st Da figure is a block diagram of the reverse three-dimensional (3D) processor used within the receiver of the system of Drawing 1, the 3rd dimension from which Drawing 75 constitutes a part of reverse 3D processor of Drawing 3 -- a connoisseur -- it is an abbreviated block diagram of a generator -- digital image frame of plurality [ Drawing / 76 ] 0./, ...... Ahupsilon and Drawing 77 are schematic illustrations changed into space-#4 No. 1r and 2D resolution picture signal corresponding to each block in the frame of J / 6 figure in the figure showing f and the arrangement which carried out block trap composition, and the 7thg figure is Sotozono showing the f-th frame and the cosine transform coefficient of a B number block. 1 ... transmitting machine control device, 2 ... 3D conversion processor [ ... Decoder 7 / ... Reverse 3D grotesque Sette 8 / ... Transmission line ] 3 ... A coefficient processor, 4 ... Statistics Cog 5 ... A receiver control device, 6 Clean room (with no change by the contents) FIG of a drawing, -7 FIG, -10 FIG, -12 FIG, -16 FIG, -17FIG, -18 A page [ 1st ] continuation Zero shot Ming person Ralph Emerging Nichols United States of America California 95030 Los Gatos long wood drive 15970 Zero shot Ming person Albert Edwards racket American California State 94086 Sunnyvale lily avenue 1066 Written amendment Method 59.7.25 Showa Date 1, display of an incident The Showa 59 patent application No. [ two ] 604.89, title of an invention An image signal disposal method, device 3, those that do amendment Relation with an incident Applicant 4 Representative 5, date of an order for amendment Fair copy of June 26, Showa 59 drawing (with no change by the contents).