JPH01231582A - Method and device for hierarchical coding of picture data - Google Patents

Abstract

PURPOSE:To improve the transmission efficiency by transmitting repetitively the La data on an untransmitted block line after thinning said data every (m) block lines after the end of transmission of one screen and using the received La data at the reception side to interpolate the untransmitted block line for display. CONSTITUTION:In a hierarchical transmission mode, the La data (reference level data) is transmitted after thinning it every prescribed (m) block lines at the transmission side. Then the La data on an untransmitted block line is repetitively transmitted after thinning said data every (m) block lines after the end of transmission of one screen. At the reception side the received La data is used for interpolation and display of the untransmitted block line. Thus the La data can be transmitted for each block line by thinning the data every (m) block lines. Then it is possible to divide the first stage of the hierarchical transmission into (m) pieces of stage. As a result, the overall contents of a picture can be known early for selection of pictures. Then the picture data can be transmitted with high efficiency.

Description

[Detailed description of the invention] [Summary Image 1 (i) - An image in which data is compressed and transmitted using a hierarchical encoding method, and the transmitted data is received and reproduced] Regarding hierarchical encoding method and apparatus i:l,
The first invention aims at transmitting image data with efficiency Jz<1], and the first invention is a hierarchical encoding method for image data (transmitting, receiving and re-coupling the transmitted data.46 Hierarchical encoding of image data In the method, on the transmitting side, there are a step of dividing image data into blocks each consisting of a plurality of adjacent pixels, a step of thinning out the reference rubel data in the block to σ every predetermined block line and transmitting the same, and after completing transmission of one screen. Furthermore, it includes a step of thinning out the reference level data of the untransmitted block lines for each predetermined block line and transmitting the data.
On the side, it is configured to include a culm that interpolates untransmitted block lines using the transmitted reference level data.

The second invention is a hierarchical encoding device for image data that divides image data into blocks each consisting of a plurality of adjacent pixels, transmits reference rubel data for each pixel in the block as an initial stage, and reproduces the transmitted data. , a buffer 1 for storing one screen worth of reference level data thinned out for every predetermined block line in the block is connected to a first buffer memory located on the right by the number of lines; an arc selection section that selects data; a variable length code generation section that variable length encodes the data input from the buffer memory via the data selection section; and a variable length code generation section variable length encoded by the variable length code generation section. A second buffer memory for storing the received code is on the transmitting side, a basic level decoding section for decoding the received code to the reference level, and an image IK+ of the 11 screen receiving the output of the reference level decoding section.
The receiving side is equipped with an image memory for storing data and an address control unit that interpolates unpaid block lines based on the reference level data output from the bending reference level decoding unit and stores the interpolated block lines in the image memory. and configure.

[Field of Industrial Application] The present invention relates to a hierarchical encoding method for image data that compresses and transmits image data using a hierarchical encoding method, and receives and reproduces the transmitted data. METHODS AND APPARATUS.

The amount of information necessary to represent the image Y-ta increases by an order of magnitude compared to numerical data. This increase in the amount of information is remarkable in image data, especially in multi-value halftone images and color images. In order to accumulate such image data or transmit it at high speed and with high quality, it is necessary to encode tone information for each image with high efficiency.

In addition, in image information search and image communication,
When the information transmission speed is sufficiently fast, interactive and efficient searching requires hierarchical transmission that first displays a rough overall image and gradually improves the image quality until the user is satisfied.
progressive build-up) functionality is required. With this function, transmission of unnecessary images can be stopped at an early point, making it possible to obtain desired images with sufficient image quality.

[Prior Art] As a highly efficient compression method for image data that has a hierarchical transmission function, for example, a multi-gradation adaptive block coding method (
There is a preliminary draft of the 1986 National Conference of the Society of Image Electronics Engineers 6).

Next, the multi-gradation adaptive block coding method (GOnQral)
ized Block Truncation C
oding (hereinafter abbreviated as G 13 TC) will be explained. G13TC divides the image into blocks consisting of NXN pixels, and divides each pixel (X + a > into 21 (n =
0.1, 2. -·=) and quantizes with a level of 100 pixels, expresses the matrix level of each pixel in a bit plane format, and encodes gradation information and bit plane information.

Next, N=4. The case where n = 2 will be described in detail. FIG. 10 shows the GBTC algorithm. Each block consists of the difference between the maximum pixel level (MAXL) and minimum pixel level (MINL) within the block and the encoding parameter T1
．． It is classified into the following three encoding modes (Δ, B, C) according to T2 (TI < T2 ).

Mode A: When MAXL-MINL≦T1, the pixels in the block are quantized to Lebel (Po).

Mo1'B: TI < MAXL-M rNL≦l-
For z, the pixels within the block are quantized to two levels (P+, P2).

Mode C: M A X L-M I N L;・
In the case of 1'2, the pixels within the block are grouped into four equally spaced levels (01 to Q4).

The quantization level is lQu level 1-a of the block.

Level interval 1d and level designation signal for each pixel (φ1) +
J, (φ2), J. Average value processing is performed using AVE (
), each value required for encoding is calculated as follows.

Mode A: Po”'AVE (X t J)
-La(φ1)tJ=O, (φ2)+J=0 (for all i, j) Mode B: Pr-AVE (X + = ≧(MA
XL+MINl )/2) P2-AVE (x I J 〈(MAX I-+Mr
NL>/2) la-(Pi +P2)/2゜Ld=P+P2 (φ+)tJ=o(Xtj≧(MΔXL+MINL)/
2) Ku(φI>+j””1(XiJri(MAXL.

+MINL)/2) (φ2):==O (for all i, j) E-doC:Qt =AVE (X IJ≧(3MAXl)
-+MINi)/4) Q4 -AVE (x H; ≦ (MAXI-
13MINL)/4> La + (Q-+ 4-04), /2.

Ld -2 (Q+ -04)/3 Q2 - La + Ld / 'I.

Q3 = l a -1-d /'1(φ+)+
;=O, (φ2)l J −0for x (;
〉I-a +Ld/2(φ+)+J=
O, (φ2)+==1(La −) Ld
/2 > X i j ≧ l-a)! (φ1)1. - Noboru, (φ2) 1==0 (La > In the case of 1d/2, when obtaining a decoded image of even higher quality in GBTC, the difference signal between the normally used decoded image and the original image is converted into matrix and variable length coded.

The GBTC performs highly efficient data compression as described above, but also has a hierarchical transmission function for image retrieval on a display and a sequential transmission function for hard copy devices without frame memory. As shown in Figure 11,
In hierarchical transmission, first, only La is transmitted, then Ld
, φ1. Additional transmission is performed in the order of φ2° difference signals to improve image quality from consistent images to fine images. In contrast, in conventional sequential transmission, data is transmitted in units of one block line.

[Problems to be Solved by the Invention] In the prior art, when an image is divided into blocks each consisting of a plurality of pixels and transmitted hierarchically, only the reference level is transmitted for each block in the first stage, and Lebel's roughness per block is transmitted. It was displaying an image.

When performing an image search, the first stage is usually to display many types of images side by side in small sizes, and then quickly select the images and do a quick look-up.
It is desirable to be able to see the entire image at an early point in time. However, in the conventional technology, reference level data (GI
In 3TC, the above-mentioned l-a data) is encoded into l〕1'' CM code and transmitted sequentially for each block line, so there was a drawback that image breaks could not be seen until one screen worth of reference level data was played back. .

The present invention has been made in view of these points,
An image app that can efficiently transmit image data
Evening l! ! ! The purpose of the present invention is to provide an i-layer plagiarism method and apparatus.

[Means for Solving the Problems] Fig. 1 is a flowchart showing the principle of the method of the present invention (first invention), and Fig. 2 is an original B1 of the apparatus of the present invention (second invention).
+ It is a block diagram.

The method of the present invention first includes, on the transmitting side, a step of dividing image data into 4r blocks from a plurality of adjacent pixels, and a step of dividing the image data into 4r blocks from a plurality of adjacent pixels, and performing lit q! The second step is to thin out the level data to 0 every predetermined block line and transmit it, and after the completion of one screen transmission, further thin out the reference level data of the untransmitted block lines to every predetermined block line and transmit it. and a step of interpolating untransmitted block lines using the transmitted reference level data on the receiving side.

Next, the configuration of the device of the present invention will be explained in 3゜Second section.
In the figure, 1 is a first buffer memory having barafuns for storing gt!'It level data for one screen whose contents are thinned out every predetermined block line, and 2 is the buffer memory. 3 is a variable-length code generator that variable-length encodes the data input from buffer 1 through data selector 2; 4 is variable-length code generator 3; This is the second buffer memory that stores the variable-length coded code.The above is the configuration of the transmitting side.

Reference numeral 5 denotes a reference level decoding unit that decodes the code 8 received from the transmitting side to the reference level; 6 an image memory that receives the output of the reference level decoder 5 and stores image data for one screen;
Reference numeral 7 denotes an address control unit that interpolates untransmitted block lines based on the reference level data outputted from the reference level decoding unit 5 and writes the interpolated block lines into the image memory 6. The above is the configuration of the receiver 5 side.

[Operation] The buffer D1 contains buffers for storing level data for one screen, which is thinned out for each predetermined block line within the block. The reference level data stored in these buffers is selected by the data selection section 2, subjected to variable length encoding by the variable length code generation section 3, and then transmitted via the baratu no meshiri 4.1 After the reference level data for the screen has been transmitted, the next 19 level data stored in buffer 1 is transmitted in the same manner as necessary.

On the other hand, on the receiving side, the transmitted code is restored to reference level data by a reference level decoding section 5.

The restored base piece level data is stored in the image memory 6 in units of one screen. The same reference level data is also entered in the address control section 7, and the address control section 7 interpolates untransmitted block lines based on this reference level data and writes it into the image medium 6.

In this way, the present invention provides L
The a data (reference level data) is thinned out every m block lines and transmitted, and after one screen transmission is completed, the (-a data) of the untransmitted block lines is thinned out every m block lines and transmitted repeatedly. The LA data transmitted on the receiving side is used to interpolate and display untransmitted block lines.By performing thinning transmission every n1 block lines, the LA data is divided into 1 block line. The first stage of hierarchical transmission, which is transmitted sequentially for each step, can be divided into m (11 stages).

According to the present invention, since the contents of the entire image can be known at an early stage and images can be selectively selected, efficient transmission can be performed.

FIG. 3 shows the principle of hierarchical transmission of the present invention in the case where m=4. The transmission order number p-q in FIG. 3 indicates that the qth block line data (-a) is transmitted from the transmitting side in the first stage. However, M is the number of block lines in the image. .la of each block line
Data is considered from the first block line (p -1)
It will be transmitted xM//ILqth.

Here, the 4th n→-1, 4n +2.4. n+3.4
n → ~4 (n = 0.1.2...) block line La data as Ia (1) and l-a, respectively
<2>.

La (3), La (4) and later, Chapter 1.2.
3゜4 stages, l a data is La (1), L, a
(3), La (2), L, and a (4) are transmitted in this order.

On the other hand, on the receiving side, the interpolation shown in FIG. 4 is performed on the block lines transmitted at each stage to create a restored image. Therefore, in the first stage, the first
．． Fifth. 9th. . . . The la data of the block line is transmitted, and the receiving side replaces it with this la data every four block lines. Next, in the second stage, the third stage. 7th. 1st
1. ...Block line 1. a data is transmitted, and on the receiving side, the same block line and the third .a data are transmitted. 4th. 8th. 1st
2. ...Replace the block line with this la data. In the third stage, the second stage. 6th. 10th. ...The la data of the block line is transmitted, and the receiving side replaces the same block line with this la data. Similarly, in the fourth stage, 4. 8th. 12th. ...The [a data of the block line is transmitted, and the receiving side replaces the same block line with this la data.

In other words, the input image is completely reproduced only after it is transmitted up to the fourth stage. L transmitted at each stage
Block lines for the a data are sequentially selected at positions that divide the untransmitted block lines into two so that there are as few block lines as possible to interpolate in the hierarchically restored image and the image quality of intermediate stages is as good as possible.

[Example] Hereinafter, examples of the method and apparatus of the present invention will be described in detail with reference to the drawings.

5 and 6 are block diagrams showing one embodiment of the present invention, in which FIG. 5 shows the transmitting side and FIG. 6 shows the receiving side, respectively. The same parts as in FIG. 2 are not shown with the same numbers -15 to 舅. First, the transmitting side in FIG. 5 will be explained.

In the figure, multi-value image data input from a terminal 11 is stored in a buffer memory 1 for one block line, and then multi-value image data xiJ is read out from the buffer memory 1 one block at a time.

That is, as shown in FIG. 7(b), the original image is divided into blocks each consisting of 4×4 pixels, and one block (not shown in FIG. 7(a)) of multivalued image data xiJ are read out in sequence.

The gradation change amount detection unit 13 detects the input multivalued image data XI.
From J to maximum gradation MAXL and minimum gradation Ml (T+ <T2
) and decide whether to use the one shown in FIG. 8 or two or more. Further, when it is determined that the number of representative gradations 1 G- is 2 or more, the difference value is compared with the second encoding parameter T2, and it is determined whether the number of representative gradations is 2 or 4 or more. For D1, by examining the spatial distribution of gradations within the block and adaptively changing the value of the encoding parameter, it is possible to improve the reproducibility of image edges and suppress noise.

Next, the representative gradation value determining unit 15 determines the representative gradation within the block using nonlinear quantization and linear quantization, according to the obtained number of representative gradations. In other words, representative 1Il! The i key is “
1'', the average value within the block is determined, and the reference level generator 16 sets this value as the standard value 1-a, and also uses the resolution component φ1. As shown in FIG. 9(a) as φ2,
Assign 0 to all pixels. Also, the number of representative gradations is “2”.
”, the representative gradation value determination unit 15 first calculates (maximum gradation MAXL + minimum gradation M I N L )÷2=LM
is determined as an intermediate value, and then the average value of the gradation values of each pixel having a gradation value within the range of 1 to maximum gradation value MAXL is obtained, and in the same way, the intermediate value km L M to the minimum gradation value The average value of the range M I N L is determined. Then,
The average value of each is taken as the representative gradation P+, P2, and 3g
The average value (P
I + P2)/2 - (l lj (1-a) is calculated, and the difference value generating section 18 calculates the difference value 1 between both average values.

! seek. The resolution component is shown in Figure 9(b),
φ2 becomes O in all pixels, and if the representative gradation is P+, φ1-02, and if the representative gradation is 22, it becomes φ1-1.

Furthermore, when the number of representative gradations is "4", the representative gradation value determination unit 15 first determines the maximum gradation MAXL and the minimum gradation MINL.
The space between the two is divided into /1 equal parts, the maximum gradation MΔ is divided equally, and two equal values between the average value and the average value are obtained as the average value. Multiply the difference between the average values by a predetermined coefficient ⅔ ('J) and obtain the value.

On the other hand, as shown in Figure 9 (C), the resolution component is φ1-O1φ2=0 when the number of representative gradations is Q+, and φ+-0 when the number of representative gradations is 02. φn・−1 When the number of representative gradations is 03, φl−1. φ2−−〇When the number of representative gradations is 04, φ+=1. It becomes φ2-1.

By performing the above processing, the representative gradation, reference value, and difference value within the block are calculated.

In the present invention, four buffer memories 20, 21, 22, and 23 are connected after the reference level generator 16, and the reference value 1a is repeated every four block lines and written into each buffer memory one block line at a time. In other words, for the reference value "a,"
1st. Second. Third. The data of the fourth block line are stored in buffer memories 20, 21, 22, and 23, respectively. Similarly, the 11th. n +1, 4th n+2. Fourth
n +3, 4th n(-4(n = 1.2, 3...
) The la data of the block lines are repeatedly stored in the buffer memories 20, 21, 22, and 23, respectively. The same applies below.

The code generating section 27 generates the difference value 1- obtained by the difference value generating section 18.
d is variable-length coded, and this variable-length code is stored in the buffer memory 28. The gradation value storage unit 17 stores the representative gradation of the block of interest determined by the representative gradation value determination unit 15.

The comparator 19 compares one multi-valued image data
are read out again pixel by pixel and compared with the representative gradation stored in the gradation value storage section 17, and the resolution component φ1. Outputs φ2. The resolution components φl and φ2 are stored in resolution component storage buffer memories 29a and 29b, respectively. Then, the contents of the buffer memories 29a and 29b are sent to the encoder 30.
The signals are read out, subjected to a known MMR encoding process to suppress redundancy, and then stored in the resolution component storage buffer memories 31a and 31. Code generator 25.27
The encoder 30 and the encoder 30 correspond to the variable length code generator 3 in FIG. 2, and the circuits existing between the buffer memory 1 and these code generators correspond to the data selector 2 in FIG.

On the other hand, the reference values La (1), La (2), and La in the buffer memories 20, 22, 21, and 23, respectively
(3), L-a (4) are multiple coolies 2
4, the data are selected and read out from the buffer memories 20.21 and 22.23 in this order, subjected to variable length encoding by the code generator 25, and stored in the buffer memory 2G. Finally, when the processing for one screen is completed, 1-a, 1-d.

φ1. The encoded signal of φ2 is sent via the multiplexer 32 to the respective buffer memories 26, 28, 3Ia, 31b.
(corresponding to the buffer 7 memory 4 in FIG. 2) are sequentially selected and read out, and sent to the receiving side in this order. Therefore, in the present invention, the reference value La is encoded as data for every four block lines, and is transmitted for four screens.

Next, the receiving side encoding circuit shown in FIG. 6 will be explained.

Terminal/10, 1, First, L every 4 block lines
The encoded signal of J is input to the reference level decoding section 41.
: is decoded. The decoded lJ reference level l-a is controlled by the address control unit 7 in each of the first to fourth stages, and the image memory 6 for one screen is transferred to the next J: sea urchin tears by four block lines. Change.

In the 5th and 1st to 4th stages, the 11th. n
-1-1 block line la data 1. -a(1)
, 18 data l of the 411th + 3rd block line, -
a (3).

11th n + 2 block 1 kk line la data la
(2) The ia data Ja(4) of the 4nth→-4 block line is decoded. The contents of the image memory for each block line are as shown in FIG. 4 by interpolating the untransferred block lines.

As the stage moves from the first stage to the higher stage, the block line thinned out phased image gradually becomes finer.

In the fourth stage, each block of the image becomes an image expressed in one gradation. As a result, for up to the fourth stage, the image number S1 is stored in the image memory 6 as S+ =la for the pixels of the [Note 1] block. The fifth stage and subsequent stages are similar to the conventional technology, in which the difference value 1d is decoded, bit plane information φ1 and φ2 are sequentially decoded, and the L
An image is created that displays up to 2 gradations per block from d and φ1 (fifth stage), and an image that displays up to 4 gradations per block from φ2 of Ld.

The images hierarchically restored on the image screen 6 are stored on the CRT 51.
will be displayed. Images from coarse to fine are sequentially restored on the CRT 51 by the above-described image restoration.

The operator 53 observes the image of CR'T' 42 and determines that it is an unnecessary image. When such a determination is made, the keyboard 52 instructs the transmitting side through the terminal 50 to stop the transmission, and requests transmission of the code of the next image. If the image is necessary, continue restoring it until the image quality is sufficient. In the present invention, the first stage of the conventional method is subdivided into four stages, but if the image is a commonly used 512 x 512 pixel size, it is possible to know the content of the image even in the subdivided intermediate stage of the present invention. Now, you can perform image searches more efficiently than before.

In the figure, 41t-1 (level decoding section 41 in FIG. 6 corresponds to reference level decoding section 5 in FIG. 2).

- In the above description, the case where the present invention is applied when performing hierarchical decoding using block head encoding has been described. but,
The present invention is not limited to GBTC, but can be similarly applied to quick lookup stages of other systems in which an image is divided into blocks and reference levels are transmitted in hierarchical decoding.

[Effects of the Invention] As described in detail above, according to the present invention, when transmitting image data using hierarchical transmission, the transmitting side thins out 3a data every m block lines and transmits the data, After one screen transmission is completed, the la of the untransmitted block line is
The image data is thinned out for each line and transmitted repeatedly, and the untransmitted lock line is interpolated and displayed using the IA data transmitted on the receiving side. It is possible to provide a method and apparatus for hierarchically encoding image data, which can efficiently transmit image data.

[Brief explanation of the drawing]

FIG. 1 is a flowchart showing the principle of the method of the present invention, FIG. 2 is a block diagram of the principle of the apparatus of the present invention, FIG. 3 is an explanatory diagram of hierarchical transmission of reference level data [a], and FIG. 5 and 6 are block diagrams showing an embodiment of the present invention. FIG. 7 is an explanatory diagram of block division of a multivalued image. Fig. 8 is an explanatory diagram of the amount of gradation change and the number of representative Wi keys, Fig. 9 is an explanatory diagram of the resolution component, Fig. 10 is a diagram showing an overview of the GBTC algorithm, and Fig. 11 is a diagram showing hierarchical transmission by GBTC and sequential It is an explanatory diagram of transmission. In FIG. 2, 1.4 is a buffer memory, 2 is a data selection section, 3 is a variable length code generation section, 5 is a reference level decoding section, 6 is an image memory, and 7 is an address control section.

Claims (3)

[Claims] (1) In a hierarchical encoding method for image data in which image data is transmitted using a hierarchical transmission method, and the transmitted data is received and reproduced, the transmitting side divides the image data into blocks each consisting of a plurality of adjacent pixels. process (step [1]), a process (step [2]) of thinning out the reference level data in the block for each predetermined block line and transmitting it, and after one screen transmission is completed, the reference level data of the untransmitted block lines is further reduced. The process includes a step of thinning out data for each predetermined block line and transmitting the data (step [3]), and a step of interpolating untransmitted block lines on the receiving side using the transmitted reference level data (step [4]). ) A method for hierarchically encoding image data. (2) The method for hierarchically encoding image data according to claim 1, characterized in that a multi-gradation adaptive block coding (GBTC) method is used as the hierarchical transmission method. (3) A hierarchical encoding device for image data that divides image data into blocks consisting of a plurality of adjacent pixels, transmits the reference level data of each pixel in the block to an initial stage, and reproduces the transmitted data. , a first buffer memory (1) having buffers for storing one screen worth of reference level data thinned out for each predetermined block line within the block for the number of thinned block lines; and the first buffer memory (1). a data selection section (2) for selecting data from the buffer memory (1); a variable length code generation section (3) for variable length encoding data input from the buffer memory (1) via the data selection section (2); a second buffer memory (4) that stores the code variable-length encoded by the variable-length code generator (3); and a reference level decoder (5) that decodes the received code to the reference level on the transmitting side. and an image memory (6) that receives the output of the reference level decoder (5) and stores one screen worth of image data, and a Image memory (6) by interpolating untransmitted block lines
1. A hierarchical encoding device for image data, characterized in that each receiving side is provided with an address control unit (7) for writing data into the image data.