JPH02143679A - Picture encoding method - Google Patents

Abstract

PURPOSE:To improve the encoding efficiency by revising the location of a referenced picture element and number of picture elements in response to the picture size in a picture element encoding method applying dynamic encoding for each state decided by the referenced picture element. CONSTITUTION:A reference picture element decision section 3 forming a CPU together with a state discrimination section 4, an arithmetic encoding section 5 and a memory 6 for arithmetic operation result or the like decides the reference picture element number and the position of picture element in response to the picture element size by a size detection sensor and a picture size input section 2. Then a priority symbol and a skew value or the like at the arithmetic operation in the encoding section 5 are decided via a state control section 4 based on a data corresponding to the decision picture element in the picture data read from the input section 1. Thus, the symbol number assigned to each state in response to the picture size from the output section 7 is sufficiently large and the skew value is converged in each state and a picture code encoded efficiently is outputted.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an image encoding method in image communications and image databases.

[Conventional technology]

Conventional image encoding methods include the MH method, MR method, MMR method, etc. that use a succession of black and white images (run length), but when halftone images are encoded using these methods, images with short run lengths are created. The disadvantage of this method is that it requires more data than the original image.

Arithmetic code (G.

G, Langdon and J, J, R15sane
n ”Compression of Black-Wh
ite Images with Arithmeti
cCoding, IEEE Trans, Commu
n, C0M-29°1981). This method is known as a highly efficient encoding method because the encoding efficiency is determined only by the probability of occurrence of white runs and black runs, regardless of the run lengths.

[Problem that the invention is trying to solve] However,
In the above conventional example, the probability of appearance of an inferior symbol is required at the time of encoding, so for example, it is necessary to first pre-scan the image to be encoded, take statistics, and find the probability of occurrence of an inferior symbol according to the image. Ta.

In addition, regardless of the image size for dynamic arithmetic naturalization that does not require Briscan, a phenomenon occurs in which encoding efficiency decreases for small image sizes because a fixed number of pixels are referenced. .

That is, when the image size is small and the number of reference pixels is large, the probability of appearance of each state decreases. Therefore, the encoding ends before the initial value converges to the optimal value, and the encoding efficiency decreases.

[Means to solve the problem]

The present invention has been made in view of the above points, and is an image encoding method that performs dynamic encoding for each state determined by reference pixels during predictive encoding. This provides an image encoding method that detects and changes the reference pixel position and reference pixel number according to the image size.This provides an image encoding method that changes the reference pixel position and number of reference pixels according to the image size. The number of reference pixels can be changed dynamically according to the image size.

〔Example〕

The present invention will be explained below using preferred embodiments.

FIG. 1 is a block diagram showing the configuration of an image processing apparatus to which the present invention is applied.

2 is an image size input section for inputting the image size (number of data) based on the operation of keys for size input or the output of a sensor for size detection; 3 is input from the image size input section 2; 4 is a reference pixel number determination unit that determines the number of reference pixels based on the image size value, 4 is an input image state determination unit assuming a Markov model, and 5 is arithmetic coding for image data. The arithmetic encoding section, 6 is a memory for storing the determined state and encoded data, and 7 is a code data output section for outputting the encoded data.

FIG. 2 shows the reference pixel determining section 3 and state discriminating section '4 shown in FIG. 1.
, is a configuration diagram when the arithmetic encoding unit 5 is realized by a microprocessor.

8 is an image data input unit corresponding to the image data input unit l shown in the first diagram; 9 is a ROM in which a program for encoding image data is stored; A CPU 511 whose main element is a microprocessor for processing is a RAM that stores data processed by CPUl0, and coded data 12 corresponds to the coded data output unit 7 shown in the first diagram that outputs the data processed by CPUl0 as coded data. Output section, 13
is an image size input section corresponding to the image size input section 2 shown in the first diagram. The processing performed by the reference pixel determining unit 3, state determining unit 4, and arithmetic encoding unit 5 in FIG.
It is configured to do this.

FIG. 3 is a flowchart showing the flow of processing in CPU10 in FIG.

First, when the program starts, data indicating the size of the image to be encoded is input from the image size input section 13 in step Sl. In CPU10, step S2
The number of reference pixels is determined as shown in Table 1 according to the image size (number of pixels) input in . For example, the total number of pixels is 1
If it is 00000, the number of reference pixels is 7, and the pixels numbered 1 to 7 shown in FIG. 4 are referred to.

Here, * indicates a pixel of interest.

Next, in step S3, the dominant symbol (MPS), skew value (Q), 2-pixel counter CN, and inferior symbol counter LCN in the arithmetic code.

Initialize the dominant section counter CNL. For example, if we consider 7 pixel references, there are 27 dominant symbols and 27 skew values. Here, 27
dominant symbols, skew value 2 pixel counter,
The inferior symbol counter and the superior section counter, respectively.
MPSt, Qt, CNt, LCNt.

It will be expressed as CNLt (t=o, 1...127).

Next, in step S4, image data is input from the image data input section 8. Then, in step S5, the pixel of interest
By the reference pixel already determined for (11j), 1
The state is classified in 28 (=2') ways. As a classification method, consider that the value of a pixel with a dot is 1 and the value of a pixel without a dot is 0, and the state St of the pixel of interest X (ITJ) is 5t=x (i, j) +2'- x (i, j-1
)+2”-x (i+1.j-1)+2”-x(i-
1,j-1)+2'・x (i-2,j-1)+2"x
(i-1, j-2)+2' -x (i+1. j-2
) is expressed numerically. Similarly, if 16 pixels are referenced, 5t=x (i-1, j)+2・x (i, j-1)
+-=+216 x (i-3, j-1).

Next, in step S6, the MPSt in the classified state.

The pixel of interest is arithmetic encoded using Qt, and the encoded data is stored in RAM II. Furthermore, if it is necessary to update MPSt and Qt, these values are updated. In this way,
Input pixel data is sequentially encoded.

Finally, in step S7, it is determined whether all of the image data has been encoded. If encoding has not been completed, the next pixel is input, and the state fraction and arithmetic encoding are repeated. If all the data has been encoded, in step S8, the encoded data stored in RAM II is output to the image data output device 12, and the process ends. At this time,
If the number of reference pixels and the reference pixel position are added as header information of encoded data, this encoded data can be uniquely decoded.

FIG. 5 is a detailed flowchart of the dynamic arithmetic encoding operation in step S6 of FIG. Arithmetic code predicts the value of the pixel of interest from surrounding pixels, and the symbol when the predictions match is the dominant symbol (1), the symbol when the predictions do not match is the inferior symbol (0), and the probability of occurrence of the inferior symbol is P This information is used for encoding.

That is, the binary arithmetic code for the code sequence S is C(S),
If the supplementary amount is A (S), then depending on whether the pixel of interest is a dominant symbol or not, encoding is proceeded by the arithmetic operation of A (null) = O111・1. Note that P (S):
By approximating it to 2-Q(S), multiplication can be accomplished by only shifting binary numbers. Q is called a skew value, and by changing this parameter, arithmetic codes can be used dynamically.

In step Sll, set formula (A), and then
In step S12, the pixel of interest is the dominant symbol MPS
It is determined whether it is the same as t. If the pixel of interest is the same as the dominant symbol, C(SO)←C(S) is set in step S13, and if the pixel of interest is the inferior symbol, C(Sl
)←C(S) +A (So), and the count LCNt of inferior symbols is incremented by 1.

Next, in step S15, the 2-pixel counter CNt is set to 2.
If it is 2, MPSt and Q are updated in step S16, and if CNt is 1, CNt
Update to 2. As a result, the skew value Q and dominant symbol MPS are dynamically updated for every two pixels based on previously encoded data.

To decode the data encoded as above, the binary signal sequence S=S' x S'' is restored, and when S' is restored, C(S) and C(S'') + A (S' O) are compared. ,
It is decoded as 1 when C(S) >C(S') +A (S' 0), and as 0 otherwise.

FIG. 6 shows MPSt in step S16 of FIG.

It is a detailed flowchart of update processing of Qt.

'First, in step S21, if there are no inferior symbols in the two pixels, then in step S22, the dominant symbol (
Dominant section counter CN indicating whether two sections are consecutive
Increment Lt by 1. Further, 1 is substituted into the inferior symbol counter LCNt for later calculation. When there is one inferior symbol and when there are two inferior symbols, the symbols are substituted into the respective steps from step S26. Here, [ ] is a Gaussian symbol and represents the largest integer less than or equal to the value in the symbol.

Then, in step S29, it is determined whether CNLt is 1 or not, and when CNLt is 1, the probability of occurrence of the inferior symbol is greater than that of the superior symbol.
At 30, invert MPSt (0 if l, 1 if 0)
do.

Next, in step S23, the probability of occurrence of an inferior symbol is calculated from LCNt and CNLt and substituted into Qt. Then, step S24. If Qt exceeds 1 in S31, Qt is set to 1, and in step S25, the count LCNt of inferior symbols is returned to 0, and the 2-pixel counter C
Reset Nt to 1.

As mentioned above, when dynamic encoding is performed, a considerable number of symbols are required for the value of Qt to converge, so
By setting the number of reference pixels to match the image size, a sufficient number of symbols can be classified into one state.
Thereby, encoding efficiency can be improved.

Incidentally, as a method of performing dynamic encoding at high speed, it is possible to calculate in advance in step S23 of FIG. 6 and create a table as shown in Table 2. If it is made into a table, it can be easily made into hardware.

Table 2 Furthermore, it can be applied not only to read image data from the league but also to output data such as an image database. In this case, since the image size is often included as header information, the header information can be used to input the image size.

Furthermore, encoding efficiency can be improved by inputting the attributes of two images, for example, the distinction between a halftone image and a character image, and referring to pixel positions that match each attribute.

[Detailed description of the invention]

As explained above, by changing the reference pixel position and the number of reference pixels according to the image size, the number of symbols assigned to each state becomes a decimal. will improve.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of the present invention, FIG. 2 is a block diagram when a microprocessor is used in the present invention, FIG. 3 is an overall flowchart of the present invention, and FIG. 4 shows reference pixel positions. 5 is a flowchart of dynamic arithmetic encoding processing, and FIG. 6 is a flowchart of updating the dominant symbol skew value. is an image size input section.

Claims (1)

[Claims] In an image encoding method that performs dynamic encoding for each state determined by reference pixels during predictive encoding, the image size of the image to be encoded is input or detected, and the reference pixel position and reference are determined according to the image size. An image encoding method characterized by changing the number of pixels.