JP2841197B2 - Method of compressing gradation image data - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a method of compressing gradation image data for compressing gradation image data such as medical images such as X-ray photographs and CT images.

BACKGROUND OF THE INVENTION With the advancement of digital technology, digitized gradation images are frequently stored, transmitted, and subjected to various digital image processings. However, a tone image has a larger amount of information than a binary image, and thus has a problem of a large amount of data when the tone image is digitized. In particular, in the case of medical images, the number of images and the number of bits required for each pixel when digitizing are enormous, for example, 4 million pixels and 8 to 10 bits in a chest X-ray photograph, which is very efficient in storing and transferring data. Is bad.

Therefore, today, a data compression technique for compressing enormous data of a gradation image including a medical image to make it compact has been spotlighted.

Data compression technologies can be broadly classified into lossless compression and irreversible compression.However, lossless compression can only achieve a low compression ratio of about 1/2 to 1/3, so a high compression rate of 1/5 or more can be obtained. Irreversible compression schemes, especially transform coding schemes, are attracting attention.

Transform coding is a lossy compression method that divides the entire image into small blocks, performs orthogonal transform on a block basis, and quantizes and encodes the transform coefficients obtained thereby. This is the most suitable compression method for compression.

In transform coding, the distribution of the AC component of the transform coefficient is known to approximate a Gaussian distribution having a peak at zero, and the quantization of the transform coefficient having such a distribution includes zero in the quantization determination level. When classified into Mid-riser type quantization and Mid-trace type quantization including a zero in the quantization output level, it is reported that Mid-trace type quantization has less random noise in a block and is more preferable. Have been.

In order to obtain a high compression rate in transform coding, there are a method of cutting high frequencies and a method of increasing the quantization width. In the case of cutting a high frequency, there is a property that blur increases in a restored image as the cut rate increases. If the quantization width is increased without cutting high frequencies, Mi
In the d-trace type quantization, there is a property that a component-specific pattern image is easily generated (in the case of Mid-riser quantization, random noise increases). In order to obtain a high quality image, that is, a restored image that is faithful to the original image, all frequency components may be quantized with a small quantization width. However, since the amount of codes at the time of encoding increases, the compression rate decreases.

The present inventors examine the relationship between the high-frequency cut degree, the quantization width, and the image quality, and adjust the quantization width and the high-frequency cut degree in accordance with the properties of each block to obtain a high compression ratio, high-fidelity restored image. I found that I could get it.

 The outline is described below.

The image quality of the restored image in the case of performing mid-trace type quantization has a strong correlation with the number of transform coefficient AC components quantized to zero in the quantization stage at the time of compression. That is, since the image of each block of the restored image is a weighted sum of a pattern unique to the component of the transform coefficient AC component quantized to a value other than zero at the time of quantization, the number of transform coefficient AC components quantized to a value other than zero is reduced. If the number is small, a component-specific pattern image appears in the restored image. Therefore, in blocks where component-specific pattern images tend to appear, the quantization width is first reduced to reduce the number of transform coefficient AC components that are quantized to zero, and then the code amount due to the reduced quantization width is reduced. The increase is suppressed by cutting high frequency components. The degree of high-frequency cut is increased for blocks with a smaller quantization width. The degree of high frequency cut is also changed by the compression ratio. To obtain a high compression rate as a whole, the degree of cut is increased. A method using the number of transform coefficient AC components quantized to zero as a parameter is suitable for a method of detecting a block in which a component-specific pattern image is likely to appear. For example, a plurality of quantization widths having different widths are set in advance, and the number of AC components quantized to zero when using the plurality of quantization widths is checked for each quantization width, and one block is determined. A quantization width is detected such that the number of AC components quantized to zero per minute is equal to or less than a certain threshold, and the AC component of the block is quantized with this quantization width. Also, the number of AC components quantized to zero is determined using one predetermined quantization width, and a small quantization width is assigned to a large number of blocks, and a large quantization width is assigned to a small number of blocks. A method of assigning is also conceivable. In addition, the variable coefficient AC components are rearranged in ascending order of absolute value, and a component whose number of components counted from the component having the smallest absolute value is equal to or less than a threshold value is detected, and the absolute value of the component or an approximate value of the absolute value is detected. May be used as a quantization width.
The method of classifying (ranking) blocks by paying attention to the number of AC components quantized to zero is smaller than the conventional method using the sum of squares or the sum of absolute values of AC components. It is excellent in considering the rounding of data due to computerization.

Further, the conventional method has a property that the quantization width is determined largely by the influence of the component having a large absolute value. The present invention differs from the conventional method in that when determining the quantization width, only the number of components having a large absolute value is considered and the magnitude of the value is not reflected in the determination of the quantization width. .

Next, how to determine the degree of cutting off the high frequency is defined by defining the number of components to be encoded (or quantized) as a quantization width or a function of the quantization width and the compression degree requested by the user. At this time, the function is set such that the number of components to be coded decreases as the block having a smaller quantization width and the required degree of compression increase.

In addition, among the components quantized to a value other than zero by quantization, the number of components to be encoded is defined as a quantization width or a function of the quantization width and a compression degree requested by a user. A method is also conceivable. In this case, the function is such that the number of components to be encoded (however, components quantized to a value other than zero) decreases as the block with a smaller quantization width and the required compression degree increase.

Further, a method of determining a high-frequency cut degree in which the ratio of the number of components to be coded to the total number of components quantized to a value other than zero by quantization is defined as a function of the quantization width can be considered. In this case, the function is such that the ratio of the component to be coded (the component quantized to a value other than zero) becomes smaller as the block has a smaller quantization width and the required compression degree is larger.

(Object and Configuration of the Invention) The present invention has been made in view of the above points, and
Achieved this objective by compressing data of a gradation image including a medical image using a transform coding method that performs trace-type quantization and obtaining a high-resolution restored image with a high compression ratio. To perform the Mid-trace type quantization, the blocks are classified into several classes using the frequency of occurrence of the transform coefficient AC component quantized to zero in the block as a parameter, and each block is suitable for each class. Compression is performed according to the quantization width and the high-frequency cut degree.

(Example) Hereinafter, the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of a gradation image data compression method according to the present invention.

In the figure, a frame memory 1 stores gradation image data to be compressed (in this example, the number of bits per pixel is 8 bits). Are read in block units (in this example, the block size is 16 pixels in the line direction and the column direction, respectively). The read block data is converted into a two-dimensional discrete cosine transform (2D-DC
T) The cosine transform is performed by the device 3 to obtain 256 transform coefficients for one block.

Next, of the 256 transform coefficients thus obtained, 255 AC components excluding one DC component are supplied to the first quantization device 4.
Basic is uniformly quantized by the quantization width w _o, to obtain a pair of the coefficient number and polarity number is a fixed length code to each transform coefficient by. The pair of the coefficient number and the polarity number is temporarily stored in the block buffer memory 6, and the coefficient number is sent to the ranking device 5.

The ranking device 5 performs a process of sorting the blocks into a plurality of ranks according to the coefficient number group of each block sent from the first quantization device 4, and obtains the resulting 1/2 quantization width (quantization). Is output to the second quantization device 7. The second quantization device 7 reads a pair of a coefficient number and a polarity number from the block buffer memory 6, and
The second quantization is performed using a quantization width determined based on a 1/2 quantization width and a compression degree requested by a user and a cut degree of a high-frequency component. The encoding device 8 converts the fixed-length code of the AC component obtained by the second quantization device 7 into a variable-length code and outputs the variable-length code to the terminal 9. The code data is transmitted from the terminal 9 or stored in the memory.

 FIG. 2A shows an example of the first quantization device 4.

First, 255 AC components of the conversion coefficients for one block excluding the DC component are input from the terminal 41. The absolute value of the 255 AC components is taken by the absolute value circuit 42, and at the same time, the polarity determination circuit 43 determines whether the conversion coefficient is positive or negative. The polarity judgment circuit 43 uses the judgment result as a polarity number at terminal 47.
Output to Here, the polarity number is represented by a 1-bit code j, and is “1” when the conversion coefficient is negative and “0” when the conversion coefficient is positive.

On the other hand, the absolute value of the transform coefficient output from the absolute value circuit 42 is divided by the basic quantization width w _o by the division circuit 44. The division result is truncated to the decimal point by the truncation circuit 45 (this result is represented by a coefficient number k (k = 0, 1, 2,...)
), And output to the terminal 46 and the terminal 47. The terminal 46 is connected to the ranking device 5, and the terminal 47 is connected to the block buffer memory 6. The polarity number and coefficient number obtained for one transform coefficient input from the terminal 41 form a pair of two, and are temporarily stored in the block buffer memory 6 from the terminal 47.

On the other hand, the DC component of each block has a sufficiently small quantization width w.
It is uniformly quantized by _dc , or temporarily stored in the block buffer memory 6 in the same manner as the other 255 AC components without performing quantization.

 FIG. 3 shows a state of the first quantization of the AC component.

The horizontal axis indicates the value of the conversion coefficient, and the horizontal axis indicates the frequency of occurrence of the conversion coefficient. The number sequence shown in FIG. 3A is a coefficient number k column output from the truncation circuit 45, and the number sequence shown in FIG. 3B is a polarity number j column output from the polarity determination circuit 43. It is.

Next, FIG. 4 shows an example of the ranking device 5 when the number of ranks is eight.

The terminal 51 is connected to the terminal 46 in FIG. 2 (A), and the coefficient number k output from the first quantization device 4 is input thereto. Next, the coefficient number k (k = 0,1,2,3,4
..) Are converted into 8-bit counter control codes C _i (i = 0, 1, 2,..., 7) by the counter control code table 52.

FIG. 5 (a) shows an example of the counter control code table 52.

Then the counter control circuit 53 takes in the counter control code C _i, the counter from the counter 0 in the counter 54 7
Eight counters to be controlled by on / off state of each bit of C _i. Counter control code C _i counter 0 in order from the bit lower bits of the counter 1, counter 2,
... corresponding to the counter 7 and the bit is turned on (=
In the case of 1), the counter control circuit 53 counts up the corresponding counter by +1. When the bit is off (= 0), the counter does not count up. Fifth
Figure (b) shows the correspondence between the bits and the counters C _i.

FIG. 6 (a) is a diagram showing the operating range of each counter.

Each counter operates only for the coefficient number k in the range indicated by the arrow in FIG. Since the counter 7 operates for all coefficient numbers, the count value of the counter 7 at the end of the ranking operation for one block is 255. When the count value of the counter 7 becomes 255, the comparison circuit 5
5 starts the comparison process. As will be described later, since the operation range of these counters is equal to half the value of the quantization width, the lower limit (operation range of counter 0) and the upper limit (operation of counter 6) Is determined by the lower and upper limits of the quantization width.

The threshold value S for the count value is given to the comparison circuit 55 from the terminal 56 at the start of the compression processing. Compression rate number P
Is a parameter indicating the magnitude of the compression ratio requested by the user from a terminal (not shown). In this example, P = 0, 1, 2
Are prepared. When the user requests a high compression ratio, 2 is assigned, and when the user requests a low compression ratio, 0 is assigned. Performs a comparison of the comparison circuit 55 and the threshold value S and the count value of each counter, sets the 1/2 quantization width k _s based on the comparison result. An example of the comparison procedure of the comparison circuit 55 is shown in the flowchart of FIG. First, counter 6
Is compared with the threshold value S (F-1). If the threshold value S is larger than the count value of the counter 6, the operation range for the coefficient number k of the counter 6 (9 in this example)
It was set as a half quantization width k _s (F-2), and outputs a 1/2 quantization step size k _s from the terminal 57 (F-16). If the threshold value S is equal to or smaller than the count value of the counter 6, the count value of the counter 5 is compared with S (F-3).

Following a similar manner determined operating ranges for the coefficients number k of the counter count value is less and the maximum threshold value S, the set this as 1/2 quantization width _{k s (F-1) ~} (F
-14). However, in comparison with the count value of the counter 0 and the threshold value S, S is equal to or smaller than the count value of the counter 0, and 1/2 quantization width k _s is a predetermined value (2 in the case of this example) (F-15).

When k _s is set, k _s is output from the terminal 57 (F−
16) The ranking operation for one block is completed, and the count values of the counters 0 to 7 are cleared to zero (F-17).

FIG. 6 (b) shows the threshold value S, the count value of each counter, and 1 /
Is a diagram illustrating the relation of 2 quantization width k _s.

The graph of each count value indicates the cumulative frequency of the coefficient number k counted by each counter of a certain block.
Assuming that the threshold S takes one of the values S ₁ , S ₂ , and S ₃ , in this graph, when S = S ₃ , k _s = k _s3 (= 4) When S = S ₂ , k _s = k _s2 (= 3). However, in order to prevent the taking smaller than necessary k _s is the when larger than the count value are both S value of each counter as in the case of S = S _1, assigns a value determined in advance in k _s . An unnecessarily small quantization width only results in an increase in the code amount, and has little effect on improving the image quality of the restored image. In this example, is set to S = when _{S 1 k s = k s1 (} = 2).

Thus 1/2 quantization width k _s obtained, the coefficient number k
Is the quantization width for. When half the quantization width with respect to the previous change coefficient of the first quantized with the basic quantization width w _o (expressed in w) and w _s, w _s is expressed by _{w s = k s · w o} .

As a method of determining the 1/2 quantization width k _s , w _s by the above-described ranking device 5, the following method may be used.

FIG. 6 (c) shows the operation of each counter at the resolution (= w _o ) of the coefficient number k, and the half quantization width w at the resolution before the first quantization at the basic quantization width w _o. FIG. 9 is a diagram showing a method for obtaining _s .

As in the one-dot chain line shown in FIG, Musube the count value of each counter in interpolation curve, the absolute value coordinates of the intersection point C of the curve and the S value, which is 1/2 quantization width w _s . When obtaining the interpolation curve, it is preferable to process the start point (the side having the smaller absolute value) and the end point (the side having the larger absolute value) of the curve as follows. Point a first diagram at the start point side, make straight portion absolute value as indicated by the point b becomes constant, so that no small w _s than necessary. This means that the lower limit of the quantization width is determined, and a meaningless increase in the code amount due to a quantization width that is too small is prevented. Then to prevent the occurrence large w _s than necessary by a point d not the end point of the curve at the end point, for example point e. This means that the upper limit of the quantization width is determined, and image quality deterioration due to an excessively large quantization width is prevented. Incidentally, it is possible to reduce the amount of computation required to obtain the w _s The use of linear interpolation for the calculation of w _s.

FIG. 6D shows the number i + 1 of counters to be set.
(1 = 7 in this example) The number of counters I ′ = 1 smaller than 1
Counts in (in this example I '= 3), and the counter number is I + 1 in the case the same resolution 1/2 quantization width k _s
And it illustrates a method for determining the w _s.

Again the absolute value coordinates of the point c by the same interpolation as that shown in FIG. 6 (c), which may be the w _s. w _s
To obtain k _s from, quantum points on both sides of w _s (in this case,
4w ₀ ) and calculate the width from 0 to this quantum point by the coefficient number k
Is represented by k _s (k _s = 4 in this example) or w _s
The closest quantum point on the zero side (4w _{0 in} this example) is obtained, and the width from 0 to this quantum point represented by a coefficient number k is represented by k _s (k _s = 4 in this example). ). The device scale can be reduced by reducing the number of counters as in this example.

Compression ratio have been several available, an apparatus capable of selecting a compression ratio for each compression ratio in advance determined the upper limit of 1/2 the quantization width k _s, is output from a terminal 510 of the ranking system 5 1/2
When the quantization width k _s exceeds the upper limit, it is better to perform a process of replacing the value of k _s in this limit. In particular, when a low compression ratio is selected, it is desirable to set the upper limit to a small value and increase the fidelity of the restored image.

FIG. 8 shows an example of the case where the above operation is performed using a table. For example, immediately after the 1/2 quantization width is output from the terminal 510 of the ranking device 4, the above-described table is passed through the table having such characteristics. Perform the operation. Figure 8 is entered
The 1/2 quantization width is represented by k _sin , and the output 1/2 quantization width is
_Expressed as k _sout . Also, c shows an example of a table when a high compression ratio is selected, where k _sout = k _sin for all k _sin . b shows an example of a table when a medium compression ratio is selected, and k _sin <
For s _s2 , k _sout = k _sin , and for k _sin ≧ k _s2 , k _sout = k _s2 . a shows an example of a table when a low compression ratio is selected, and when k _sin <k _s1 , k _sout = k _sin , k _sin ≧ k
_{In s1} , k _sout = k _s1 .

 FIG. 9 (a) shows another example of the ranking device 5.

The terminal 51 is connected to the terminal 46 of FIG. 2 (a). When the coefficient number k output from the first quantization device 4 is input from this terminal, the rearrangement 52 'is stored in the buffer memory 53'. The coefficient number k is stored. At this time, the rearrangement circuit 5 arranges the coefficient numbers k in the buffer memory 53 'in ascending order.
2 'stores the coefficient number k.

When the rearrangement of coefficient numbers for one block is completed,
Comparator circuit 54 'is the buffer memory 53' checks the value k of S-th coefficient number counted from the smallest in the 1/2 quantization width k _s k
Set to _s = and output from terminal 510, where = 0
In the case of, the predetermined value (2 in this example) is k _s
Assign as

This method realizes the same processing as the method shown in FIG. 6A by rearranging instead of using a counter. Incidentally, using interpolation as described sorting of the device or performed on the absolute value of the transform coefficients w ₀ prior to the quantization by the basic quantization width w _0, or 6 (c) 1/2 quantization width w _s at the resolution before performing basic quantization
Can be requested.

FIG. 9 (b) shows still another example of the ranking device 5.

The terminal 51 is connected to the terminal 46 of FIG. 2 (a). When the coefficient number k output from the first quantization device 4 is input from this terminal, the counter control circuit 52 ″ sets a predetermined constant. Counter the number of k that satisfies 0 ≦ k ≦ K for K
Count at 53 ″.

When the counting of the coefficient number k for one block is completed,
Counter 53 "count value v of the quantization width table 54" is converted to half the quantization width k _s by and outputted from the terminal 510.

 FIG. 10 shows an example of the quantization width table 54 ″.

Straight ab and the straight line cd in the figure respectively set the upper and lower limits of 1/2 quantization width k _s.

FIG. 11 (a) shows an example of the second quantization device 7, which comprises a quantization condition setting section 70 and a quantization section 71.

First, in the initial condition setting stage at the start of compression processing,
From 702, the compression ratio number P is read into the quantization condition setting unit 70. When the compression ratio number P is read, the value of P
One of the three tables A, B, and C in the component cut table 703 is selected. The component cutting table 703 enter a half quantization width k _s, a table for outputting a quantization completion component number x. Here, the quantization end component number x is a number indicating that only x components are quantized and coded toward a low-frequency AC component and a high-frequency component, and the remaining high-frequency components are cut. . The selected table is applied to all blocks of one frame of the image to be compressed.

 FIG. 12 shows an example of the component cut table 703.

This table is a table to obtain an output x to the input k _s, the value of the higher values of k _s is smaller x set to be smaller (or equal), the high-frequency cut small blocks of the quantization width Increase the degree. On the other hand, the compression ratio numbers P be the same k _s is too large (the compression rate is large), and set as the value of x decreases (or equal), high-frequency cut degree as for the compression ratio larger the Enlarge. In this example, when P = 0, table A has P =
When 1, the table B is selected, and when P = 2, the table C is selected.

When the initial condition setting stage is completed, the compression process is started for each block. When ranking process for one block is completed, 1/2 quantization width k _s is read from the terminal 701 to the condition conversion table 703, the initial condition table A in component cutting table 703 selected by the setting step, table B , The quantization width k _s by one of the tables in Table C
Is converted to a quantization end component number x. Loading quantizing condition setting unit 70 1/2 quantization width k _s, Upon completion of the setting of the quantization completion component numbers x, quantization processing of the block is started.

 Next, quantization will be described.

First, the data reading circuit 712 starts reading an AC component (a pair of a coefficient number k and a polarity number j) from the block buffer memory 6 via the terminal 711. At this time, the number of AC components to be read is given by the quantization end component number x, and x AC components are read from the low frequency side in the order shown in FIG. 16 (A) or (B). Coefficient number k of the read AC component is taken into the quantization circuit 713, Mid-trace type uniform quantization is performed in quantization width 2k _s, obtaining a code number l. Fig. 11 (b) shows the quantization circuit
713 shows an example.

In the quantization circuit 713, first, as a quantization offset,
The second-order coefficient number k '(k' = k + k _s ) is obtained by adding the 1/2 quantization width k _s given from the terminal 701 to the coefficient number k. Next 'is divided by is taken to the divider circuit 713-3 quantization width 2k _s, the code number the integer part of the quotient l (l = int {k' second order coefficient number k becomes / 2k _s}) Use a fixed-length code. The code number 1 is sequentially sent to the encoding device 8 via the number 715. The polarity number j paired with the code number is sequentially sent from the data read circuit 712 to the encoding device 8 via the terminal 714.

Next, an example of the encoding device 8 is shown in FIG. 13, and the encoding means is shown in the flowchart of FIG.

(I) First, the zero determination circuit 803 reads the code number 1 from the terminal 801 (P-1), and performs zero determination of the code number 1 (P-2).

(Ii) In the determination in step (P-2), the code number l
Is zero, the zero count value Z of the zero counter 804 is incremented by one (P-3). If the code number l is not zero, it is determined whether the zero count value Z is zero (P-4).

(Iii) In the determination of step (P-4), if the zero count value Z is zero, the code l and the polarity number j (read from the terminal 802) paired with the code l are stored in the Huffman code table 805 as a variable length code hlj. To (P-
5) Output hlj from the terminal 807 (P-6).

(IV) determined in step (P-4), if not zero count value Z is zero zero count value Z was converted by B1 code table 806 to the variable length code r _z (P-7), then zero count The value Z is cleared to zero (P-8), and then r _z
Is output from the terminal 807 (P-9), and then the step (P-
5) In step (P-6), the variable length code hli for the code number 1 and the polarity number j is output to the terminal 807.

(V) The above processing steps are classified into the following three steps.

Any of the above three processing steps is repeatedly performed on x components (code number 1) for one block (P-10).

(Vi) When any of the above processing steps for the x-th component is completed, it is determined whether the zero count value Z of the zero counter 804 is zero (P-11). If the zero count value Z is zero, If the zero count value Z is not zero, the run length is calculated.
Steps (P-7) and (P) in the Huffman encoding process
After obtaining (−8) and (P−9), the encoding operation for one block is completed.

FIG. 15 (a) shows an example of the Huffman code table 805.

In this example, a common Huffman code is used for each quantization width. In the case of performing uniform quantization as in the present embodiment, there is a large difference in the compression ratio between when an optimal Huffman code is used for each quantization width and when a common Huffman code is used for each quantization width. Since there is no Huffman code for each quantization width, a simpler device can be realized.

 FIG. 15 (b) shows an example of the B1 code table 806.

This example is an example in which a B1 code common to each quantization width is used as a run-length code assigned to a zero count value z. When the gradation image data is compressed by the present method, the frequency distribution of each zero run length becomes an exponential function distribution suitable for the B1 code. Therefore, a high compression ratio can be obtained by using the B1 code. In FIGS. 15 (a) and 15 (b), C and 1 take a value of 0 or 1, and are 1-bit codes satisfying ΔC.

FIG. 16 shows the data read circuit 71 in FIG.
2 shows the data reading order.

Since the conversion coefficient of the gradation image tends to have a smaller amplitude as the frequency component increases, the probability of occurrence of a component quantized to zero also increases as the frequency component increases. Therefore, Fig. 16 (a)
Alternatively, as shown in (b), if quantization and coding are performed from the low frequency side toward the high frequency side, the probability of occurrence of a long zero run increases, and effective run length coding can be performed. When the quantization end component number is x, only x components are read from the low frequency side, and the (x + 1) th to 255th AC components are not read.

FIG. 17 (a) shows another example of the second quantizing device 7 and the encoding device 8 in the case where the high-frequency cut degree is determined by the number of components quantized other than zero.

The second quantizer 7 and the thirteenth quantizer 7 described in FIG.
The difference from the encoding device 8 described with reference to the drawing will be described.

The sign cut table 703 is taken in from the terminal 701.
The compressibility number P taken in from 2 quantization width k _s and the terminal 702 is converted to a code cut number alpha. The code cut number α is
Of the code numbers 1 output from the quantization circuit 713, l ≠ 0
This is a number meaning that α is to be encoded by counting α from the low frequency side. Zero determination circuit 803 of encoding device 8
A code detection counter 70 for counting the number of l な る 0 among the code numbers 1 output from the quantization circuit 713 is provided therein.
6b there is the count value of the code detection counter 706b (code detection count value) y _b equals sign cut number alpha, zero decision circuit 803 sends a data read stop command to the data reading circuit 712, one block of The quantization encoding operation is stopped.

The terminal 714 is the Huffman code LUT805 shown in FIG.
Is a terminal for outputting the polarity number j.
Is a terminal for outputting the code number 1 to the Huffman code LUT 805. The terminal 812 is connected to the zero counter 804.

The quantization circuit 713 operates in the same manner as described with reference to FIG. Encoding procedure are that code detection count value y _b is used instead of the code passing count value y, the encoding operations of one block when the code detection count value y _b is equal to the code cut number α No. 14 except for the end
It is performed in the same procedure as described in the figure.

FIG. 18 (a) shows an example of the code cut table 704.

Code cutting table 704 is a table to obtain the output α with respect to the input k _s, the value of the higher values of k _s is smaller α is also set small. That is, of the total number of components that are quantized to non-zero smaller blocks of the quantization width, by reducing the number of components to be encoded, as the block value of k _s is small, cut many high-frequency components Set to As in FIG. 12, depending on the value of the compression ratio number P,
The compression ratio can be changed by selecting the tables A, B, and C.

FIG. 17 (b) shows the second quantizer 7 and the encoder 8 in the case where the degree of high-frequency cut is determined by the ratio of the number of components to be coded to the total number of components to be quantized other than zero. Is still another example.

Explaining the difference from FIG. 17 (a), the code cut table 704 is a half quantization width k fetched from the terminal 701.
_s and the compression ratio number P taken from the terminal 702 are converted into a code passing ratio β. The code passing ratio β is a quantization circuit 713
Means that β × 100 (%) is counted from the low frequency side with respect to the total number of 1 where l ≠ 0 among the code numbers 1 output from. Thus the count value of the ranking device 4 1/2 quantization step size k _s than to the counter used to determine the half quantization width (counter operating range k _s) (which m
) Must be output.

The count value m output from the ranking device 4 is a terminal
706, and the code passing ratio β
To obtain the code number γ (γ = int
{Β × m}).

The code number number γ means that among the code numbers l output from the quantization circuit 704, l = ≠ 0 is counted from the low frequency side and only γ are coded, and in FIG. It works in the same way as the described code cut number α.

That is, in the zero determination circuit 803, and code detection counter 803 which counts the number of l ≠ 0 is activated among the code number l, code detection count value y _b sign number number γ
When the value becomes equal to, the zero determination circuit 803 sends a data read stop command to the data read circuit 712 to stop the quantization encoding operation for one block. Terminals 714, 811, 81
2 and the operation of the quantization circuit 713 and the encoding procedure are the same as those in FIG.

FIG. 18 (b) shows an example of the code cut table 704.

Code cutting table 704 is a table to obtain the output β with respect to the input k _s, the value of the higher values of k _s is smaller β also set small. That smaller blocks of the quantization width, of the total number of components that are quantized to non-zero, by reducing the ratio of the number of components to be encoded, a lot of high frequency components as the block value of k _s is less Set to cut.

Note that, as in FIG. 12, A, B,
The compression ratio can be changed by selecting the table of C.

Alternatively, the quantization width may be determined for the coefficient number k after the first quantization, and the quantization may be performed for the transform coefficient w before the first quantization. First quantization device 4 in this case
Has, for example, a configuration as shown in FIG. A truncation circuit is connected to the terminal 46 connected to the ranking device 5.
The coefficient number k, which is the output of 45, is output to the terminal 47 connected to the block buffer memory 6, and the absolute value | w | of the conversion coefficient w, which is the output of the absolute value circuit 42, and the output of the polarity determination circuit 43. Is output. The output of the ranking system 5 in this case is 1/2 quantization width W _s, also the configuration of the second FIG quantizer 7 is a device configured by replacing the 1/2 quantization k _s to w _s.

 FIG. 19 shows an example of the code configuration.

FIG. 19 (a) shows an example of a code obtained as a result of compressing the entire image of one frame, which is composed of a header, a code of each block following the header, and a data end code.

FIG. 19 (b) shows an example of the header section configuration, which includes the block size (16 in this example), the number of blocks in the image line direction,
The number of blocks in the image column direction and the compression ratio number P are stored.

FIG. 19 (c) shows an example of the code configuration of each block.
1/2 quantization width k _s or w _s (fixed length code), the number of coded components (fixed length code), the number of the used table (fixed length code) when there are a plurality of Huffman code tables, It is composed of a flag (fixed-length code) indicating whether or not run-length encoding has been performed, a DC component code (fixed-length code), and an AC component code sequence (variable-length code sequence) following the flag. The number of AC component code strings is q (1 ≦ q ≦ 25) as shown in FIG.
5) a coefficient code sequence (variable-length code sequence hl ₁ j ₁ , hl ₂
j _2, ... hlqjq) and the drawing (E) shows such a run-length code (variable length code r _z) is configured alternating.

Hereinafter, examples of various parameters used in this embodiment will be described.

Basic quantization width w ₀ = 0.25 S = 200 Note that the above parameter is the information amount per pixel of the original image is n.
The block size is N × N with bits (n = 8 in this example)
When (N = 16 in this example), the block image is represented by f (x,
y), a value obtained by using a two-dimensional discrete cosine transform equation (shown in the following equation) in which a transform coefficient matrix is represented by F (u, v).

In the present invention, instead of the two-dimensional discrete cosine transform device 3, another orthogonal transform device, for example, 2
A dimensional Hadamard transform device, a slant transform device, or the like may be used. In the present invention, the block size is not limited to the size of 16 × 16, but may be 8 × 8, 32 × 32, 64 × 64.
However, the number of bits per pixel of the original image is not limited to 8 bits. Further, the AC component to be referred to at the time of ranking may not be the entire AC component but a specific AC component. For example, 12 on the DC component side
It is also possible to reduce the influence of high-frequency cut by using five components as specific AC components. Further, the classification in the present invention can be performed in combination with other classification parameters.

In addition, encoding using both a Huffman code and a run-length code at the time of encoding (encoding shown in the embodiment of the present invention)
By selecting coding using only Huffman codes for each block, coding efficiency can be increased. Also, low-frequency components near the DC component in one block have a low probability of occurrence of zero, and therefore are coded using only Huffman codes. If coding using both length codes is performed, coding efficiency is further improved.

(Effects of the Invention) As described above, in the present invention, a quantization width and a high-frequency cut degree are selected according to the properties of each block.
Since the quantization and the encoding are performed, it is possible to suppress the appearance of the component-specific pattern image and the appearance of the blur of the image, thereby obtaining a high-quality restored image with a high compression ratio. Became.

[Brief description of the drawings]

FIG. 1 is a block diagram of an embodiment of a gradation image data compression method according to the present invention, and FIGS. 2 (a) and 2 (b) are block lines of a different example of the first quantization apparatus shown in FIG. FIG. 3 is a diagram showing a coefficient number and a polarity number obtained by quantizing a transform coefficient AC component in relation to a frequency of occurrence of a transform coefficient;
FIG. 4 is a block diagram of an example of the ranking device shown in FIG. 1, FIG. 5 (a) is a counter control code table, and (b)
FIG. 6 (a) shows the correspondence between the counter control code and each counter, FIG. 6 (a) shows the operation range of each counter in the counter control circuit, and FIGS. 6 (b), (c) and (d) show the quantization width. FIG. 7 is a flowchart showing the comparison procedure of the comparison circuit, FIG. 8 is an example of a table for setting the upper limit of the quantization width according to the degree of compression, FIG. (B) is a block diagram showing another example of the ranking device shown in FIG. 1, FIG. 10 is an example of the quantization width table shown in FIG. 9 (b), and FIG. FIG. 1B is a block diagram of an example of the second quantization device shown in FIG. 1, FIG. 2B is a block diagram of an example of the quantization circuit shown in FIG. FIG. 13 is an example of the three tables A, B, and C in the component cut table shown in (a), and FIG. 13 is a block diagram of an example of the encoding device shown in FIG. FIG. 14 is a flowchart showing the AC component encoding procedure by the coding apparatus shown in FIG. 1, respectively Figure 15 (a) and (b) the thirteenth
FIGS. 16 (a) and (b) show an example of the data reading sequence in the data reading circuit of FIG. 11 (a), and FIGS. 17 (a) and (b) Block diagrams showing other examples of the second quantization device and the encoding device shown in FIG. 1, and FIGS. 18 (a) and (b) are shown in FIGS. 17 (a) and (b), respectively. FIG. 19 (a), (b), (c),
(D) and (E) are diagrams showing an example of the contents of the code configuration. 1 ... frame memory, 2 ... data reading device, 3
... Two-dimensional discrete cosine transform device, 4... First quantizer, 5... Ranking device, 6... Block buffer memory, 7.

Continuation of Front Page (56) References JP-A-62-222783 (JP, A) JP-A-61-147692 (JP, A) IEEE TRANSACTIONS ON COMMUNICATIONS Vol. COM-25 [11] (1977) p. 1285-1292 (58) Field surveyed (Int.Cl. ⁶ , DB name) H04N 7/24-7/68 H04N 1/41-1/419

Claims (2)

(57) [Claims] 1. A gradation image data compression method for dividing digitized gradation image data into a plurality of blocks and quantizing and encoding a transform coefficient obtained by performing orthogonal transformation for each block, wherein each block unit A method for compressing gradation image data, comprising: determining a quantization width of a transform coefficient AC component; and determining a high-frequency cut degree of a component to be quantized according to the determined quantization width and compression degree. 2. The gradation image data compression method according to claim 1, wherein the degree of high-frequency cut is increased for a block having a smaller quantization width.