JPH06292018A - High efficiency coder and high efficiency decoder - Google Patents

Abstract

PURPOSE:To improve the picture quality of a character and the surrounding of the character. CONSTITUTION:A code quantity constant circuit 22 receives a DCT transformation coefficient to calculate a normalizing coefficient alpha1 smaller than a normalizing coefficient alpha and a normalizing coefficient alpha2 larger than the normalizing coefficient alpha based on a total activity. The DCT transformation coefficient is given also to a character detection circuit 21. The character detection circuit 21 detects a character block based on the DCT transformation coefficient and its distribution and gives normalizing coefficients alpha1, alpha2 with respect to the character block and a non-character block to a basic quantization table 6. A quantization circuit 7 uses the basic quantization coefficient multiplied with the normalizing coefficient alpha1, alpha2 to quantize the DCT transformation coefficient thereby quantizing the character block with a finner quantization width than that of other block. Thus, the code quantity allocated to the character block is increased and the picture quality of the character block is improved.

Description

Detailed Description of the Invention

[Object of the Invention]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a high-efficiency coding device and a high-efficiency decoding device suitable for a karaoke device for displaying characters.

[0002]

2. Description of the Related Art In recent years, digital compression of images has been studied. In particular, various standardization proposals have been proposed for high-efficiency coding using DCT (discrete cosine transform). The high-efficiency coding technique is for coding image data at a smaller bit rate in order to improve the efficiency of digital transmission and recording. The orthogonal transform code including DCT is most suitable for image compression, and is currently applied to a CD-karaoke device that reproduces a compressed image on a CD-I player.

In orthogonal transform coding, one frame is divided into a plurality of blocks (m pixels × n horizontal scanning lines) and D
By performing CT processing and converting the coordinate axis components into orthogonal spatial frequency components (orthogonal components), it is possible to reduce the redundancy in the spatial axis direction. The orthogonal transform coefficients are arranged in order of frequency from the horizontal and vertical low band to the high band. The redundancy of the signal of the block is reduced by quantizing the orthogonally transformed component. Generally, most of the energy of the image signal is concentrated in the low frequency components, so that the amount of data is reduced by roughening the quantization width for the orthogonal transform coefficient in the high frequency band during quantization.

Further, variable length coding such as Huffman coding is applied to the quantized output to further reduce the data amount. Huffman coding performs coding based on the result calculated from the statistical code amount of the quantized output, assigning short bits to data with a high appearance probability and long bits to data with a low appearance probability. The variable length coding to be assigned reduces the total amount of data.

As described above, since variable length coding is adopted in high efficiency coding, the amount of data after coding changes depending on the statistical distribution of the quantized output. In other words, since the compression code amount differs depending on the design, the difference in code amount must be absorbed by the buffer during transmission. However, in consideration of the buffer capacity limitation and the recording rate limitation when recording compressed data on a storage medium, it is necessary to make the compressed data a constant rate. In particular, when recording on a medium such as a digital VTR and a digital video disk, it is necessary to set a constant rate in units of one to several screens in consideration of editing and the like.

FIG. 10 is a block diagram showing a conventional high-efficiency coding apparatus and high-efficiency decoding apparatus for controlling the compression code at a constant rate on a screen-by-screen basis as described above.

The input image signal is converted into a digital signal by the A / D converter 1 and given to the memory 2. The memory 2 outputs the image data to the DCT circuit 3 in block units of m × n pixels (for example, 8 × 8 pixels). DCT circuit 3
Outputs two-dimensional DCT processing of block-unit image data to frequency components, and outputs DCT conversion coefficients to the delay circuit 5 and the code amount constant circuit 4. An image having a fine pattern or an edge portion has a large amount of information in the middle and high frequencies of the DCT transform coefficient, and a flat image concentrates information only in the low frequency region.

The output of the delay circuit 5 is given to the quantization circuit 7,
The quantizing circuit 7 quantizes the DCT transform coefficient by dividing it with a quantized coefficient from a quantized table 6 to be described later to reduce the data amount, and then supplies it to the variable length coding circuit 8. The variable length coding circuit 8 performs variable length coding on the quantized output by Huffman coding, for example, to further reduce the data amount. The output of the variable length coding circuit 8 is added to the error correction code in the error correction coding circuit 10 and then output to the transmission line.

By the way, as described above, in the variable length coding, the compression code amount differs depending on the design. Therefore, in order to adjust the total code amount by changing the quantized coefficient by the normalization coefficient α, bit allocation is performed for each block in order to make the variable length code constant rate.

That is, the code amount constant circuit 4 calculates the activity indicating the fineness of the picture for each block from the alternating current (AC) component of the DCT transform coefficient, and further accumulates the block activity for one frame to make a total (frame ) Ask for activity. Then, the normalization coefficient α is obtained from the obtained total activity and is given to the basic quantization table 6.
The normalization coefficient α increases as the amount of information on the screen increases.

FIG. 11 is an explanatory diagram showing the basic quantization coefficients of the basic quantization table 6.

The DCT transform coefficient is composed of 8 × 8 frequency components, that is, a DC component and 63 AC components. Each of the 64 basic quantized coefficients in the basic quantization table shown in FIG. It corresponds. As mentioned above, D
Since the amount of information differs depending on the horizontal and vertical frequency bands of the CT transform coefficient, the basic quantization coefficient is AC1 for each band.
To AC14. The basic quantization table 6 multiplies the basic quantization coefficient by the normalization coefficient α and outputs the quantization coefficient to the quantization circuit 7. The quantizing circuit 7 divides the DCT transform coefficient by the quantizing coefficient for quantization, and the code amount for each frame is adjusted by adjusting the quantizing coefficient with the normalizing coefficient based on the activity.

On the other hand, the block activity and the frame activity obtained by the code amount constant circuit 4 are given to the coding rate control circuit 9. The coding rate control circuit 9 distributes the code amount (set code amount) that can be used for encoding the AC component of one screen to each block in proportion to the block activity. That is, the coding rate control circuit 9 obtains the activity ratio of each block from the frame activity and the block activity by the operation shown in the following formula (1), and the code amount (bit allocation) for each block.
And outputs the obtained bit allocation data to the variable length coding circuit 8. The variable length encoding circuit 8 performs constant rate conversion by performing variable length encoding on each block within the range of the number of distributed bits.

Number of allocated bits of block = (set code amount of AC component) × (block activity) / (frame activity) (1) It is to be noted that variable length coding is performed after determining the distributed bit amount by obtaining the frame activity. In order to adjust the time in the arithmetic processing, the delay circuit 5
The T transform coefficient is delayed and given to the quantizing circuit 7.

In the receiving system, the compressed data input through the transmission path is given to the error correction decoding circuit 12 to perform error correction. The error-corrected data is variable-length decoded in the variable-length decoding circuit 13 and given to the inverse quantization circuit 14. Further, the normalization coefficient decoding circuit 15 decodes the normalization coefficient α from the error-corrected data and outputs it to the inverse quantization coefficient calculation circuit 16. The inverse quantization coefficient calculation circuit 16 uses the same basic quantization table as in FIG. 11 to calculate the quantization width in the transmission system and gives it to the inverse quantization circuit 14. The inverse quantization circuit 14 inversely quantizes the variable-length decoded output to restore the data before quantization to the inverse DC.
It is given to the T circuit 17.

The inverse DCT circuit 17 performs D by the inverse DCT processing.
The data before the CT processing is restored and given to the deblocking circuit 18. The deblocking circuit 18 converts the data in block units into one screen and gives it to the D / A converter 19, and the D / A converter 19 returns it to an analog signal and outputs it. The normalization coefficient α is determined by the code amount constant circuit 4 on a screen-by-screen basis and is a fixed value for one screen.

In this way, orthogonal transform coding is performed in block units. This method has a highly efficient coding property for natural images. However, a character or line image has a drawback that ringing called mosquito noise occurs and the image quality is significantly deteriorated. That is, by quantizing the transform coefficient, the high frequency component is deleted, and ringing noise is generated in the character portion on the screen and the surroundings, and the character quality is degraded. Also, the edges of the letters will be blurred.

In particular, recently, MPEG (Motion Pic) employing DCT conversion and motion-compensated interframe predictive coding has been adopted.
ture Experts Group) system has become widespread, and a karaoke device equipped with a CD-I player that adopts this system and highly compresses (about 1/10) a moving image to 1 to 1.5 Mb / s has been commercialized. ing. Since this device is a high compression type, in particular, a block portion containing a large amount of middle and high frequency components,
That is, the image quality of the character edge portion is extremely deteriorated, which is a problem for a person who sings while watching the character portion of the screen of the karaoke device.

[0019]

As described above, in the above-described conventional high-efficiency coding apparatus and high-efficiency decoding apparatus, when high compression is applied, ringing noise (mosquitoes) is generated in the character or line image portion. There is a problem in that the image quality is remarkably deteriorated due to noise).

It is an object of the present invention to provide a high-efficiency coding apparatus and a high-efficiency decoding apparatus capable of preventing block distortion and ringing noise in a character or line image portion.

[Structure of the Invention]

A high-efficiency coding apparatus according to the present invention is a high-efficiency coding apparatus for quantizing an orthogonal transform coefficient obtained by orthogonally transforming input image data in block units to reduce the code amount. In the character block detection means for detecting a block including a character edge according to the AC component of the orthogonal transformation coefficient and its distribution pattern, and based on the detection result of the character block detection means, a quantization width for the block including the character edge. Is provided more finely than the quantization width for other blocks, and the high-efficiency decoding device of the present invention is a device for orthogonally transforming image data block by block and quantizing the inverse code. In a high-efficiency decoding device for decoding by quantization and inverse orthogonal transform, a luminance level change of image data decoded by the inverse orthogonal transform Therefore, based on the character detection means for detecting the character area and the detection result of this character detection means, the brightness level of at least the character area of the character area of the decoded image data and its peripheral area is flattened to the brightness average level. And a flattening unit that outputs the converted data.

[0022]

In the high-efficiency coding apparatus of the present invention, the character block detecting means detects a block including a character edge based on the AC component of the orthogonal transform coefficient and its distribution pattern.
When the block including the character edge is detected, the quantization control means finely sets the quantization width for this block. As a result, the code amount assigned to the block including the character edge is increased, and the image quality of the character portion is improved.

Further, in the high-efficiency decoding apparatus of the present invention, the character detecting means detects the character area from the change in the brightness level of the decoded image data. The flattening means flattens the image data of the character area and its peripheral area to the brightness average level. As a result, it is possible to prevent the quality of the character from deteriorating due to the change in the brightness of the character and to prevent the character outline from being blurred.

[0024]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of a high efficiency coding apparatus and a high efficiency decoding apparatus according to the present invention.
In FIG. 1, the same components as those in FIG. 10 are designated by the same reference numerals.

The input image signal is supplied to the A / D converter 1. The A / D converter 1 converts the input image signal into a digital signal and gives it to the memory 2. The memory 2 outputs the image data to the DCT circuit 3 in units of blocks of horizontal m pixels × vertical n pixels (for example, 8 × 8). The DCT circuit 3 performs a two-dimensional DCT process on a block unit of m × n pixels to perform DC
The T conversion coefficient is output to the delay circuit 5, the character detection circuit 21, and the code amount constant circuit 22.

2 to 8 are explanatory diagrams for explaining the DCT transform coefficient for the character image. FIG. 2 shows a character image, and FIGS. 3 to 5 show the relationship between the character image and the DCT transform coefficient. 3 (a) to 5 (a) show a part of characters, and FIGS. 3 (b) to 5 (b) show the energy concentration state of DCT transform coefficients. 6 and 7
Shows an example of the actual DCT transform coefficient, and FIG.
The arrangement of the AC components of the T conversion coefficient is shown.

Now, it is assumed that the image of the character "ha" has a size of vertical 4 blocks × horizontal 4 blocks, as shown in FIG. That is, it is assumed that the size of the character "wa" is larger than that of the block. In addition, the character "wa" is
On the screen, the brightness is higher than that of the background, such as white. 16 blocks in rows a to d and columns 1 to 4 respectively
By using blocks a-1 to a-4 to d-1 to d
Represented by -4.

Of these blocks, block a-1,
For b-1, c-1, d-1, a-3, a-4, c-4, etc., as shown in FIG. 3A, the background and the character portion (black portion) are switched in the horizontal direction. That is, the luminance change in the horizontal direction is large and the luminance change in the vertical direction is small. Therefore, in these blocks, the AC component of the DCT transform coefficient is
The data of large absolute value is concentrated in the shaded portion of (b), that is, in the horizontal low range to high range. Further, as shown in FIG. 4A, the blocks b-3, b-4, etc. in FIG. 2 have a large vertical luminance change and a small horizontal luminance change. Energy is concentrated in the vertical component from the low band to the high band, that is, in the shaded portion in FIG. 4B. Further, when the high-intensity part of the character is formed in the diagonal direction of the block as shown in FIG. 5A, the AC component of the DCT transform coefficient is as shown by the shaded area in FIG. 5B. Energy is concentrated in the horizontal and vertical diagonal directions where the horizontal and vertical bands change similarly.

As described above, when the characters are larger than the block size, the luminance distribution of each block is as shown in FIG.
(A) to FIG. 5 (a), and the energy of the DCT transform coefficient is concentrated in the horizontal direction, the vertical direction, and the oblique direction of the horizontal and vertical directions indicated by the hatched portions in FIGS. 3 (b) to 5 (b), respectively. And distributed. Therefore, FIG. 3B to FIG. 5B
The sum of squares of the coefficient values in one row in the direction of the shaded area is extremely large.

In FIGS. 6 and 7, one block is composed of 8 × 8 pixels, and the background has a black (data 0) (hatched portion) and a white (data 255) character block of eight types of patterns A to H.
, The input of the DCT circuit 3 and the actual DCT transform coefficient are shown.

As shown in each of the patterns in FIGS. 6 and 7, the image has a flat background, the brightness of which changes rapidly at the boundary between the background and the character portion, and the brightness of the character portion is also flat. There is also a feature in that there are a small number of high coefficient values in the middle range of the DCT transform coefficient.

From the characteristics shown above, it has been proved that a block including a character edge having a certain thickness (character edge block) can be detected under the following conditions (1) to (4). There is. As shown in FIG. 8, the DCT conversion coefficient is composed of a DC component and 63 alternating current (AC) components, and the alternating current component is represented by AC1 to AC63 from the horizontal and vertical low frequencies to the high frequencies. And

(1) There are u (≧ 1) AC components whose absolute values are larger than a predetermined threshold value T1 in AC1 to AC63.

(2) k1 for indicating the AC component in the middle range,
The following expression (2) is satisfied using k2 (1 <k1 <k2 <n). Note that T2 is a predetermined threshold value.

[0035] (3) The following expression (3) or expression (4) is satisfied. In addition, T
3 and T4 are predetermined threshold values.

[0036] (4) The horizontal component sum of squares, the vertical component sum of squares, and the diagonal component sum of squares are expressed by the following equations (5), (6), and (7), respectively.
To be satisfied. Note that T5 is a predetermined threshold value.

[0037] The character block almost corresponds to any one of the above conditions (1) to (4) and also corresponds to a plurality of conditions. As a result, it is possible to determine that a block that satisfies the above-described conditions (1) to (4) is a character block.

In particular, when the characters can be limited to white characters, that is, high-intensity characters as in karaoke software, the threshold values T1 to T5 can be set to high values,
The character block can be easily determined.

Even in the case of blocks other than the character block, if the brightness change is similar to that of the character block,
The DCT conversion coefficient has a component of extremely high value (value of T1 or more) (condition (1)), particularly in the middle range (ACk1 to Ak).
A relatively high value is concentrated in Ck2) (condition (2)), the AC component value is large overall (condition (3)), and the energy is concentrated and distributed in a predetermined direction. Multiple conditions (4) may be met. However, by appropriately setting the threshold values T1 to T5, it is possible to prevent the omission of the detection of the character block.

The character detection circuit 21 operates from the DCT circuit 3 to the DCT.
The conversion coefficient is given, the character block is detected by determining whether or not the input block satisfies the above-mentioned condition, and the detection signal is given to the switch 23.

The character portion can also be detected by detecting the change in the brightness level of the output of the A / D converter 1 for each line. However, considering that the processing is performed for each block after the DCT processing, the method of detecting in units of blocks like the character detection circuit 21 of the present embodiment is advantageous. Although it is possible to detect a character block only by changing the brightness level in the block, it is difficult to detect horizontal / vertical or diagonal directionality only from the brightness level. Since the degree of concentration is large, the determination is easy and advantageous.

The code amount constant circuit 22 is connected to the DCT circuit 3 through D.
Given the CT transform coefficients, calculate the activity of each block. The code amount constant circuit 22 accumulates block activities to obtain a total activity (frame activity), and calculates a normalization coefficient α based on the obtained total activity. Further, in the present embodiment, the code amount constant circuit 22 calculates the normalization coefficient α1 for the block including the character edge and the normalization coefficient α2 for the other non-character block. That is, the coefficient value α1 smaller (smaller) than the normalization coefficient α is set for the character block, and the coefficient value α2 coarser (larger) than the normalization coefficient α is set for the other blocks.
The normalization coefficients α1 and α2 satisfy the following expression (8).

Α × (total number of blocks in one screen) = α1 ×
(Number of character blocks) + α2 × (number of blocks other than characters) (8) The normalization coefficients α1 and α2 are given to the basic quantization table 6 via the switch 23. The switch 23 is controlled by the character detection circuit 21 to select the normalization coefficient α1 when it is determined to be a character block and to select the normalization coefficient α2 when it is determined to be a non-character block to perform basic quantization. It is designed to output to table 6. The basic quantization table 6 multiplies the basic quantization value shown in FIG. 10 by the normalization coefficient α1 or the normalization coefficient α2, and outputs it to the quantization circuit 7 as a quantization coefficient. The quantizing circuit 7 quantizes the DCT transform coefficient from the delay circuit 5 using the supplied quantized coefficient and outputs the quantized DCT transform coefficient to the variable length coding circuit 8. The delay circuit 5 delays the output of the DCT circuit 3 by the time required to calculate the total activity of one screen.

The block activity and total activity from the code amount constant circuit 22 are also given to the coding rate control circuit 9. The coding rate control circuit 9 uses the above equation (1).
The bit allocation of each block is determined based on the calculation shown in (3), and the data indicating the number of allocated bits is output to the variable length coding circuit 8. The variable-length coding circuit 8 Huffman-codes the quantized output, and outputs it. In this case, the variable-length coding circuit 8 stops coding for a code exceeding the number of distributed bits by coding, and outputs each block by coding within the range of the number of distributed bits. There is. As a result, the code amount within one screen is suppressed within the set code amount.

The variable length coded output is the error correction coding circuit 10
Give to. The error correction coding circuit 10 adds an error correction code to the input variable length code, packs it, and outputs it to the transmission path. In this case, the quantization coefficients α1 and α2 are also transmitted for each block. In the storage medium system, the encoded output from the error correction encoding circuit 10 is recorded on a tape or the like via a magnetic head, for example.

On the other hand, on the decoding side (reception system or reproduction system), the transmitted data is given to the error correction decoding circuit 12. The error correction decoding circuit 12 error-corrects the transmission data using the error correction code, and then supplies it to the variable length decoding circuit 13. The variable length decoding circuit 13 performs Huffman decoding processing to restore the data before Huffman coding. Further, the normalization coefficient decoding circuit 15 decodes the normalization coefficients α1 and α2 transmitted for each block from the output of the error correction decoding circuit 12 and supplies them to the inverse quantization coefficient calculation circuit 16. The inverse quantization coefficient calculation circuit 16 multiplies the same table as the basic quantization table by the normalization coefficient α1 or the normalization coefficient α2 to obtain the inverse quantization coefficient for each block, and outputs it to the inverse quantization circuit 14.

The inverse quantization circuit 14 inversely quantizes the variable-length decoded output using the inverse quantization coefficient, restores the data before quantization, and outputs it to the inverse DCT circuit 17. The inverse DCT circuit 17 is an inverse DCT
The data is returned to the data before the DCT processing by the processing and is output to the deblocking circuit 18. The deblocking circuit 18 restores the data in m × n pixel block units to the data in the scanning order, outputs the data to the terminal a of the switch 26, and supplies the data to the flattening circuit 24 and the character detection circuit 25.

Since the output of the deblocking circuit 18 has a small quantization coefficient of the character edge block, the code amount used for this block is large, the image quality of the character edge is improved, and the distortion is remarkably reduced. However, as a phenomenon peculiar to DCT processing, the brightness level of the character itself is not stable,
Further, the edge part is slightly blurred. Therefore, in the present embodiment, the flattening circuit 24 and the character detection circuit 25 are adopted in order to improve the image quality of thick characters such as karaoke software.

That is, the flattening circuit 24 is supplied with the output of the deblocking circuit 18, and outputs the average value of the luminance levels of the input image data to the terminal b of the switch 26. The character detection circuit 25 determines that the brightness level of the output of the deblocking circuit 18 has a difference larger than a predetermined threshold value a (> 0) between adjacent pixels and the change point (pixel Number of pixels from the position) to the change point (pixel position) on the −a side x
If (is larger than a predetermined number, this x pixel area is determined to be a character area, and a detection signal is output to the switch 26 to select the terminal b. Further, when there is a change in brightness around the character. Since the edge part is easily blurred, the character detection circuit 25
The switch 26 allows the switch 26 to select the terminal b even during a period corresponding to several pixels adjacent to the pixels in the character area.

The switch 26 selects the terminal b when the output of the deblocking circuit 18 is judged to be a character area by the detection signal from the character detection circuit 25 and selects the signal of the average brightness level from the D / A converter. In other cases, the terminal a is selected and the output of the deblocking circuit 18 is directly output to the D / A converter 19. The D / A converter 19 converts the input data into an analog image signal and outputs it.

Next, the operation of the embodiment thus constructed will be described with reference to the graph of FIG. Figure 9
(A) and (b) show the input and output of the flattening circuit 24, respectively.

An image signal including a character portion is converted into an A / D converter 1
It is converted into a digital signal by and is given to the memory 2.
The memory 2 converts the data into a block unit of 8 × 8 pixels and outputs the data to the DCT circuit 3, and the DCT circuit 3 outputs the DCT.
The spatial coordinate components are converted into frequency components by the processing and output. The output of the DCT circuit 3 is delayed by the delay circuit 5 for a period for obtaining the total activity, and is given to the quantizing circuit 7.

The code amount constant circuit 22 obtains the block activity and the total activity from the DCT transform coefficient,
The normalization coefficient α is calculated based on the total activity. Further, the code amount constant circuit 22 is based on the above equation (8),
The normalization coefficient α1 smaller than the normalization coefficient α is obtained as the normalization coefficient for the character block, and the normalization coefficient α2 larger than the normalization coefficient α is obtained as the normalization coefficient for the other blocks.

It is assumed that the predetermined block data is a character block. If the character detection circuit 21 determines that this block data satisfies the above conditions (1) to (4) and detects that it is a character block, it switches
By controlling 23, the normalization coefficient α 1 from the code amount constant circuit 22 is given to the basic quantization table 6. Basic quantization table 6
Is applied to the quantization circuit 7 by multiplying the basic quantization coefficient by the normalization coefficient α1. The quantizing circuit 7 is given a small quantizing coefficient, and the encoded character block has a sufficient amount of information even in the high frequency component.

On the other hand, it is assumed that the predetermined block data is a non-character block. The character detection circuit 21 controls the switch 23 by detecting that the block data is a non-character block from the judgments of the above conditions (1) to (4). As a result, the normalization coefficient α 2 from the code amount constant circuit 22 is given to the basic quantization table 6, and the basic quantization table 6 gives a relatively large quantization coefficient to the quantization circuit 7.
This reduces the code amount of the quantized output of the non-character block. Since the normalization coefficients α1 and α2 satisfy the above equation (8), the code amount of the quantized output in one frame is constant.

The quantized output is given to the variable length coding circuit 8 and variable length coded within the range of the set code amount for each frame, and given to the error correction coding circuit 10. Error correction coding circuit
Numeral 10 adds an error correction code to form a packet and outputs it to the transmission line. In this case, information on the normalization coefficients α1 and α2 is also transmitted for each block.

As described above, on the encoding side, the character edge block is detected from the DCT transform coefficient, and the character edge block is quantized using a small quantization coefficient, so that a sufficient number of coding bits can be obtained. Allocate. As a result, it is possible to prevent block distortion and ringing noise (mosquito noise) from occurring in the character block, and it is possible to significantly reduce noise around the character, the character edge, and the character periphery.

On the other hand, on the decoding side, the transmission data is subjected to error correction in the error correction decoding circuit 12 and then given to the variable length decoding circuit 13 and the normalization coefficient decoding circuit 15. The variable length decoding circuit 13 variable length decodes the transmission data and supplies it to the inverse quantization circuit 14. On the other hand, the normalization coefficient decoding circuit 15 decodes the normalization coefficients α1 and α2 transmitted for each block and gives them to the inverse quantization coefficient calculation circuit 16. The inverse quantization coefficient calculation circuit 16 multiplies the input normalization coefficient by the basic quantization table to obtain the inverse quantization coefficient and supplies it to the inverse quantization circuit 14. The dequantization circuit 14 dequantizes the data of the character block using the dequantization coefficient based on the normalization coefficient α1, and dequantizes the data of the non-character block using the dequantization coefficient based on the normalization coefficient α2. To restore the original data. Reverse DC
The T circuit 17 performs inverse DCT processing on the dequantized output, and the deblocking circuit 18 returns the block data to a data string in the scanning order and outputs it.

Now, the output of the deblocking circuit 18 is shown in FIG.
It is assumed that the signal is as shown in (a), that is, the signal of the character portion of the original image having white (high luminance level) and black edge like karaoke software. The period A in FIG. 9A corresponds to the character area, and the other period corresponds to the background portion. The character detection circuit 25 detects a change in the brightness level, and the change is shown in FIG.
It is detected whether it is larger than a predetermined threshold value a shown in (a). Further, the character detection circuit 25 obtains the number of pixels x from the pixel position where the brightness level changes by + a to the pixel position where the brightness level changes by -a, and when the number of pixels x is larger than the predetermined number of pixels, that is, when the white level is If the width is more than the specified width, the white level period is judged as the character area and the switch
The detection signal is output to 26.

On the other hand, the flattening circuit 24 averages the luminance levels of the deblocking circuit 18 and outputs the averaged luminance levels. The switch 26 selects the output of the flattening circuit 24 for the signal portion of the character area. As a result, as shown in FIG.
The data in the character area is flattened to a predetermined white level. Further, when there is a change in brightness around the character, the edge portion is easily blurred, so the character detection circuit 25 causes the pixel determined to be the character area to be adjacent to a few pixels (see FIG. 9).
Also in the section B) of (a), the switch 26 is made to select the terminal b. As a result, the area around the character area is also the section B in FIG. 9B.
As shown in, the pixel is flattened by the average of the luminance levels. The output of the switch 26 is converted into an analog signal by the D / A converter 19 and output.

As described above, on the decoding side, the character area is detected by the change of the brightness level of the decoded image data, and the brightness level of the character area section is flattened by the average level, whereby the brightness of the character itself is changed. It suppresses the deterioration of quality.

The present embodiment is not limited to the above embodiment. For example, in the above embodiment, the bit allocation of each block by the coding rate control circuit 9 is based on the above equation (1). Although it was decided and the encoding was performed within the set code amount on a screen-by-screen basis, for the character block, the bit allocation number was increased for the character block in order to reduce the distortion due to the stop of the encoding of the high frequency component. You may.
In this case, control may be performed so as to reduce the number of coded bits of the non-character block or the block of the peripheral portion (four sides of the screen) that is inconspicuous on the screen.

[0063]

As described above, according to the present invention, it is possible to prevent block distortion and ringing noise from occurring in a character or line image portion.

[Brief description of drawings]

FIG. 1 is a block diagram showing a high efficiency encoding device and a high efficiency decoding device according to an embodiment of the present invention.

FIG. 2 is an explanatory view showing a character image.

FIG. 3 is an explanatory diagram for explaining characteristics of a character image.

FIG. 4 is an explanatory diagram for explaining characteristics of a character image.

FIG. 5 is an explanatory diagram illustrating characteristics of a character image.

FIG. 6 is an explanatory diagram for explaining characteristics of a character image.

FIG. 7 is an explanatory diagram illustrating characteristics of a character image.

FIG. 8 is an explanatory diagram for explaining DCT transform coefficients.

FIG. 9 is a graph for explaining the operation of the example.

FIG. 10 is a block diagram showing a conventional high-efficiency encoding device and high-efficiency decoding device.

FIG. 11 is an explanatory diagram for explaining a basic quantization table.

[Explanation of symbols]

3 ... DCT circuit, 7 ... Quantization circuit, 8 ... Variable length coding circuit, 14 ... Inverse quantization circuit, 15 ... Normalization coefficient decoding circuit,
21 ... Character detection circuit, 22 ... Code amount constant circuit, 24 ... Flattening circuit, 25 ... Character detection circuit

Claims (3)

[Claims] 1. A high-efficiency coding apparatus for quantizing an orthogonal transform coefficient obtained by orthogonally transforming input image data in block units to reduce a code amount, wherein a character is generated by an AC component of the orthogonal transform coefficient and its distribution pattern. Character block detecting means for detecting a block including an edge, and quantization control for making a quantization width for a block including the character edge finer than a quantization width for another block based on a detection result of the character block detecting means. A high-efficiency coding apparatus comprising: 2. A high-efficiency decoding device for decoding a code obtained by orthogonally transforming and quantizing image data in block units by inverse quantization and inverse orthogonal transform, wherein the luminance level of the image data decoded by the inverse orthogonal transform A character detection unit that detects a character region based on a change, and based on the detection result of the character detection unit, the luminance level of at least the character region of the character region of the decoded image data and its peripheral region is set to a luminance average level. A high-efficiency decoding apparatus comprising: a flattening unit for flattening and outputting. 3. The character block detection means is characterized in that the orthogonal transformation coefficient in the middle and high frequencies has an absolute value component equal to or more than a predetermined threshold, and the sum of squares of the orthogonal transformation coefficient in the middle frequency is equal to or more than a predetermined threshold. At least one of the characteristics and the characteristics in which energy is concentrated in a component having a substantially constant vertical frequency, a component having a substantially constant horizontal frequency, or a component having a substantially constant change in horizontal and vertical frequencies of the orthogonal transform coefficient. The high-efficiency coding apparatus according to claim 1, wherein the block that is included is detected as a block including a character edge.