JP3364939B2 - Image coding device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [0001] BACKGROUND OF THE INVENTION The present invention relates to an image coding apparatus.
In particular, image data is highly efficient encoded by discrete cosine transform.
The present invention relates to an image encoding device to be converted. [0002] 2. Description of the Related Art Image data is transmitted, for example, by magnetic tape.
When recording on a recording medium such as a
Various encodings are employed. For example, so-called prediction code
, Transform coding, vector quantization and the like are known. [0003] By the way, the above-mentioned transform coding is based on the image signal.
Utilizing the correlations, sample values (hereinafter referred to as image data)
) Is converted to mutually orthogonal axes and the correlation between image data
Is de-correlated to reduce the amount of data.
The basis vectors are orthogonal to each other and the average signal power before
Average power of so-called transform coefficients obtained by summation and orthogonal transform
And the power concentration on the low-frequency component is excellent.
Orthogonal transformation is adopted, for example, so-called Hadamard transformation.
In other words, Haar transformation, Karnen-Roube (KL) transformation,
Discrete Cosine Transfor (DCT)
m), discrete sine transform (hereinafter DST: Discrete Sin
e Transform), slant transformation, etc.
Have been. Here, the DCT will be briefly described.
You. DCT transforms images horizontally and vertically in spatial arrangement
Both are divided into image blocks consisting of n (n × n) pixels.
Divide the image data in the image block using the cosine function
The orthogonal transformation is performed. This DCT is a high-speed calculation algorithm.
Algorithm that enables real-time conversion of image data.
The emergence of one-chip LSI
It is widely used for transmission and recording. Also,
DCT is a coding efficiency that is a low frequency band that directly affects the efficiency.
The above KL transformation, which is an optimal transformation in terms of the power concentration to the components
It has almost the same characteristics as the replacement. Therefore,
The transform coefficient obtained by DCT is calculated as
Encoding only the data, greatly increasing the amount of information as a whole
A significant reduction is possible. Specifically, image data obtained by DCT is obtained.
The conversion coefficient to be used is, for example, C_ij(I = 0 to n-1, j = 0 to
n-1), the conversion coefficient C₀₀Is the flat in the image block
It corresponds to the DC component representing the average brightness value, and its power is usually
It is considerably larger than other components. So this DC
When the components are coarsely quantized, visually large image quality degradation and
Noise that is peculiar to orthogonal transform coding
Since the lock distortion occurs, the conversion coefficient C₀₀Many
Equal quantum by allocating the number of bits (for example, 8 bits or more)
And other components except DC component (hereinafter referred to as AC component)
Conversion coefficient C_ij(C₀₀), For example, visual space
Utilizing the visual characteristic that the frequency decreases at high frequencies,
Quantize by reducing the number of bits assigned to higher frequency components
It has become so. In the transmission and recording of image data,
Conversion coefficient C obtained by DCT of image data_ijThe above
After the quantization, so-called Hahma
Coding (Huffman coding) and run-length coding (Ru
n Length coding)
Synchronization and parity are added to encoded data for transmission and
Records are made. Further, for example, a video signal is converted into a digital signal.
Digital video tape recording on magnetic tape
Editing and variable-speed playback in a coder (hereinafter simply referred to as VTR)
And so on, one frame or one field of data
It is desirable that the data amount be constant (fixed length).
Considering the scale, the encoded data is
It is also desirable that the sync blocks collected for several minutes have a fixed length.
No. Therefore, in a VTR, a plurality of quantization widths different from each other are used.
Of the sync block
The condition for using one quantizer for image blocks is also
And the data amount of the sync block is less than or equal to a predetermined value.
Select the quantizer with the smallest quantization width to perform quantization.
It has become. This is the image block in the sync block.
When the quantization is performed by switching and selecting the quantizer for each
Quantizer information must be recorded for each image block
The amount of data (overhead) increases
Therefore, it is to avoid it. [0008] However, as described above,
Same for the image block of one sync block
When a quantizer is used, the power of the AC component (C_ij ^Two, I, j
The so-called definition (activity) defined in (0) is different.
Image blocks are mixed in the same sync block
In both cases, there are many image blocks with high activity
And the conversion coefficient C_ijConcentration in the low-frequency components of the
A quantizer having a large quantization width will be selected. This
In the case of, the activity is low, that is, the pattern is monotonous
Image blocks with a small dynamic range
The quantization width is relatively large (coarse) with respect to the
It is quantized, so-called quantization distortion or block distortion
There is a problem that only the eye is visually conspicuous. Therefore, the activity of each image block
Is detected, and the conversion coefficient C is determined according to the activity._ijWeight
It is conceivable to attach and quantize, but a monotonous pattern
Even for blocks, for example, horizontal or vertical
Activity) in the image block where
The problem of high detection and coarse quantization occurs. The present invention has been made in view of such circumstances.
High-definition only image blocks with complex patterns
Image blocks of different degrees.
When performing quantization by weighting the permutation coefficient,
And quantization distortion can be made visually inconspicuous,
To provide an image encoding device
It is the target. [0011] SUMMARY OF THE INVENTION Image coding according to the present invention
In order to solve the above-mentioned problem, the apparatus spatially converts the image data.
The block is divided into n × n blocks as one block.
Dividing means to be divided and each block from the blocking means.
The image data of rock is transformed by orthogonal transformation using the cosine function.
Discrete cosine transform means for calculating a commutation coefficient, and said discrete cosine transform
High-frequency component in the horizontal direction from the conversion coefficient calculated by the means
And the region that contains the vertical high-frequency component.
For each region of the region that contains the high-frequency component in the oblique direction
Region coefficient detecting means for detecting a coefficient;
More than the threshold value in each area among the transform coefficients detected in the step
Is calculated for each area, and based on the number,
A fineness detection means for detecting the level of fineness of the area, and all
If high definition is detected in the area of
Definition determining means for determining that the definition is high, and the definition
Weighting function according to the level of definition determined by the determining means
To the conversion coefficient calculated by the discrete cosine conversion means.
Weighting factor multiplying means for multiplying, and the weighting factor multiplying means.
From the multiple blocks.
Quantization using the same quantizer in a processing unit
And conversion means. [0012] [0013] [0014] In the image encoding device according to the present invention, the image data
Block with n × n blocks in the spatial arrangement as one block
The image data of each block using the cosine function.
Orthogonal transform to calculate a conversion coefficient, and
Area that contains the high-frequency component in the horizontal direction and the height in the vertical direction.
Intrinsic region and oblique high-frequency component
Conversion coefficients for each area of the
The number of transform coefficients above the threshold in each area
Determined for each area, and based on the number, the degree of definition of each area
Is detected, and if high definition is detected in all regions,
If it is determined that the definition of the block is high,
The conversion coefficient obtained by multiplying the weight coefficient according to the level of the definition is
The same quantizer is used for the processing unit consisting of multiple blocks.
And quantize. [0015] [0016] [0017] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The following is an embodiment of an image coding apparatus according to the present invention.
Will be described with reference to the drawings. FIG. 1 applies the present invention
FIG. 2 shows a circuit configuration of an image encoding apparatus according to the first embodiment.
Is a digital video
-Precoder (hereinafter simply referred to as VTR) recording system circuit
FIG. 3 shows a circuit of a reproduction system of a VTR.
2 shows the configuration. First, the VTR will be described. this
The VTR converts the analog video signal as shown in FIG.
Digital signal, and convert the resulting image data into a so-called conversion code.
After performing data processing such as encryption and data compression,
Recording system for recording on magnetic tape 1 via air head 21
And, as shown in FIG.
1 to binarize the reproduction signal reproduced by
After performing data processing such as encoding, it is converted to an analog signal.
And a playback system that plays back analog video signals.
You. The above recording system, as shown in FIG.
Sampling the video signal and converting it to a digital signal
An analog / digital converter that forms image data
Lower A / D converter) 11 and the A / D converter 11
These image data are divided into n × n
Image block G_h(H = 0 to H, H is 1 frame
Number of pixels in one frame or one field and one image block
Number of pixels n^TwoCircuit 12 divided into
And the image data from the blocking circuit 12 as a cosine function
Orthogonal transform (hereinafter DCT: Discrete Cosine Tran)
sform) for each image block G_hConversion coefficient C_ij
Discrete remainder for calculating (i = 0 to n-1, j = 0 to n-1)
A string transformation circuit (hereinafter referred to as a DCT circuit) 13 and the DCT circuit
Conversion coefficient C from road 13_ijTo a plurality of image blocks G_h
, For example, for each sync block that is one unit of transmission
Circuit 14 which quantizes data to form quantized data
And the quantized data from the quantizing circuit 14
Coded data VLC encoded by a so-called variable length code
_ijCode forming (i = 0 to n-1, j = 0 to n-1)
Encoding circuit 15 and encoded data V from the encoding circuit 15
LC_ijFor example, Paris for error detection and error correction
Parity adding circuit 1 for adding a parity to each sync block
7 and the parity from the parity adding circuit 17 are added.
Encoded data VLC_ijSync signal, etc.
Synchronization signal insertion time that is added for each
Path 18 and the parallel data from the synchronization signal insertion circuit 18.
Converts transmission data sent as serial data to serial data
Parallel / serial (hereinafter referred to as P / S) converter 1
9 suitable for recording on the transmission data from the P / S converter 19.
The modulation is performed to generate a recording signal.
1 channel encoder (hereinafter referred to as ENC)
C) 20). The recording system is connected to the
Convert video signals supplied as analog signals to image data
After the conversion, for example, for one frame or one field
Convert image data to image block G_hDivided into image blocks
Ku G_hDCT of the image data of_ijCalculate
And the conversion coefficient C_ijIs quantized for each sync block
While forming the quantized data, the variable length code
Encoded data VLC by encoding encoded data_ijForm
It has become. Also, this recording system uses encoded data
VLC_ijA sync signal is added to each sync block for transmission.
After forming the transmission data, the transmission data
Perform modulation, for example scrambling or NRZI modulation processing,
The recording is performed on the magnetic tape 1 by the magnetic head 21.
ing. Thus, an image coding apparatus according to the present invention,
That is, the main part of the VTR configured as described above is
A block circuit 12 to a quantization circuit 14;
Physically, it looks like this: The blocking circuit 12 is, for example, one frame.
Or a memory with a recording capacity for one field?
And, for example, as shown in FIG.
For example, supplied as so-called luminance signal Y and color difference signals U, V
Image data is sequentially stored and arranged in a spatial arrangement as described above.
Of n × n, for example, 8 × 8 in one block
Image block G_hAnd read it out to the DCT circuit 13.
Supply. The DCT circuit 13 includes, for example, a so-called DSP (Di-
gital Signal Processor) etc.
Image block G from circuit 12_hImage data supplied for each
Is orthogonally transformed using the cosine function as described above.
Number C_ijIs calculated, and the conversion coefficient C_ijTo the quantization circuit 14
Supply. The quantizing circuit 14 is provided as shown in FIG.
The conversion coefficient C from the DCT circuit 13_ijHorizontal direction
High frequency component, vertical high frequency component, diagonal high frequency component
Based on each image block G_hThe so-called definition of
A)_h(H = 0 to H)
Activity detection circuit 51 and the activity detection circuit
Activity A from 51_hWeighting factor K based on_h
(H = 0 to H), a weight coefficient generation circuit 52,
Weight coefficient K from weight coefficient generation circuit 52_hWith the above DCT
Conversion coefficient C from circuit 13_ijImage block G_hEvery power
And a multiplier 53 having different quantization widths from each other.
The weighted conversion coefficient (K_h× C
_ij) Are quantized to form the same image block G_hTo
To form different quantized data
The resulting quantizer Q_m(M = 1 to M). The activity detection circuit 51
Conversion coefficient C_ijOf the region that contains the horizontal high-frequency component of
Transformation of regions with large numbers and vertical high-frequency components
The relationship between the magnitude of the coefficient and the area that contains the high-frequency component in the oblique direction
Detects the magnitude of a number and calculates the logical product of those coefficients.
Find each image block G_hActivity A_hDetect
It is supposed to. More specifically, the activity detection circuit 51
Is, for example, as shown in FIG._ijHorizontal, vertical
Gate circuit 61 for separating each high-frequency component in the direct and oblique directions
And high-frequency components in each direction from the gate circuit 61.
62h, 6 which are respectively compared with the threshold value TH in
2v, 62d and the outputs of the comparators 62h, 62v, 62d.
For example, the number of high level (hereinafter referred to as H level)
Counters 63h, 63v, 63d for detecting
Each output of the counters 63h, 63v, 63d (hereinafter referred to as a counter)
Value) is compared with the threshold value th in the corresponding direction.
Devices 64h, 64v, 64d and the comparators 64h, 64
AND gate 65 for calculating the logical product of each output of v and 64d
It is composed of The gate circuit 61 is, for example, as shown in FIG.
As shown, the image from the DCT circuit 13
Block G_hConversion coefficient C supplied for each_ijThe whole area 70
Is a size of, for example, 4 × 3 which contains a high-frequency component in the horizontal direction.
Region 71 and a high-frequency component in the vertical direction.
The region 72 of 3 × 4 size and the high frequency component in the oblique direction
Is divided into existing areas 73 having a size of 4 × 4, for example.
Conversion coefficient C of 71_ij(I = 4-7, j = 0-2)
And the conversion coefficient C in the area 72_ij(I = 0 to
2, j = 4 to 7) are supplied to the comparator 62v, and the
Conversion coefficient C_ij(I = j = 4 to 7) is supplied to the comparator 62d.
I do. The comparator 62h calculates the conversion coefficient C of the area 71._ij
Is the horizontal threshold TH_hAnd the conversion coefficient C_ijIs large
When the threshold is high, for example, an H level is output and the comparator 62v
Is the conversion coefficient C of the area 72_ijIs the threshold TH in the vertical direction._vWhen
By comparison, the conversion coefficient C_ijIs large, for example, H level
And the comparator 62d outputs the conversion coefficient C_ij
Is the threshold value TH in the oblique direction._dAnd the conversion coefficient C_ijIs large
When the threshold is high, for example, an H level is output. The counters 63h, 63v and 63d compare
The number of H level of each output of the devices 62h, 62v and 62d is
Each of them is counted, and each count value is compared with a comparator 64
h, 64v, 64d. That is, the counter 6
3h, the threshold value TH in the area 71_hThe above conversion staff
Number C_ijAre output, and the counter 63v indicates that the area 7
Threshold value TH at 2_vThe above conversion coefficient C_ijNumber of output
From the counter 63d, the threshold value TH in the area 73 is obtained.
_dThe above conversion coefficient C_ijIs output. The comparators 64h, 64v, 64d are
The respective count values from the data 63h, 63v, 63d correspond to the count values.
Threshold th_h, Th_v, Th_dCompare with each
When the count value is large, for example, output the H level
I do. That is, from the comparator 64h, the region 71
Threshold TH_hThe above conversion coefficient C_ijIs the threshold th_hthat's all
, The H level is output, and the comparator 64v outputs
Threshold TH in region 72_vThe above conversion coefficient C_ijNumber of thresholds
Value th_vAt this time, the H level is output and the comparator 64
From d, the threshold value TH in the area 73 is obtained._dConversion factor above
C_ijIs the threshold value tth_dIn the above case, the H level is output.
It is. Specifically, for example, as shown in FIG.
Image block G_hImage data area 90 is dark on the right side
The area 91 (the shaded area represents darkness) and the bright area on the left side
Image block having a vertical edge
Ku G_hThen, the conversion coefficient C_ijMany horizontal high-frequency components
That is, the threshold value TH is set in the area 71._hThe above conversion coefficient C
_ijOccur, and the threshold value TH is set in the area 73._dthat's all
Conversion coefficient C_ijAre generated, and the comparators 64h and 64d
Outputs an H level signal, and the comparator 64v outputs an L level signal.
A level signal is output. Also, for example, as shown in FIG.
Rock G_hImage data area 90 is the upper dark area 9
3 and lower bright area 94, horizontal edge
Image block G having_hThen, the conversion coefficient C_ijVertical
The high-frequency component in the direction
_vThe above conversion coefficient C_ijAre generated, and the area 7
The threshold value TH is 3_dThe above conversion coefficient C_ijMany occur, the ratio
H level signals are output from the comparators 64v and 64d.
An L level signal is output from comparator 64h. For example, an image block G having a complicated pattern
_hOr, for example, as shown in FIG._hof
The image data area 90 is obliquely above the dark area 95 on the upper side.
It consists of a lower bright area 96 with diagonal edges.
Image block G_hThen, the conversion coefficient C_ijHorizontal and vertical
The high-frequency component in both directions is large,
The thresholds TH are respectively assigned to the areas 71, 72, 73_h, TH_v, T
H_dThe above conversion coefficient C_ijOccur in all comparators
H level signals are output from 64h, 64v, 64d
You. The AND gate 65 is obtained as described above.
Of the outputs of the comparators 64h, 64v, 64d
Ask for. As a result, from the AND gate 65, a complicated
Image block G with picture_hWith diagonal edges
Image block G_hH level signal is output
Image blocks with horizontal or vertical edges
Gook_h, An L-level signal is output. Soshi
The AND gate 65 is thus detected.
Signal at the H level_hIs high
Activity A at the L level_hIs low
The weight coefficient generation circuit 52 via the terminal 67.
Supply. The weight coefficient generation circuit 52 performs the
Tee A_hActivity A based on_hIs low
Large, activity A_hWhen the weight is high
Number K_hAnd the weight coefficient K_hIs supplied to the multiplier 53.
I do. The multiplier 53 has a conversion coefficient C_ijWeight coefficient K_hThe bro
Conversion coefficient C multiplied by each weight and multiplied by a weight coefficient_ijTo
Quantizer Q_mTo supply. Quantizer Q_mHave different quantization widths
And an image block, for example, as shown in FIG.
G_hConversion coefficient C_ijRegion 100 into 16 regions 101
１１６116, and the higher frequency components are coarsely quantized. Concrete
Specifically, for example, as shown in Table 1 below, the quantizer Q₁
Is the conversion coefficient C in the regions 101 to 103_ij1/2
And transform coefficients C in the regions 104 to 106_ijTo 1 /
4 times, and transform coefficients C in the regions 107 to 108_ij1
/ 6 times, and transform coefficients C in regions 109 to 111_ijTo
Multiply by 1/8 and transform coefficient C in area 112_ijIs 1/1
0, and transform coefficients C in regions 113 to 116_ij1
After multiplying by / 16, quantization is performed with a predetermined quantization width q.
The quantizer Q_TwoIs converted in the areas 101 to 103
Coefficient C_ijIs multiplied by 、, and the conversion coefficient C
_ijIs multiplied by 1/4, and the conversion coefficients are
C_ijIs multiplied by 1/6, and the conversion
Number C_ijIs multiplied by ８, and the conversion coefficient C_ij
Is multiplied by 1/10, and the conversion coefficients are calculated in regions 113 to 115.
C_ijIs multiplied by 1/16, and the conversion coefficient C_ij
Is multiplied by 1/32 and then quantized with a predetermined quantization width q.
The same image block G_h
Quantized data with different data amounts
Formed. Then, these quantized data are encoded
Supply to road 15. [0037] [Table 1]The encoding circuit 15 operates as shown in FIG.
In addition, the quantizer Q_mEach with a different amount of data from
Each of the quantized data is encoded by a variable length code,
Codes with different data amounts for the same sync block
Encoded data VLC_ijEncoders COD respectively forming_m
(M = 1 to M) and each encoder COD_mEncoding from
Data VLC_ijRespectively, and have a predetermined storage capacity.
Buffer memory BUF_m(M = 1 to M) and
Buffer memory BUF_mEncoding read from each
Data VLC_ijSelector 54 for selecting one of
Each buffer memory B_mIs detected by detecting the overflow of
The selector 54 is provided by a quantizer selection signal described later.
And a control circuit 55 for controlling The encoding circuit 15 performs each quantization.
Container Q_mQuantized data of different data amount from
For example, so-called Huffman code and run-length
Code by Run Length code
Codes with different data amounts for the same sync block
Encoded data VLC_ijAnd each of these encoded data
Data VLC_ijTo the buffer memory BUF_mRemember each
And each of these buffer memories BUF_mOh no
Detects bar flow, does not cause overflow, and
Quantizer Q for maximum data amount_mThe quantity for choosing
The quantizer selection signal, that is, the quantizer Q_mSelect number m
To the encoded data selected by the selector 54.
Data VLC_ijThrough the terminal 5 to the Paris shown in FIG.
Output to the tee adding circuit 17. This result
As a result, the encoding circuit 15 outputs the data of the sync block.
Data amount is within the specified amount and the data amount is maximized.
Coded data VLC quantized with the minimum quantization width_ij
Is output. In other words, a predetermined number of image blocks G_h
And the fixed length of the sink block
Conversion coefficient C within the range of fixed length data capacity_ijThe most detailed
Coded data VLC obtained by quantization_ijIs output
It is. Synchronization signal insertion with parity addition circuit 17
The circuit composed of the circuit 18 is as shown in FIG.
A parity generator 56 for generating parity and a synchronization signal
A synchronizing signal generator 57 for generating a signal and an ID,
Activity detection circuit 51, selector 54-generation of synchronization signal
Activity A respectively supplied from the device 57_h,amount
Child device Q_mNumber m, encoded data VLC_ij, Parite
MUX58 for time division multiplexing of synchronization signal and ID
It is composed of From the MUX 58, for example,
One sync block includes a synchronization signal, an ID,
Quantizer Q used in sync block_mNumber m, each
Image block G_hActivity A_h, A certain number of images
Block G_hEncoded data VLC_ijFrom parity
Is output. As described above, in this image coding apparatus,
The image data supplied via the terminal 4 is arranged in a spatial arrangement.
N × n image blocks G_hSplit into each image
Block G_hAfter DCT of the image data of
Conversion coefficient C_ijWith a predetermined number of image blocks G_hShin consisting of
Block is fixed length and the amount of data allowed
Quantizer Q with minimum quantization width within_mAnd quantized using
Variable length coding of the obtained quantized data
When forming and transmitting this transmission data via the terminal 5
And each image block G_hActivity A_hBased on
For example, activity A_hWhen weight is low, large weight
Coefficient K_hAnd activity A_hSmall when high
Weight coefficient K_hTo the image block G_hConversion coefficient C for each_ij
Is multiplied and quantized, so that all the
Image block G_hThe same quantizer Q_mFor
Activity A_hLow image block G_hof
Conversion coefficient C_ijCan be relatively finely quantized,
Visually distinguish block distortion and quantization distortion during playback
Good image quality can be obtained. In particular, as described above, each image block G_h
Activity A_hActivity A based on_h
High image block G_hIs coarse, activity A_hBut
Low image block G_hIs finely quantized
But activity A_hWhen detecting
Region 71 with high-frequency component inside, vertical high-frequency component inside
Region 72 and the region in which the high-frequency component in the oblique direction exists.
In each area of 73, the conversion coefficient C equal to or more than the respective threshold_ijof
Number and detect each image block based on their intersection.
Ku G_hActivity A_hTo detect
For example, as shown in FIGS. 6a, 6b
Is an image block G having edges only in the vertical direction_hTo
Then Activity A_hCan be detected lower
Image blocks G with complex patterns_hFor
Activity A_h6a and 6b can be detected.
Image block G with a monotonous pattern as shown_hFinely quantified
To prevent so-called mosquito noise
can do. For example, as shown in FIG.
Block G with Directional Edge_hThen, I mentioned above
Sea urchin A_hIs detected high and coarsely quantized
However, since the diagonal component is visually inconspicuous,
It doesn't matter. Here, activity A_hThe other side of detection
The method will be described. In the above embodiment, the regions 71 and 7
Conversion coefficients C equal to or larger than the threshold value in 2, 73_ijBased on the number of
Activity A_hIs detected, but the
In the regions 71, 72 and 73, the sum of the absolute values of
Each is calculated, and the sum of the absolute
And the activity is determined by the logical product of the comparison results in each area.
Tee A_hTo be detected. As a result, the implementation
As in the example, image blocks with horizontal or vertical edges
Rock G_hExcept for complex patterns with high-frequency components
Image block G_hCan be detected, and the above-described embodiment and
Similar effects can be obtained. Next, the reproduction system of this VTR will be described.
I do. This reproducing system, as shown in FIG.
Signal reproduced by the magnetic head 31 from the loop 1
Is subjected to signal processing such as NRZI demodulation to transmit data
Channel decoder (hereinafter simply referred to as DEC)
) 32 and transmitted from the DEC 32 as serial data.
To convert incoming transmission data into parallel data.
An al / parallel (S / P) converter 33;
When synchronizing transmission data from S / P converter 33,
Both are encoded data VLC_ijTo play the sync signal
Path 34 and the encoded data VLC_ijOccurs when playing
A time axis correction circuit (hereinafter, TBC:
Time Base Corrector) 35 and TBC 35
Coded data VLC_ijError correction,
Encoded data VLC that could not be corrected_ijError
An error correction circuit 36 for setting a flag EF;
-Variable-length coded when recording from the correction circuit 36
Encoded data VLC_ijAnd decode the quantized data again.
The decoding circuit 37 to generate and the quantum from the decoding circuit 37
Conversion processing by subjecting the quantized data to signal processing such as inverse quantization_ij
And an inverse quantization circuit 38 for reproducing the
Conversion coefficient C_ijImage data by orthogonal transformation of
Inverse discrete cosine transform circuit (hereinafter referred to as IDCT circuit) 39
And the image block G from the IDCT circuit 39._hSupply each
1 frame or 1 field from image data
A reverse blocking circuit 40 for forming image data of
Based on the error flag EF from the error correction circuit 36
Error in the image data from the deblocking circuit 40
An error correction circuit 41 for performing correction, and the error correction circuit 4
Converts image data from 1 to analog signal and outputs
Digital / analog converter (hereinafter referred to as D / A converter)
) 42. Next, of the reproduction system configured as described above,
The operation will be described. DEC32 is magnetic tape 1
Binarize the reproduction signal reproduced by the magnetic head 31
After performing NRZI demodulation, for example,
The transmission data is reproduced by performing
Is supplied to a synchronization signal detection circuit 34 via an S / P converter 33.
Pay. The synchronizing signal detection circuit 34 is provided with the S / P converter 3
From the transmission data converted to parallel data in step 3
Signal, the synchronization is pulled in, and the encoded data VL
C_ijTo reproduce the encoded data VLC_ijTo TBC35
To supply. The TBC 35 is an encoded data VLC_ijtime
Perform inter-axis correction to absorb fluctuations in the time axis that occur during playback
And the time-axis corrected encoded data VLC_ijThe Ella
To the correction circuit 36. The error correction circuit 36 outputs the encoded data VL
C_ijError correction using the parity added when recording
And has an error that exceeds the error correction capability.
Encoded data VLC_ijError flag EF
And the error-corrected coded data VLC_ijDecrypt
To the conversion circuit 37. The decoding circuit 37 outputs a Huffman code
Encoding encoded by a signal and a run-length code
Data VLC_ijTo reproduce the quantized data,
Is supplied to the inverse quantization circuit 38. The inverse quantization circuit 38 generates the encoded data VLC
_ijQuantizer Q reproduced together with_mBased on the number m of
Quantizer Q used for recording_mRecognizes this quantum
Chemistry Q_mInverse quantization of quantization data with quantization width corresponding to
And the encoded data VLC_ijPlayed with
Activity A_hIs multiplied during recording based on
Each image block G_hWeighting factor K_hRecognize and quantize
Weighted coefficient K_hMultiply the reciprocal of
C_ijAnd the conversion coefficient C_ijTo the IDCT circuit 39
Supply. The IDCT circuit 39 has a conversion matrix for recording.
Using the transpose of_ijBy orthogonal transformation
Data to image block G_hPlayback each time, this image data
It is supplied to the inverse blocking circuit 40. The deblocking circuit 40 generates the image block G
_hOne frame or one frame from the image data reproduced every time
Error correction circuit 4 for forming image data for the field
Feed to 1. The error correction circuit 41, for example,
Image data that could not be corrected by the error correction circuit 36.
Interpolation using error-free image data near the data
Of the image data that could not be corrected
Error correction is performed, and the image data with this error corrected
Is supplied to the D / A converter 42. The D / A converter 42 outputs the error-corrected image.
Converts image data into analog signals, and
Video signals as a luminance signal Y and color difference signals U and V, for example.
Output. As described above, when recording, the sync block
All image blocks G_hThe same quantizer for
Q_mUsing the conversion coefficient C_ijIs quantized into an image block G
_hActivity A_hActivity A based on
_hLow image block G_hConversion coefficient C_ijIs relatively thin
Activity A_hHigh image blocks
G_hConversion coefficient C_ijIs relatively coarsely quantized,
And the quantizer Q_mNumber m and each picture
Image block G_hActivity A_hRecord
By using this information during playback,
Block reproduction and quantization distortion
Plays a video signal with good image quality that is not visually noticeable
Can be In particular, activity A when recording_h
Detection based on horizontal, vertical, and diagonal high-frequency components.
That is, in three regions 71, 72, 73
Conversion coefficient C above threshold_ijThe number of, or the converter for each area
Number C_ijBased on the sum of absolute values of
, The conversion coefficient C equal to or larger than the threshold_ijNumber or conversion factor
C_ijHigh activity A when the sum of absolute values of
_hBy setting the horizontal or vertical edge
Image block G having_hExcluding high frequency components
Image block G with complicated pattern_hCan detect
Mosquito noise can be prevented. As is clear from the above description, the present invention
Then, the image data is divided into n × n pieces in the spatial arrangement by one block.
Image data of each block
Is orthogonally transformed using a cosine function to calculate a transform coefficient.
For the transform coefficients of
Area, the area that contains the high-frequency component in the vertical direction, and the
Detects conversion coefficients for each region of the region containing high-frequency components
Of the detected conversion coefficients,
Is calculated for each area, and based on the number,
Detects the degree of definition of the area. And in all areas
If high definition is detected, the definition of the block
Is determined to be high, and the weight according to the level of the determined fineness is determined.
Process of transforming multiplied coefficients into multiple blocks
By quantizing using the same quantizer in units,
Quantize transform coefficients of blocks with low definition relatively finely
Visualization of block distortion and quantization distortion
In order to obtain good image quality.
it can. [0059]

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram illustrating a circuit configuration of an embodiment of an image encoding device to which the present invention has been applied. FIG. 2 is a block diagram showing a circuit configuration of a recording system of a digital video tape recorder to which the image encoding device is applied. FIG. 3 is a block diagram showing a circuit configuration of a reproduction system of a digital video tape recorder to which the image encoding device is applied. FIG. 4 is a block diagram illustrating a circuit configuration of an activity detection circuit included in the image encoding device. FIG. 5 is a diagram illustrating a conversion coefficient area for explaining the operation of the activity detection circuit; FIG. 6 is a diagram schematically illustrating a picture image block having an edge. FIG. 7 is a diagram illustrating a region of a transform coefficient for explaining an operation of a quantizer of an activity detection circuit included in the image encoding device. [Description of Reference Numerals] 12, 62 ... blocking circuit 13 ... DCT circuit 14 ... quantization circuit 51 ... activity detecting circuit Q _m ... quantizer

Continuation of front page (56) References JP-A-2-105792 (JP, A) JP-A-2-84894 (JP, A) Kazuyuki Murata, "DCT coding method using adaptive quantization-digital Application to Color Copier-", Journal of the Institute of Image Electronics Engineers of Japan, Japan, The Institute of Image Electronics Engineers of Japan, October 25, 1991, Vol. 20, No. 5, p. 467-475 (58) Field surveyed (Int.Cl. ⁷ , DB name) H04N ^7/ 24-7/68 H04N 1/41-1/419

Claims (1)

(57) [Claim 1] Blocking means for dividing image data into blocks of n × n blocks in a spatial arrangement, and image data of each block from the blocking means. A discrete cosine transform unit for performing an orthogonal transform using a cosine function to calculate a transform coefficient; and an area having a horizontal high-frequency component and a vertical high-frequency component from the transform coefficient calculated by the discrete cosine transform means. a domain coefficient detecting means for detecting the transform coefficients of each region in the region underlying the high frequency component of the intrinsic region and the oblique direction, each region of the transform coefficients detected by the region coefficient detection means
The number of conversion coefficients equal to or larger than the threshold in
To detect the degree of detail of each area based on the number of
Detecting means and, if high definition is detected in all the regions,
And a weight for multiplying the conversion coefficient calculated by the discrete cosine conversion means by a weight coefficient corresponding to the level of the fineness determined by the fineness determination means. Coefficient multiplying means, and transform coefficients multiplied by the weighting coefficients by the weighting coefficient multiplying means , using the same quantizer in a processing unit consisting of a plurality of blocks.
And a quantizing means for performing quantization by using the image encoding apparatus.