JPH0787586B2 - Image signal coding control method - Google Patents

Description

The present invention relates to an image signal coding control system for highly efficiently coding image signals in a video conference, a videophone, etc., and relates to a code allocation that improves coding efficiency and considers playback image quality. And a transform cosine transform unit for performing a discrete cosine transform on an input image signal or an inter-frame difference signal of the input image signal, and a transform coefficient from the discrete cosine transform unit or a transform coefficient of the transform cosine transform unit. A quantizer for quantizing the inter-frame difference signal, an encoder for variable-length coding the quantized output signal from the quantizer, and an image determiner for determining the property of the input image signal, The image determination unit causes a flat portion, an edge portion,
It is configured to determine which of the fine pattern portions corresponds and to control the quantization step of the quantization unit based on the determination result.

[Industrial application field]

The present invention relates to an image signal coding control system for highly efficiently coding an image signal in a video conference, a video telephone and the like.

Various types of high-efficiency coding schemes for image signals have already been proposed, and recently, the discrete cosine transform (DCT) has attracted attention as one of them. In this discrete cosine transform, a plurality of pixels are set as one block, and the transform coefficient is orthogonally transformed into a transform coefficient by a transform matrix. Generally, the transform coefficient tends to concentrate on the low frequency region side, although it varies depending on the property of the image. In this case, the transform coefficient is quantized and variable-length coded to perform high-efficiency coding of the image signal, and the coded signal is transmitted or recorded in the image recording device. Then, at the time of reproduction from the receiving side or the image recording device, variable length decoding, inverse quantization, and further inverse discrete cosine transform are performed to obtain a reproduced image signal.

It is desired to further improve the coding efficiency in the method of coding such an input image signal by performing the discrete cosine transform.

[Conventional technology]

FIG. 15 is a block diagram of a conventional example, the transmitting unit includes a discrete cosine transform unit 60, a quantizing unit 61, a scanning unit 62 and an encoding unit 63, and the receiving unit includes a decoding unit 64 and reverse scanning. It includes a unit 65, an inverse quantization unit 66, and an inverse discrete cosine transform unit 67.

The input image signal is an image signal obtained by capturing a moving image or a still image such as a drawing with a television camera, an image signal reproduced from an image recording device, or the image signal subjected to interframe difference processing or the like. This signal is a signal converted into a digital signal by an AD converter (not shown), and is added to the discrete cosine conversion unit 60. In the discrete cosine transform unit 60, for example, as shown in FIG.
The image f of each block is divided into blocks of N pixels.
The discrete cosine transform is applied to (u, v), and the transform coefficient F (i, j)
Are quantized individually in the quantizing unit 61, and the scanning unit 62 scans the quantized output signals in a two-dimensional array into a one-dimensional array by scanning in the horizontal direction, the vertical direction, or the zigzag direction. In this case, entropy coding is performed by variable-length coding or the like in which short codes are assigned to those with a high appearance probability and long codes are assigned to those with a low appearance probability.

On the receiving side, the decoding unit 64 entropy-decodes the received code by a process reverse to that of the encoding unit 63, and the inverse scanning unit 65 performs a scanning process reverse to that of the transmitting side scanning unit 62, and an inverse quantizing unit. 66
, And the inverse cosine transform section 67 performs the reverse process of the discrete cosine transform section 60 to produce an output image signal, which is added to a display device (not shown) for transmission. The image from the side is displayed.

The discrete cosine transform is expressed by the following equations (1) and (2).

When the pixel f (u, v) of each block in one screen is converted into the conversion coefficient F (i, j) based on the equations (1) and (2), it corresponds to the value of i, j. It shows the frequency component,
The smaller the value of i, j, the lower the frequency component, and the F of i, j = 0
(0,0) indicates the DC component. For example, in the block consisting of N × N pixels on the screen of FIG. 16, N = 8
Then, the shaded block is enlarged and shown below, and by performing the discrete cosine transform, the transformation coefficient in the low frequency region in the arrow a direction and the transformation coefficient in the high frequency region in the arrow b direction are obtained. And for a typical screen,
By performing the discrete cosine transform, the value of the conversion coefficient of the low-frequency component is large, the value of the conversion coefficient of the high-frequency component is small, and many values are 0.

This transform coefficient F (i, j) is quantized in the quantizer 61 and scanned in the scanner 62. In the lower block of FIG. 16, c is a zigzag scan and d ( solid line)
Indicates horizontal scanning, and e (dashed-dotted line) indicates vertical scanning. Any one of them scans from a low frequency region to a high frequency region or in the opposite direction, and a two-dimensional array is a one-dimensional array. , And the encoder 63 sequentially performs entropy coding.

As described above, since the high frequency component of the conversion coefficient is often 0, scanning is performed from the high frequency region side to
The end-of-bound code EOB is added to the position where is no longer. For example, with respect to the quantized signal of the block of 8 × 8 pixels shown in FIG. 17, zigzag scanning is performed from the high frequency region side as shown by the arrow, and the position where it is not 0 (position of value −7: upper left pixel position Is 1, and the lower right pixel position is 64, the end-of-bound code EOB is added to the pixel position "46").

Then, when zigzag scanning is performed from the low frequency region side, 1031
(DC component), -9, -5,3,20,15,0, ... 0, -7 (EOB)
The transform count of the one-dimensional array is obtained, entropy-encoded in the encoding unit 63, and all the transform coefficients after the end-of-bound code EOB are 0, so the encoding is omitted. The coding efficiency is improved.

On the receiving side, by detecting the end-of-bound code EOB, the end of the coded signal of the block can be identified and decoding can be performed.

[Problems to be Solved by the Invention]

By applying the discrete cosine transform to the block corresponding to the input image signal, the transform coefficients concentrated on the low frequency region side are obtained, and 0 is increased on the high frequency region side. Therefore, the end of bound code on the high frequency region side is obtained. 0 after EOB
, So that the coding efficiency is improved.

However, in the conventional method of performing image signal coding using discrete cosine transform, high efficiency coding is uniformly performed regardless of the nature of the image to be transmitted or recorded. Since high-efficiency coding is not performed, there is a drawback that reproduction image quality is deteriorated when the coding efficiency is simply improved.

It is an object of the present invention to improve coding efficiency and enable code allocation in consideration of reproduction image quality.

[Means for Solving the Problem] The image signal coding control method of the present invention is for determining the property of the input image signal and controlling the quantization step, and will be described with reference to FIG.

Discrete cosine transform unit 1 for performing a discrete cosine transform on an input image signal or an inter-frame difference signal of the input image signal
And a quantizing unit 2 for quantizing the transform coefficient from the discrete cosine transform unit 1 or the inter-frame difference signal of the transform coefficient.
And an image determination unit 4 for determining the property of the input image signal, and an encoding unit 3 for variable-length encoding the quantized output signal from the quantization unit 2. It is determined whether a flat portion, an edge portion, or a fine pattern portion corresponds to a block composed of a plurality of pixels of the image signal, and the quantization step of the quantization unit 2 is controlled based on the determination result. .

Furthermore, a motion compensating unit for compensating motions between the frames of the input image signal is provided, and the image judging unit 4 controls the quantizing unit 2.

When the image determination section 4 determines that the area is flat,
The quantization step of the quantizer 2 is made fine, and when it is determined to be an edge portion, it is equal to or coarser than the quantization step in the case of a flat portion, and when it is determined to be a fine pattern portion, the edge portion is The quantization step is set to be equal to or coarser than the quantization step.

Further, the image determination unit 4 determines that the standard deviation of the pixel values of the block is a flat portion when the standard deviation is equal to or less than the first threshold value, and
When it exceeds the second threshold and is less than or equal to the second threshold, it is determined to be an edge portion, and when the second threshold is exceeded, it is determined to be a fine pattern portion.

Further, the image determination unit 4 applies a Laplacian mask to the pixels of the block, and as a result, when the number of pixels exceeding the predetermined threshold is less than or equal to the third threshold, it is determined to be a flat portion, and the third When it exceeds the threshold value and is equal to or less than the fourth threshold value, it is determined as an edge portion, and when it exceeds the fourth threshold value, it is determined as a fine pattern portion.

The image determination unit 4 also obtains an average value of the pixel values of the block or the sub-block obtained by subdividing the block, and the absolute value of the difference between the average value and the pixel value of the block or the sub-block is a predetermined value. The number of pixels exceeding the threshold value is counted, and when the number of pixels is less than or equal to the fifth threshold value, it is determined as a flat portion, and when the number of pixels is greater than the fifth threshold value and less than or equal to the sixth threshold value, it is determined as an edge portion, Further, when the number of pixels exceeds the sixth threshold value, it is determined as a fine pattern portion.

[Action]

The discrete cosine transform unit 1 performs a discrete cosine transform on the input image signal or a discrete cosine transform on the inter-frame difference signal, and the quantization unit 2 quantizes the transform coefficient. The quantizer 2 has a configuration for switching the quantization step or a configuration for switching a quantizer having a plurality of different quantization steps, and the switching is controlled by the image determination unit 4. The coding unit 3 also performs variable length coding such as Huffman coding on the quantized output signal.

The image determination unit 4 determines which of a flat portion such as a sky or a wall, an edge portion such as an outline portion of an object, or a fine pattern portion such as a pattern corresponds to a block corresponding to an input image signal. Then, the quantization step of the quantization unit 2 is switched.

Further, the motion compensation unit detects a motion vector between the block of the input image signal of the current frame and the block of the input image signal of the previous frame, and the inter-frame difference signal for the block having large motion in the image signal is detected. It does not increase.

Also, the visual characteristics have good sensitivity to flat parts and edge parts, and relatively low sensitivity to fine pattern parts,
Even if there is some distortion, it cannot be identified. Also, if the quantization step in the quantizer 2 is made rough, the reproduced image will be blurred, and if the high-frequency component of the transform coefficient obtained by performing the discrete cosine transform is forced to zero, block distortion occurs. Tend. Therefore, the image determination unit 4
It is judged for each block of the input image signal, the quantization step is finely divided in the case of a flat portion, the quantization step is the same as or coarser than that of the flat portion in the case of an edge portion, and the quantization step is edged in the case of a fine pattern portion. The setting is similar to or coarser than the part. As a result, the coding efficiency can be improved without deteriorating the reproduced image quality.

In addition, the image determination unit 4 determines, in correspondence with the block of the input image signal, which one of a flat portion, an edge portion, and a fine pattern portion, based on the values (luminance) of the pixels forming the block. As one of the means, first, the standard deviation is obtained with respect to the pixel value of the block of the input image signal.
This standard deviation is compared with the first and second threshold values, and the first
If it is less than or equal to the threshold value, it is composed of relatively average pixels, and if it exceeds the second threshold value, it is a flat part such as a sky or wall. If it exceeds the second threshold value, it is composed of pixels of different values. A fine pattern part such as a pattern, which exceeds the first threshold and the second
When the threshold value is less than or equal to the threshold value, the pixel configuration is intermediate between the flat portion and the fine pattern portion and includes the contour portion such as characters and figures.

Another means in the image determination unit 4 applies a Laplacian mask to the pixels of the block of the input image signal.
As a result, the number of pixels exceeding a predetermined threshold is detected and counted. That is, the number of pixels corresponding to edges is counted. When the number of pixels is less than or equal to the third threshold, the number of pixels corresponding to the edge is small, and thus the flat portion. When the number of pixels exceeds the fourth threshold, the number of pixels corresponding to the edge is large. If the threshold value is less than or equal to the fourth threshold value, the pixel configuration is intermediate between the flat portion and the fine pattern portion, so that the pixel is determined to be the edge portion.

Still another means in the image determination unit 4 finds the average value of the pixel values of the block or sub-block. Then, it is identified whether the absolute value of the difference between the average value and the value of each pixel exceeds a predetermined threshold value, and the number of pixels exceeding the threshold value is counted. When the number of pixels is less than or equal to the fifth threshold value, the difference in the pixel value of the block is small, and thus the pixel is determined to be a flat portion.
Further, when the number of pixels exceeds the sixth threshold value, there is a large difference in the pixel values of the block, so it is determined to be a fine pattern portion. When the number of pixels exceeds the fifth threshold value and is equal to or less than the sixth threshold value, the difference in the pixel value of the block is larger in the flat portion and smaller than the fine pattern portion, so that it is determined as the edge portion.

〔Example〕

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 2 is a block diagram of an embodiment of the present invention. On the transmitting side, a discrete cosine transform unit 11, a quantizing unit 12, and a scanning unit 15 are provided.
And an encoding unit 13, a transmission unit 16, and an image determination unit 14. The receiving side includes a receiving unit 17, a decoding unit 18, an inverse scanning unit 19, an inverse quantizing unit 20, and an inverse discrete cosine transform unit 21.

An input image signal such as a moving image signal from a television camera or the like is subjected to a block cosine transform in a discrete cosine transform section 11, a quantizing section 12 quantizes the transform coefficient, and a scanning section 15 Converts the two-dimensional array corresponding to the block into a one-dimensional array by zigzag scanning, horizontal scanning, vertical scanning, etc., performs variable length coding in the coding unit 13, and modulates in the transmission unit 16, etc. Then, the encoded signal is transmitted to the receiving side.

Further, in the image determination unit 14, it is determined whether a flat portion, an edge portion or a fine pattern portion corresponding to the block of the input image signal,
The determination signal is applied to the quantizing unit 12 and the transmitting unit 16, the quantizing step in the quantizing unit 12 is controlled, and the quantizing step is sent from the transmitting unit 16 to the receiving side as auxiliary information. In this case, as in the case of the normal entropy coding transmission method, the amount of transmission information is determined from the amount of use of the transmission buffer, and the quantization unit 12 is controlled so that the transmission buffer does not overflow or underflow. Configurations can also be added.

On the receiving side, the receiving section 17 receives the encoded signal, the decoding section 18 decodes it, and the inverse scanning section 19 converts the one-dimensional array signal into a block-corresponding two-dimensional array signal. Converted,
Inverse quantization is performed in the inverse quantization unit 20. At this time, the receiving unit 17
The auxiliary information indicating the separated quantization step is added to the quantizer 20 and inversely quantized by the same quantization step as on the transmitting side. The inverse quantized output signal is inverse discrete cosine transformed by the inverse discrete cosine transform unit 21 to obtain a received image signal,
The display is provided in addition to the display device (not shown).

The image determination unit 14 determines, for each block of the input image signal, the number of pixels having a difference from the average value or more, the standard deviation,
The flat part, the edge part,
It is determined which one of the fine pattern parts. With this determination signal, the quantization step of the quantization unit 12 is fine for the flat portion, the same as or flatter than the flat portion for the edge portion, and the edge portion for the fine pattern portion. It is the same or coarser.

FIG. 3 is a block diagram of another embodiment of the present invention, in which 31 is a discrete cosine transform unit (DCT), 32 is a quantizer (Q), and 33 is a quantizer.
Is an encoding unit (VLC), 34 is an image determination unit, 35 is an addition circuit, 3
6 is a high frequency component removing circuit (HFC), 37 is an inverse quantizer (), 38 is an adding circuit, 39 is a frame memory (FM), 40
Is a switching unit for switching between interframe coding and intraframe coding.

This embodiment shows a case where the inter-frame difference is obtained and encoded in the transform coefficient region, the discrete cosine transform unit 31 performs a discrete cosine transform on the input image signal, and the transform coefficient by this is stored in the frame memory 39. The conversion coefficient of the previous frame and the conversion coefficient of the previous frame are added to the addition circuit 35 to obtain the difference, and the quantization unit 32
Is quantized by, and the conversion coefficient of the high frequency component is deleted by the high frequency component deleting circuit 36, and the encoding unit 33 performs variable length coding and transmits to the receiving side. Alternatively, the image is recorded in addition to the image recording device.

Further, the transform coefficient from which the high frequency component has been removed by the high frequency component removal circuit 36 is inversely quantized by the inverse quantization unit 37, and the conversion coefficient of the previous frame accumulated in the frame memory 39 and the addition circuit
The addition is made at 38 and the result is stored in the frame memory 39 as the conversion coefficient of this frame.

Further, the generated information amount between the inter-frame coding and the intra-frame coding is compared, and when the generated information amount in the intra-frame coding is small, the switching unit 40 is controlled to switch from the frame memory 39 side to the ground side. In this way, intraframe coding can be performed.

Further, the image determination unit 34 determines whether the input image signal is a flat portion, an edge portion, or a fine pattern portion corresponding to a block, and the quantization unit 32 uses the determination signal.
To switch the quantization step between the inverse quantization unit 37 and the inverse quantization unit 37.

Further, the high frequency component removing circuit 36, for example, performs a zigzag scan c, a horizontal scan d, or a vertical scan e on the two-dimensional array of blocks in FIG. The quantized output signal of (N × N) / 2 transform coefficients including the component is taken out, and the subsequent high frequency components are removed and added to the encoding unit 33, so that the high frequency component of zero or a value close to zero is obtained. The transform coefficient is not coded, which also improves the coding efficiency. The high frequency component removing circuit 36 may be provided in the preceding stage of the quantizing unit 32.

Further, the image determination unit 34, similarly to the above-described embodiment, determines whether the flat portion, the edge portion, or the fine pattern portion in correspondence with the block of the input image signal, and the quantization unit 32 and the inverse quantization unit 37. It controls the quantization step.

The functions of the respective units such as the discrete cosine transform unit 31 described above can be realized by respective dedicated integrated circuits or by the arithmetic function of the digital signal processor.

FIG. 4 is a block diagram of still another embodiment of the present invention,
The case where motion compensation is performed is shown. In the figure, 41 is a discrete cosine transform unit (DCT), 42 is a quantization unit (Q), 43 is a coding unit (VLC), 44 is an image determination unit, 45 is an addition circuit, and 46 is a high frequency component deletion. Circuit (HFC), 47 is the inverse quantizer (), 48
Is an adder circuit, 49 is a frame memory (FM), 50 is a switching unit,
Reference numeral 51 is an inverse discrete cosine transform unit (IDCT), and 52 is a motion compensation unit.

The input image signal is added to the image determination unit 44, the addition circuit 45, and the motion compensation unit 52. When the switching unit 50 is switched to the motion compensation unit 52 side as shown in the figure, the motion vector is detected by comparing the input image signal and the image signal in the previous frame of the frame memory 49 in block correspondence, The difference is controlled so that the inter-difference becomes small. Switching unit 50
Is switched to the ground side, intraframe coding is performed, and when switched to the frame memory 49 side, interframe coding without motion compensation is performed. This switching unit 50
The control of the switching can be performed by comparing the amount of information generated by each encoding method and using the encoding method with the smallest amount of information generated.

The difference between the input image signal and the image signal of the previous frame is obtained by the adder circuit 45, the discrete cosine transform unit 41 performs discrete cosine transform on the difference signal, and the quantization unit 42
And the transform coefficient of the high-frequency component quantized by the high-frequency component removing circuit 46 is deleted, variable-length coding is performed in the coding unit 43, and coding is performed including auxiliary information such as a quantization step and a motion vector. The converted signal is sent to a receiving side (not shown) or recorded in the image recording device.

The inverse quantization unit 47 inversely quantizes the transform coefficient, the inverse discrete cosine transform unit 51 performs inverse discrete cosine, and the adder circuit 48 adds the image signal of the previous frame to the frame as the image signal of the current frame. It is stored in the memory 49.

The image determination unit 44 determines whether the input image signal is a flat portion, an edge portion, or a fine pattern portion corresponding to a block, and controls the quantization steps of the quantization unit 42 and the inverse quantization unit 47.

In this embodiment, when a moving image signal with a large motion is input, the coding efficiency can be improved by the motion compensation, and further, the quality of the image in that case is judged to deteriorate the reproduced image quality. The quantization step can be controlled so as not to do so.

FIG. 5 and FIG. 6 are flowcharts of the image determination by the standard deviation in the image determination unit. First, F (8,8)
As shown in, the size of the block is 8 × 8 pixels, the standard deviation is σ, the first and second thresholds are TH1, TH2, and the average value is H (a
1) First, the average value H is set to 0 (a2), I = 1 (a3) and J = 1 (a4) (the first pixel position of the block), and then the average value H = H + f (I, J). (A5), then J = J
As +1 (a6), it is determined whether or not J> 8 (a7), J> 8.
If not, the process proceeds to step (a5). If J> 8, I = I + 1 is set (a8), and it is determined whether or not I> 8 (a
9), if I> 8 is not satisfied, go to step (a4), I
In the case of> 8, since the accumulation processing in the 8 × 8 block is completed, H = H / 64 is set (a10), and the average value H is obtained.

Next, set σ = 0 (a11), I = 1 (a12) and J = 1.
As (a13), the standard deviation σ is σ = σ + (f (I, J) −
H) ² (a14), that is, the square of the difference between the pixel value and the average value is accumulated. Then, set J = J + 1 (a1
5), it is determined whether or not J> 8 (a16). If J> 8 is not satisfied, the process proceeds to step (a14). If J> 8, I = I +
1 (a17)), it is determined whether I> 8 (a18), I> 8
If not, the process proceeds to step (a13), and if I> 8, the cumulative value in the block of 64 pixels is obtained. (A19) and let this be the standard value σ.

Then, it is determined whether or not σ <TH1 (a20), and when the standard deviation σ is smaller than the first threshold value TH1, there is little difference in the values of the pixels in the block, so it is determined that there is a flat portion such as the sky or wall. judge. When the standard deviation σ is not equal to or less than the first threshold TH1, it is determined whether TH1 ≦ σ ≦ TH2 (a21). When the standard deviation σ is greater than the first threshold TH1 and less than or equal to the second threshold TH2, ,
Since there are some variations in the pixel values within the block, it is determined that the edge portion includes the outline portion of a character or a figure. When the standard deviation σ exceeds the second threshold value TH2, there are many variations in the pixel values in the block, so it is determined that the pattern portion is a fine pattern such as a striped pattern.

FIG. 7 and FIG. 8 show flowcharts of image determination by edge detection, and show the case of edge detection by Laplacian. That is, a mask M of Laplacian 3 × 3 size, a block f (8,8) of 8 × 8 pixel size, and a block f ′ (8,8) after the mask M of Laplacian M is applied.
8) The threshold TH, the third and fourth thresholds TH3 and TH4, and the COUNT of the number of edges are determined (b1), and I = 2 (b2) and J = 2 (b3). That is, the Laplacian mask M is applied to the pixels excluding the pixels around one block. The Laplacian mask M in step (b1) shows a case where the central pixel is "4", the pixels above and below and to the left and right are "-1", and the pixels in the diagonally up and down directions are "0". It is also possible to use a mask other than the configuration.

Then, the calculation shown in step (b4) is applied to the pixel f (2,2) with I = 2 and J = 2 to apply the mask M, and then J = J + 1 (b5), J> 7. It is determined whether or not (b6), and if J> 7 is not satisfied, the process proceeds to step (b4),
When J> 7, I = I + 1 is set (b7), and it is determined whether or not I> 7 (b8). When I> 7 is not satisfied, the process proceeds to step (b3), and when I> 7, 8 × 8. 6x6 in a block of pixels
The Laplacian mask M has been applied to the pixels, and the number of edges is counted next.

That is, the number of edges COUNT in the initial state is set to 0 (b9), and I
= 2 (b10) and J = 2 (b11), the absolute value ABS (f '(I, J of the block f' (I, J) after mask M is applied.
J)) is larger than the threshold value TH (b12), and if larger, the number of edges is set to COUNT = COUNT + 1 (b13) and counted up.

Then, J = J + 1 (b14), it is determined whether or not J> 7 (b15), I = I + 1 (b16), and it is determined whether or not I> 7 (b17), and J> 7 and I> 7. Then, counting of the number of edges in the block is completed, so it is determined whether or not COUNT <TH3 (b18). Then, the number of edges COUNT is the third threshold TH
If the number is 3 or less, the number of edges is small, and therefore it is determined to be a flat portion. Next, it is determined whether TH3 ≤ COUNT ≤ TH4 (b19), the edge count COUNT exceeds the third threshold TH3, and the fourth threshold TH
When it is 4 or less, the number of edges is relatively large, so it is determined to be an edge portion. When the edge number COUNT exceeds the fourth threshold TH4, the number of edges is larger than that when it is determined to be an edge portion. Judge as a fine pattern part.

FIG. 9 and FIG. 10 show a flow chart of image determination based on the average value, which is an 8 × 8 block B (8,8), a threshold value THa, fifth and sixth threshold values TH5, TH6, a count value COUNT and an average value. H is determined (c1), and steps (a2) to (a1) shown in FIG. 5 are performed.
8 × 8 by steps (c2) to (c10) similar to 0)
The average value H of the block of pixels is determined. When the absolute value of the difference between the pixel value B (I, J) and the average value H is larger than the threshold value THa by the steps (c11) to (c19) (| B (I, J)
-H |> THa) [step (c14)], COUNT = COUNT + 1
(C15).

Next, it is determined whether or not COUNT <TH5 (c20), and the fifth threshold TH5
In the following cases, since the value of each pixel in the block is close to the average value, it is determined to be a flat portion. Then TH5 ≦ COUNT ≦
If the count value COUNT is greater than the fifth threshold value TH5 and less than or equal to the sixth threshold value TH6 (c21), the number of pixels having a difference from the average value H is relatively large. Therefore, if it is determined as an edge portion and is larger than the sixth threshold value TH6, the number of pixels having a difference with respect to the average value H is larger than that in the case of being determined as an edge portion. Judge as a part.

FIG. 11, FIG. 12 and FIG. 13 are flowcharts of image determination based on the average value of small blocks. A block of 8 × 8 pixels is subdivided. For example, for a small block (sub-block) of 2 × 2 pixels. The average value is calculated, and the difference between the average value and the pixel value exceeding the threshold value is counted up. The block B (8,8) of 8 × 8 pixels, the threshold value THb,
The fifth and sixth thresholds TH5 and TH6, the count value COUNT, and the average value H of 2 × 2 pixels are determined (d1). The large blocks are I and J, and the four pixel positions of the small block are II and JJ (II = 0,1, JJ = 0,1)
When the average value H in the small block is obtained by steps (d2) to (d15), and then the absolute value of the difference between the average value H and the pixel value is larger than the threshold value THb by steps (d16) to (d30). And count up as shown by COUNT = COUNT + 1 (d22).

Next, it is determined whether or not COUNT <TH5 (d31), and the fifth threshold value TH5
In the following cases, since there are many pixels having a value close to the average value H, it is determined to be a flat portion, and then it is determined whether TH5 ≤ COUNT ≤ TH6 (d32). If the sixth threshold value is TH6 or more, , Average value H
Since there are many pixels with a large difference with respect to, it is determined that it is a fine pattern portion, and it exceeds the fifth threshold TH5 and the sixth threshold T
In the case of H6 or less, the pixel having a large difference from the average value H is
Since it exists in a small block to some extent, it is determined to be a fine pattern portion. 5th and 6th in case of judgment by this small block
The thresholds TH5 and TH6 of the
It is also possible to set a value different from the sixth threshold values TH5 and TH6.

The thresholds TH, THa, THb and the first to sixth thresholds TH1 to TH6 in the above-described embodiments of pixel determination correspond to application conditions when the input image signal has many still images and many moving images. And can be set. Also, in the step (a19) for obtaining the standard deviation σ in FIG. 6, the 1/64 division process may be omitted, and the first and second threshold values TH1 and TH2 may be set to large values. The action and effect can be obtained.

In the image determination based on the standard deviation in FIGS. 5 and 6, when the standard deviation σ can be assumed to be in the range of 0 to 60, for example, the first threshold TH1 is 10 and the second threshold TH2 is Can be 30. Also, in the image determination by the edge detection in FIGS. 7 and 8, the number of edges is in the range of 0 to 64. For example, the third threshold TH3 is 5, the fourth threshold T
H4 can be set to 15. In the image judgment based on the average value in FIGS. 9 and 10, the fifth threshold TH
5 can be set to 5 and the sixth threshold TH6 can be set to 10.

FIG. 14 shows a flow chart of the quantization step control,
The quantizers 12, 32, 42 are provided with a plurality of quantizers A, B, C, and the inverse quantizer 20 on the receiving side or the inverse quantizers 37, 47 for interframe encoding have a plurality of inverses. This corresponds to the case where the quantizers A ′, B ′, C ′ are provided for switching control. First, the quantizers A, B having different quantization steps from the block F ′ (8, 8) of 8 × 8 pixels are used. , C
, And the inverse quantizers A ′, B ′, C ′ are determined (e1), the image determination signal is input and selection (SEL) is performed (e2). That is,
It is judged whether the judgment signal indicates a flat portion (e
3) In the case of the flat portion, select a quantizer A with a small quantization step, and similarly select an inverse quantizer A'corresponding to this quantizer A at the receiving side or in the previous stage of the frame memory. To do.

If it is not a flat part, it is judged whether it is an edge part (e
4), in the case of the edge part, the quantization step is the quantizer A
A quantizer B having a quantization step that is the same as or larger than that is selected, and a corresponding inverse quantizer B ′ is selected at the receiving side or in the preceding stage of the frame memory. If it is not an edge portion, a quantizer C having a quantization step whose quantization step is the same as or larger than that of the quantizer B is selected, and the corresponding inverse quantum is selected at the receiving side or in the preceding stage of the frame memory. Select the converter C '.

Instead of selecting the quantizers A, B, and C as described above, it is also possible to have a configuration in which the quantization steps in the quantizers 12, 32, and 42 are switched and controlled. Akabe 20,37,
Also in 47, the inverse quantization step can be switched and controlled.

The block determined to be a flat portion is a block in which the pixel value changes little, the high-frequency component of the transform coefficient by the discrete cosine transform is further reduced, and the reproduction image quality is less deteriorated even if the high-frequency component is deleted. Further, since the quantization step is made finer, the deterioration of the reproduced image quality in the flat portion having high visual sensitivity is reduced. In addition, a block that is determined to be an edge portion such as the outline of a character or a figure can perceive deterioration of the reproduced image quality due to a quantization error of a high frequency component to some extent, but since it has lower visual sensitivity than a flat portion, The block which is determined to be the same or rougher as the flattening step and the finer pattern portion such as a striped pattern contains a relatively large amount of high-frequency components of the transform coefficient by the discrete cosine transform. Since the sensitivity is low, the deterioration of the reproduced image quality due to the quantization error can hardly be perceived. Therefore, the quantization step is the same as or coarser than the case of the edge portion, and the coding efficiency is improved.

The present invention is not limited to each of the above-described embodiments, and the image determination threshold value and the like are optimally set in consideration of the characteristics of the image.

〔The invention's effect〕

As described above, the present invention allows the image determination unit 4 to
It is determined whether a flat portion, an edge portion, or a fine pattern portion corresponds to the block of the input image signal, and the quantization step in the quantization unit 2 is controlled according to the determination result, and the discrete cosine transform unit 1 performs block correspondence. The transform coefficient of the input image signal subjected to the discrete cosine transform is quantized, and the quantized output signal is variable length coded in the coding unit 3. By the control, there is an advantage that the encoding efficiency can be improved so that the deterioration of the reproduced image quality cannot be perceived.

Further, the image determination unit 4 determines the standard deviation σ as the first and second thresholds TH1 and TH.
By comparing with 2, or by comparing the number of detected edges with the third and fourth thresholds TH3, TH4, or by comparing with the number of pixels of the difference from the average value with the fifth, sixth thresholds TH5, TH6 By doing so, the image determination of the average portion, the edge portion, or the fine pattern portion is performed, and this image determination processing can be easily executed during the block corresponding processing period in the discrete cosine transform unit 1. . Therefore, there is an advantage that the quantization step of the quantization unit 2 is controlled based on the determination signal for each block, and the coding efficiency can be improved without deteriorating the reproduced image quality. It is also possible to further improve the coding efficiency by including processing such as motion compensation.

[Brief description of drawings]

FIG. 1 is a diagram for explaining the principle of the present invention, FIG. 2, FIG. 3, and FIG.
FIG. 5 is a block diagram of a different embodiment of the present invention.
FIGS. 6 and 6 are flowcharts of image determination by standard deviation, FIGS. 7 and 8 are flowcharts of image determination by edge detection, FIGS. 9 and 10 are flowcharts of image determination by average value, FIG. 11, 12 and 13 are flowcharts for image determination based on small block average values, FIG. 14 is a flowchart for quantization step control, FIG. 15 is a block diagram of a conventional example, FIG. 16 is a block scanning explanatory diagram, 17
The figure is a scanning explanatory view of a conventional example. Reference numeral 1 is a discrete cosine transform unit, 2 is a quantization unit, 3 is an encoding unit, and 4 is an image determination unit.

 ─────────────────────────────────────────────────── ───Continuation of front page (56) References JP-A-64-4186 (JP, A) IEEE TRANSACTIONS ON COMMUNICATIONS VOL. COM29 [12] (1981) P. 1799-1808 IEICE Communications Division National Conference 522 P. 1-201 (Sho 61)

Claims (3)

[Claims] 1. A discrete cosine transform unit (1) for performing a discrete cosine transform on an input image signal or an inter-frame difference signal of the input image signal, and a transform coefficient or the transform from the discrete cosine transform unit (1). A quantizer (2) for quantizing the inter-frame difference signal of coefficients, an encoder (3) for variable-length coding the quantized output signal from the quantizer (2), and the input image signal An image determination unit (4) for determining a property, wherein the image determination unit (4) corresponds to a block including a plurality of pixels of the input image signal, and a standard deviation of a pixel value of the block is equal to a first standard deviation. When it is less than or equal to the threshold value, it is determined to be a flat portion, when it exceeds the first threshold value and is equal to or less than the second threshold value, it is determined to be an edge portion, and when it exceeds the second threshold value, it is determined to be a fine pattern portion. Quantization step of the quantization unit (2) based on An image signal coding control method characterized by controlling the. 2. A discrete cosine transform unit (1) for performing a discrete cosine transform on an input image signal or an inter-frame difference signal of the input image signal, and a transform coefficient or the transform from the discrete cosine transform unit (1). A quantizer (2) for quantizing the inter-frame difference signal of coefficients, an encoder (3) for variable-length coding the quantized output signal from the quantizer (2), and the input image signal An image determination unit (4) for determining a property, the image determination unit (4) applying a Laplacian mask to the pixels of the block corresponding to the block composed of a plurality of pixels of the input image signal. The number of pixels exceeding a predetermined threshold value is counted, and when the number of pixels is less than or equal to a third threshold value, it is determined as a flat portion, and when the number of pixels is greater than the third threshold value and less than or equal to a fourth threshold value, it is determined as an edge portion. And when the fourth threshold is exceeded, a detailed sentence The image signal coding control method, wherein the quantization step of the quantization unit (2) is controlled based on the determination result. 3. A discrete cosine transform unit (1) for performing a discrete cosine transform on an input image signal or an inter-frame difference signal of the input image signal, and a transform coefficient or the transform from the discrete cosine transform unit (1). A quantizer (2) for quantizing the inter-frame difference signal of coefficients, an encoder (3) for variable-length coding the quantized output signal from the quantizer (2), and the input image signal An image determination unit (4) for determining a property, wherein the image determination unit (4) corresponds to a block composed of a plurality of pixels of the input image signal, the pixel of the block or a sub-block obtained by subdividing the block. The average value of the values is calculated, and the number of pixels whose absolute value of the difference between the average value and the value of each pixel in the block or sub-block exceeds a predetermined threshold value is counted. It is judged to be a flat part at the following times and the fifth
Is determined to be an edge portion when the threshold value is exceeded and is equal to or less than the sixth threshold value, and a fine pattern portion is determined when the threshold value is exceeded, and the quantization step of the quantization unit (2) is performed based on the determination result An image signal coding control method characterized by controlling the.