JP2000287210A - Method and unit for compressing image - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain a method and a unit that compress an image in a way that image quality of a complicated image is kept to be a prescribed level or over and image quality of a simple image is not excessively degraded. SOLUTION: A lower limit Rmin of a minimum transfer rate is set in a step (S6) in order to suppress degradation in the image quality resulting from a decreased GOP rate of an image with less code quantity in the case of encoding an image at a variable transfer rate. Thus, the image compression is attained, where high sharpness is maintained as a whole while minimizing the image degradation of a comparatively complicated image pattern and not degrading the image quality of the image with a comparatively simple image pattern.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention prevents the amount of code assigned to a relatively simple picture from becoming unnecessarily low and prevents the image from being excessively deteriorated, so that the overall image can be obtained most clearly. The present invention relates to a method and apparatus for compressing an image.

[0002]

2. Description of the Related Art When image information is conveyed through a transfer path or stored and stored, an image compression method for expressing the image information with as few bits as possible is useful for effective utilization of processing capacity. Method. A conventional image compression method will be described with reference to a flowchart of FIG. 7 illustrating a conventional image compression method. In this method, encoding at a variable transfer rate is performed in two stages, provisional encoding and main encoding. In the provisional encoding, the input image is divided into blocks called macroblocks, and motion-compensated predictive encoding is performed for each block (S21).

[0003] Discrete cosine transform (DCT, Discrete
A DCT coefficient is calculated by performing discrete cosine transform in block units in a Cosine Transform unit (S22), quantized using a standard quantization value in a quantization unit (S23), and variable length coding is performed in an encoding unit (S23). S24). The output code amount is accumulated by a code amount counter (S25), and the code amount when the standard quantized value is used is obtained. In this coding, motion compensation prediction coding (S28), discrete cosine transform (S29), coding (S31)
Is performed in the same manner as the temporary encoding. At the time of quantization (S30), a group of pictures (GOP, Group of Pictur) obtained by provisional encoding
e) For each generated code amount, the total code amount is obtained while performing a non-linear conversion, and the GOP rate is set so as to be the maximum within a range not exceeding the capacity of the storage medium (storage medium). At this time, the GOP rate of an image having a large generated code amount is limited so as not to exceed the maximum transfer rate Rmax of the recording medium.

[0004] The input image includes an image which is not suitable for compression encoding because of a rapidly moving picture, and an image which is close to a still image in which compression encoding is very easy.
Therefore, when coding is performed at a fixed transfer rate, the image quality of a complicated picture is greatly reduced as compared with the input picture, and the picture quality of a simple picture becomes almost the same. Therefore, by increasing the amount of code allocated to complex patterns and reducing the amount of code allocated to simple patterns, the variable transfer rate encoding enables high image quality as a whole even at low average transfer rates. A scheme has been proposed. When performing encoding at the variable transfer rate, provisional encoding for grasping in advance the difficulty of compression of an input image (complexity of a picture and intensity of movement) and a code amount obtained by the provisional encoding are used. GOP
2 of the main encoding for calculating and assigning the assigned code amount for each
Perform in stages. When calculating the allocated code amount for each GOP, the upper limit is set as described above, and the GOP rate is limited so as not to exceed the maximum transfer rate of the recording medium (for example, Japanese Patent Application Laid-Open No. 8-130736).

[0005]

SUMMARY OF THE INVENTION In the evaluation of subjective image quality, when the coding at a fixed transfer rate is compared with the coding at a variable transfer rate, it is pointed out that the latter deteriorates the image like a still image. There is. This is because only the upper limit is set when calculating the assigned code amount for each GOP, so that the transfer rate for an image close to a still image becomes unnecessarily low and the quantization value may become high. is there. Also, if the upper limit is set low in order to suppress the deterioration of an image close to a still image, encoding using the maximum transfer rate of the recording medium cannot be performed, and the amount of code allocated to a complicated picture decreases. No improvement in image quality can be expected for very complicated pictures. The image compression disclosed in the above-mentioned Japanese Patent Application Laid-Open No. 8-130736 intends to set an optimal quantization step based on a difference in generated code amount of quantization depending on a maximum value / minimum value of the quantization step. However, it is not possible to prevent the image quality of a relatively simple picture from being excessively deteriorated. In addition, JP-A-10-164588, JP-A-10-215460, JP-A-10-234037, and the like disclose image compression, but all of them have excessive image quality of a relatively simple picture image. No means for preventing deterioration is disclosed. SUMMARY OF THE INVENTION It is an object of the present invention to provide a method and apparatus capable of performing image compression while maintaining the image quality of a complicated image at or above a predetermined level without excessively degrading the image quality of a simple image.

[0006]

According to the method of the present invention, in a method of coding an image at a variable transfer rate, a lower limit value is set in addition to an upper limit value of a target transfer rate, and a code amount allocated to each image group is set. This is an image compression method characterized in that image compression is performed without decreasing the value below a predetermined value, and image compression may be performed while increasing or decreasing the lower limit value for each image. Further, the apparatus of the present invention includes a motion compensation unit, a discrete cosine transform unit, a quantization unit, an encoding unit, and a quantization control unit, and also sets an upper limit value and a lower limit value of a target transfer rate in the quantization control unit. The image compression apparatus according to claim 1, wherein a code amount is assigned to each image group, and an appropriate quantization is performed by a quantization unit so as to achieve the assigned code amount.

Hereinafter, the present invention will be described in detail. The present invention
In coding at a variable transfer rate that can change the amount of code assigned to each image group, set an upper limit and a lower limit when setting a target transfer rate for each GOP, and compare it with coding at a fixed transfer rate. It is possible to improve the image quality by increasing the code amount allocated to a complicated pattern by the upper limit value, and to suppress the deterioration of the image quality by preventing the code amount allocated to the simple pattern from becoming unnecessarily low by the lower limit value. The configuration is provided.

[0008] In an image compression method using a normal variable transfer rate, the main focus is on image compression without deteriorating the image quality of a complicated picture image, and relatively simple picture images are assigned. The code amount decreases.
In particular, to compensate for the more complex the pattern of the former image. The code amount allocated to the image of the relatively simple picture is excessively reduced, and the image quality is deteriorated. Even if the image has a relatively simple pattern, if the deterioration is severe, the adverse effect on the entire image is increased, and the image quality is deteriorated more than the image quality of the image of the complicated pattern is improved. According to the method of the present invention, it is possible to suppress excessive image quality deterioration of an image of a relatively simple pattern even if the image quality of an image of a complicated pattern is somewhat sacrificed, and to increase the sharpness of the entire image. Can be maintained.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, embodiments of image compression according to the present invention will be described based on examples shown in the accompanying drawings, but the embodiments do not limit the present invention. FIG. 1 is a flowchart showing a series of steps of the image compression method of the present invention, FIG. 2 is a flowchart showing the provisional encoding and the respective steps of the encoding of FIG. 1, and FIG. 3 is a flowchart of the quantization control unit of FIG. It is a flowchart which shows an internal structure in detail. As shown in FIG.
The image compression unit of the present embodiment divides one image into blocks called macroblocks composed of 16 × 16 pixels, for example, and a motion compensation unit 2 that performs motion compensation prediction coding in block units, for example, 8 × 8 A DCT unit that performs a discrete cosine transform on a block-by-pixel basis; a quantization unit 4 that divides a transform coefficient output from the DCT unit by a coefficient called a quantization value; A coding unit 5 for compressing by a redundancy compression coding method such as coding, a quantization control unit 7 for setting a quantization value from a code amount output from the coding unit, and a storage medium 6 for accumulating codes. Be composed.

Next, the overall operation of the image compression means having such a configuration will be described in detail. MPEG (Movi
In the motion compensation unit 2 standardized by the ng Picture Experts Group), the code amount fluctuates at the GOP cycle due to the coding structure. Therefore, the transfer rate is changed in GOP units,
Temporary encoding and actual encoding are two types of encoding at the variable transfer rate.
Perform in stages. In the tentative coding, the input image 1 is first divided by the motion compensation unit 2 into blocks called macroblocks composed of 16 × 16 pixels, and motion-compensated predictive coding is performed on a block-by-block basis (S1).

Next, the DCT unit 3 performs a discrete cosine transform in units of 8 × 8 pixels to calculate DCT coefficients (S
2) The quantization unit 4 performs quantization using the standard quantization value 705 (S3), and the encoding unit 5 performs variable length encoding (S4).
The code amount output here is accumulated by the code amount counter 701 (S5), and finally the code amount when the standard quantized value 705 is used is obtained. On the other hand, in the present coding, motion compensation prediction coding (S8), discrete cosine transform (S9), coding (S11)
Is performed in the same manner as the temporary encoding. At the time of quantization (S10), the total code amount is obtained while performing non-linear transformation on the generated code amount Gi 'for each GOP obtained by the provisional encoding. The GOP rate setting section 70 changes the non-linear conversion coefficient so that the total code amount stored in the code amount memory 702 becomes maximum within a range not exceeding the capacity of the storage medium (recording medium) 6.
3 sets the GOP rate Gi. At this time, the GOP rate of an image having a large generated code amount is limited so as not to exceed the maximum transfer rate Rmax of the recording medium. In addition, a minimum transfer rate Rmin is set to suppress the GOP rate of an image having a small amount of generated codes and to suppress the deterioration of image quality (S
6). The conversion function f (x) is as follows.

[0012]

(Equation 1)

[0013]

(Equation 2)

[0014]

(Equation 3)

Calculate a and b satisfying (Equation 1) and (Equation 2), determine the function f (x), and replace (Equation 3) with each GOP
Set as the target transfer rate for each. The parameters, maximum transfer rate, and minimum transfer rate in the equation can be set as appropriate according to the input image. As shown in FIG. 3, the quantization control unit 7
In the above, the code amount output from the encoding unit 5 at the time of provisional encoding is accumulated by a code amount counter 701, and the value is stored in a code amount memory 702. Subsequent GOP rate setting unit 703
In this case, the total code amount stored in the code amount memory at the time of the main encoding is set to the target total code amount (capacity of the recording medium) and the coefficient of the function (a and b) so as not to exceed the limit of the maximum / minimum code amount To determine. The code amount allocating unit 704 allocates the code amount at the same ratio as the generated code amount in MB units with respect to the code amount in the provisional encoding from the set GOP rate. Appropriate quantization is performed by the quantization unit 4 so that the allocated code amount is obtained. At this time, the setting is made so as not to exceed the capacity of the video buffer.

FIG. 4 is a flowchart showing in detail the internal configuration of the quantization control section 7a of another embodiment of the image compression means in the method of the present invention. In the quantization control unit 7a of the present embodiment, the coefficient is determined without considering the condition of the lower limit value for the time being. When the GOP rate at the time of setting the GOP rate falls below the minimum transfer rate, the GOP
Add the amount of code to the rate and raise it to the minimum transfer rate,
At this time, the difference between the original transfer rate and the minimum transfer rate is stored. By setting the target GOP rate low (decreasing the coefficient) at the time of encoding after this GOP, the generated code amount is suppressed below the target, and the code amount has a surplus. The remainder generated here is added to raise the minimum transfer rate to the above-described minimum transfer rate, and is allocated to the consumed code amount, so that the total code amount becomes equal. As a result, the processing at the time of coefficient setting can be reduced (one condition is reduced).

The quantizers 7 and 7a perform feedback control of the quantized value in accordance with the target code amount of each macroblock. Since the target code amount of each macroblock is based on the actual code amount obtained by the provisional encoding, the control can be performed stably. FIG. 5 is a graph showing the relationship between the fixed transfer rate (FIG. 5a), the conventional variable transfer rate (FIG. 5b), and the variable transfer rate (5c) of the present embodiment, in relation to time and code amount. As shown in FIG. 5, by setting the upper limit and the lower limit when obtaining the transfer rate for each GOP, the upper limit and the lower limit are set closer to a still image in which compression encoding can be easily performed as compared with a fixed transfer rate. It is possible to suppress the deterioration of the image quality of a simple image, and to improve the image quality of an image that is not suitable for compression encoding (an image with a sharp movement). In other words, by setting the lower limit of the code amount, the quantization value of the portion having a small code amount is reduced, and according to the present embodiment, the conventional control of the quantization value that takes a value between the fixed transfer rate and the variable transfer rate is achieved. it can. The relationship between the fixed transfer rate, the conventional variable transfer rate, and the image quality for each image of the variable transfer rate according to the present embodiment is as follows.

Still image: Fixed transfer rate> This embodiment> Conventional variable transfer rate Image with rapid movement: Conventional variable transfer rate> This embodiment> Fixed transfer rate

FIG. 6 shows a case where the lower limit is set low (FIG. 6).
7A is a graph showing the relationship between the time and the code amount when the value is set high (FIG. 6B). When the lower limit is lowered as shown in FIG. 6A (when the transfer rate approaches 0 bps (bits per second)), a large amount of code can be allocated to an image having a large generated code amount, and the image quality of an image having a sharp motion can be improved.
By setting the lower limit to a high value (making the minimum transfer rate close to the average transfer rate), it is not possible to allocate a large amount of code to an image with a large amount of generated code, but the code amount can be allocated to an image with a small amount of generated code. Deterioration of image quality of an image such as a still image can be suppressed. As described above, even if the maximum transfer rate is the same, the allocation of the code amount can be changed by setting the lower limit.

[0020]

According to the present invention, in a method of encoding an image at a variable transfer rate, a lower limit value is set in addition to an upper limit value of a target transfer rate, and a code amount allocated to each image group is reduced to less than a predetermined value. This is an image compression method characterized in that image compression is performed without performing image compression. According to this method, as in the prior art, not only can the image quality of this complicated pattern be improved by increasing the code amount allocated to the complicated pattern by the upper limit value, but also the code amount allocated to the simple pattern by the lower limit value Can be minimized so that the image quality of this simple picture is not reduced unnecessarily. Therefore, all of the available code amounts can be used most efficiently, and image compression at a level not available in the past becomes possible.

In the present invention, it is not necessary to fix the upper limit and the lower limit to constant values during a series of image compression steps, and increase or decrease both or any of them according to the image being processed during the step. Then, more appropriate image compression can be performed. Further, in the method of the present invention, it is not always necessary to set the lower limit, and the difference from the minimum transfer rate is stored in the difference memory, and the excess is used to compensate for the shortage, thereby suppressing the deterioration of the overall image quality. You may do it. Also, the device of the present invention
It comprises a motion compensation unit, a discrete cosine transform unit, a quantization unit, an encoding unit, and a quantization control unit. The quantization control unit sets an upper limit value and a lower limit value of a target transfer rate and codes each image group. An image compression apparatus characterized in that an amount is allocated and an appropriate quantization is performed in a quantization section based on the allocated code amount, and the use of this apparatus makes it relatively complicated for the same reason. It is possible to suppress the excessive deterioration of the image quality of the image of the simple pattern while maintaining the image quality of the image of the simple pattern substantially.

[Brief description of the drawings]

FIG. 1 is a flowchart showing a series of steps of an image compression method according to an embodiment of the present invention.

FIG. 2 is a flowchart showing the relationship between provisional encoding and encoding steps in FIG. 1;

FIG. 3 is a flowchart showing in detail an internal configuration of a quantization control unit in FIG. 2;

FIG. 4 is a flowchart showing a series of steps of an image compression method according to another embodiment of the present invention.

FIG. 5 is a graph showing the relationship between time and code amount at a fixed transfer rate (a), a conventional transfer rate (b), and a variable transfer rate according to the embodiment (c).

FIG. 6 is a graph showing the relationship between time and code amount when the lower limit is set low (a) and when the lower limit is set high (b) in the embodiment of the present invention.

FIG. 7 is a flowchart illustrating a conventional image compression method.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Input image 2 Motion compensation part 3 Discrete cosine transformation part 4 Quantization part 5 Encoding part 6 Recording medium (storage medium) 7 Quantization control part 701 Code amount counter 702 Code amount memory 703 GOP rate setting part 704 Code allocation part 705 Standard quantization value

Continued on the front page F term (reference) 5C059 KK01 MA00 MA05 MA23 MC11 MC38 ME02 PP01 PP04 TA46 TA57 TB03 TC01 TC10 TC37 TD02 TD16 TD17 UA02 UA32 5C078 AA04 BA57 CA21 DA00 DA01 DA07 9A001 EE04 EE05 HH27

Claims (4)

[Claims] In a method of encoding an image at a variable transfer rate, a lower limit value is set in addition to an upper limit value of a target transfer rate, and the image amount is reduced without decreasing a code amount allocated to each image group to less than a predetermined value. An image compression method comprising performing compression. 2. A method of encoding an image at a variable transfer rate, wherein a lower limit value is set in addition to an upper limit value of a target transfer rate, and a code amount allocated to each image according to image quality required for each image group. Image compression while increasing or decreasing the lower limit. 3. A method of encoding an image at a variable transfer rate, performing image compression by setting only an upper limit value of a target transfer rate, storing an excess or shortage code amount in a difference memory, and storing the stored code. An image compression method comprising performing image compression while compensating for a difference between a minimum transfer rate and a target rate according to an amount. 4. A motion compensation unit, a discrete cosine transform unit, a quantization unit, a coding unit, and a quantization control unit, wherein the quantization control unit sets an upper limit value and a lower limit value of a target transfer rate. An image compression apparatus, wherein a code amount is assigned to each image group, and an appropriate quantization is performed by a quantization unit based on the assigned code amount.