JP3191950B2 - Image coding device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image encoding apparatus for compressing and encoding an image.

[0002]

2. Description of the Related Art Generally, when performing compression encoding of an image,
There is a demand for encoding one screen with high efficiency within a predetermined code amount.

In response to this request, Japanese Patent Application Laid-Open No.
No. 0 and Japanese Unexamined Patent Application Publication No. Hei 2-262786 disclose that the code amount assigned to each block is set according to the activity of the block and the entire image, and when the coding of each block is performed, the code amount is set to be within the assigned code amount. If the allocated code amount is likely to be exceeded, a method has been proposed in which the coding is discontinued halfway. At this time, a quantization scale that determines a quantization step width when performing quantization of each block is calculated based on the activity of the entire image.

Further, Japanese Patent Application No. 1-28 filed by the present applicant has been filed.
In the system described in No. 3761, the transform coefficient orthogonally transformed for each block is quantized by a provisional quantization step width, a Huffman code is assigned to the quantized value, and the code amount in the block and The code amount of the entire image is obtained.

[0005] From these code amounts, the code amount allocated to each block and the quantization scale (coefficient by which the initial quantization step width is multiplied) are set, and the orthogonal transform is performed again for each block using these code amounts. The transform coefficients are quantized using the quantization step width obtained by multiplying the quantization scale by the default quantization step width, zigzag scanning is performed from the low frequency component side, and the quantized values arranged one-dimensionally are sequentially arranged from the DC component. Encoding is performed, and when the code amount allocated to each block is likely to be exceeded, the encoding is discontinued halfway. That is, it is cut off from the high frequency component side. The remaining code amount in each block is added to the allocated code amount of the next block to be coded. With such an image encoding device, encoding can be performed within the code amount required for one screen.

[0006]

However, in the above-described image encoding apparatus, when there are many fine structures including many high-frequency components in an image, the activity of the entire screen or the provisional code amount increases. The quantization scale also has a large value.

In the above-described image, most of the code amount is allocated to a block having a large activity or provisional code amount with respect to the allocated code amount of the block.
Only a small amount is allocated to a block having a small activity or provisional code amount.

As described above, when the quantization scale is large, the initial setting quantization step width is weighted so that the quantization step width is small for low frequency components and large for high frequency components. Even so, since the quantization scale by which the default quantization step width is multiplied is large, the quantization step width of the low-frequency component exceeds the limit and becomes large, and the quantization distortion of the low-frequency component increases. Also, if the block allocation code amount is small, not only high frequency components but also visually important low frequency components are not coded and are truncated.

When the data thus encoded is decoded, a large discontinuous step at a block boundary, which is called "block distortion", is generated, resulting in extremely remarkable deterioration. Therefore, an object of the present invention is to provide an image coding apparatus that controls the amount of code to suppress occurrence of block distortion and reduce visually noticeable deterioration.

[0010]

In order to achieve the above object, the present invention divides an input image into blocks each including a plurality of pixels, and controls the code amount of the input image by using the block and the input image. In an image coding apparatus for orthogonally transforming each block using an assigned code amount and a quantization scale generated based on the overall statistic and information amount, and quantizing and encoding output transform coefficients, an input image And an upper limit value setting means for setting an upper limit value of the quantization scale in accordance with the set code amount of the image encoding apparatus.

[0011]

According to the image coding apparatus having the above configuration,
An upper limit value is set for the quantization scale, and the upper limit value of the statistical (information) amount in the block is calculated so that the quantization scale predicted by the statistical (information) amount of the entire image falls within the above upper limit value. When the amount is within the above range, the amount of code allocated to the block becomes larger than that when the above processing is not performed for a block with a small amount of statistical (information) in the block, and the quantization scale does not exceed the upper limit. The chemical strain does not increase more than a certain degree.

Further, in a block in which the intra-block statistic is restricted by the upper limit, the allocated code amount is reduced as compared with the case where the above processing is not performed, and the coding of the high frequency component is discontinued earlier.

[0013]

Embodiments of the present invention will be described below in detail with reference to the drawings. FIG. 1 is a block diagram illustrating a schematic configuration of an image encoding device according to an embodiment of the present invention.

First, an input image is stored in a frame memory 1, and the read image data is transformed by a discrete cosine transform unit (DCT) 2 in a predetermined block unit. Note that the orthogonal transform is not limited to the discrete cosine transform, and may be an orthogonal transform using another method. Then, the image data read from the frame memory 1 is converted and input to the quantizer 3.

Next, a tentative quantizer step width is obtained by multiplying the quantization table 4 storing the weighted initial quantization step width and the default quantization scale from the scale factor calculator 5. The quantization is performed on each transform coefficient using the provisional quantizer step width. The DC component of the quantized transform coefficient is input to the DC DPCM 6, and the difference between the DC component and the quantized DC component of the previous block is calculated.

The difference value enters the DC Huffman coding unit 7, and only the word length of the Huffman code word corresponding to the input value is read from the DC Huffman code table 8 and output to the code amount calculation unit 9.

On the other hand, the AC component of the quantized transform coefficient is
As shown in FIG. 4, zigzag scanning is sequentially performed from the low frequency component side, and input to the two-dimensional Huffman encoding unit 10. The two-dimensional Huffman encoding unit 10 obtains a run length of “0” of the input AC component in the zigzag scan order, and collects the immediately following coefficient value other than “0” into one Huffman code. Word length of the code word, further, when it came to the end of the block, the word length of the E OB code word is read from the two-dimensional Huffman coding unit 10, is output to the code amount calculating section 9 .

The code amount calculator 9 adds up the codeword lengths of the input blocks and outputs the result to the block code amount storage memory 12 and the one-pass code amount limiter 13 as the block code amount. Block code amount storage memory 1
2 stores each block code amount. Here, a specific configuration of the one-pass code amount limiter 13 is shown in FIG. 2 and will be described.

First, the code amount CI set in the screen is input to the code amount limit value setting unit 21 in the first pass, and based on the relationship between the quantization scale and the code amount as shown in FIG. The value αu on the boundary line b is set to the upper limit. Here, FIG.
Both the vertical and horizontal axes are logarithmic. This set code amount CI
Can be obtained from the straight line a . Here, the straight line a is an average when the relationship between the quantization scale and the code amount is obtained for various images.

The block code amount of the transform coefficient quantized using the above-described initialization quantization scale α _I is input to the block code amount integration and code amount generation frequency distribution creating unit 22, where one screen worth of image data is input. The total code amount is obtained, and FIG.
A frequency distribution table of the block code amount as shown in FIG.

Here, when the total code amount Ca on both surfaces is larger than the upper limit code amount Cu, the boundary value of the hatched portion 31 in FIG. 3A is changed to the upper limit of the block code amount as shown in FIG. By setting the value, it is possible to satisfy Ca ≦ Cu. Such a block code amount upper limit value is calculated by the block code amount upper limit value calculation and the total code amount calculation unit 23. As a concrete method of this,
The code amount of the hatched portion 31 in FIG.

[0022]

(Equation 1) (Here, C (t) is the frequency at the time of the code amount t, x is the upper limit, and the following expression (1) is referred to as “A”). This is performed by obtaining the maximum x satisfying the following equation: C _a −A ≦ C _u (2) When x is obtained, C _a -A (3) is output to the scale factor calculation unit, and the value of x is output to the block code amount limiter 24.

In the block code amount limiter 24, the block code amount stored in the memory is set within the upper limit x, and the block allocation code amount calculation unit 1 shown in FIG.
4 is output. In the scale factor calculation unit 5,
The set code amount and the total code amount Ca-A after the limiter processing, which is the output of the one-pass code amount limiter 13, are input.
From these quantities a predicted quantization scale is calculated. here
Then, the predicted quantization scale is calculated from a straight line parallel to the straight line a shown in FIG. 5 and passing through the coordinates (αI, Ca− A). In the block allocation code amount calculation unit 14, the set code amount C
And a total code amount Ca-A calculated in I and one pass code amount limiter processing unit 13 calculates an allocation scale beta. β
= Set code amount CI / 1 pass code amount limiter 13
Calculated total code amount Ca-A

By multiplying the allocation scale β by the block code amount subjected to the limiter output from the one-pass code amount limiter 13, the block allocation code amount is calculated. The data in the block code amount storage memory 12 is rewritten with these calculated values.

The above is the processing of the first pass, and then the processing shifts to the processing of the second pass. In the second pass processing, the image data stored in the frame memory 1 is extracted for each block in the same manner as in the first pass processing, and the DCT 2
Perform DCT, and the quantizer 3 quantizes the transform coefficient. The quantization step width at this time is obtained by multiplying the predicted quantization scale calculated in the first pass by the default quantization step width.

Then, of the quantized transform coefficients, the difference between the DC component and the DC component of the previous block is obtained in the same manner as in the first pass, and the Huffman code word corresponding to this difference value is obtained by the DC Huffman coding. The Huffman code word is read from the DC Huffman code table 8 by the unit 7, and this Huffman code word is
The signal is sent to the multiplexer 15.

As in the first pass, the AC component of the transform coefficient is processed by the two-dimensional Huffman encoding unit 10 in the zigzag scan order as shown in FIG. The coefficient values that are not "" are combined into one, the corresponding Huffman code word is read from the two-dimensional Huffman code table 11, and this code word is sent to the multiplexer 15. Further, the DC Huffman encoding unit 8
In addition, the word length of the codeword encoded by the two-dimensional Huffman encoding unit 10 is input to the code amount calculation unit 9 and the code amount in the block is integrated (the block code amount in the second pass: Σ).
C _II (i)) The truncation processing unit 16 _performs comparison with the allocated code amount _Callot of the block currently being encoded stored in the block code amount storage memory 12. Here, as shown in FIG. 6, the index i is an index when a “0” run length of the transform coefficient zigzag-scanned from the low-frequency component and the subsequent coefficient values other than “0” are grouped. Then

[0028]

(Equation 2) , The coding is performed up to the maximum Y, and thereafter the coding is not performed, and the coefficients are discarded. Remainder of allocated code amount

[0029]

(Equation 3)

Is added to the code amount allocated to the next block. The Huffman codewords input to the multiplexer 15 are sequentially output to the transmission path. This process is performed on all blocks in the image, and the image can be encoded within the set code amount.

The above-described processing uses the provisional code amount for one pass to calculate the code amount control parameters (block allocation code amount and quantization scale) for two passes.
The same operation can be performed by obtaining an activity at the time of passing and calculating a code amount control parameter using this value.

The moving picture coding apparatus according to the present invention described in detail above provides an upper limit value for the quantization scale, which is predicted by the statistics of the entire image and which is within the above upper limit value. The upper limit value of the information amount in the block is obtained, and by setting the value within the value, a block having a smaller information amount is assigned a larger block allocation code amount than when the above processing is not performed, and the quantization scale is further increased. Since the value does not exceed the value, the quantization distortion for these blocks described above is smaller than when the process of obtaining the upper limit is not performed, and a code amount capable of coding a certain low-frequency component is assigned. ) With a small amount of blocks, block distortion that often occurs can be reduced. The present invention can be applied to both the encoding of a still image and the encoding of a moving image using the same principle. In addition, the present invention is not limited to the above-described embodiment, and it goes without saying that various modifications and applications are possible without departing from the spirit of the invention.

[0033]

As described above in detail, according to the present invention, it is possible to provide an image coding apparatus which controls the amount of code to suppress the occurrence of block distortion and reduce visually noticeable deterioration.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a schematic configuration of an image encoding device according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating a specific configuration of a one-pass code amount limiter processing unit illustrated in FIG. 1;

FIG. 3A is a diagram illustrating a frequency distribution of a block code amount, and FIG. 3B is a diagram illustrating a frequency distribution of a block code amount in a range up to a block upper limit value.

FIG. 4 is a block diagram illustrating blocks in order from a low frequency component side;
It is a figure showing the state where a zigzag scan is performed.

FIG. 5 is a diagram illustrating a relationship between a quantization scale and a code amount;

FIG. 6 is a diagram illustrating a DC / AC conversion coefficient and an index;
FIG. 6 is a diagram illustrating a configuration in which components are arranged in a zigzag scan order from low frequency components.

[Explanation of symbols]

1. Frame memory, 2. Discrete cosine transform unit (DC
T), 3 quantizer, 4 quantization table, 5 scale factor calculation unit, 6 DC DPCM, 7 DC Huffman coding unit, 8 DC Huffman code table, 9 code amount calculation unit, 10 ... two-dimensional Huffman coding unit, 11 ... 2
Dimensional Huffman code table, 12: block code amount storage memory, 13: one-pass code amount limiter processing unit, 14: block allocation code amount calculation unit, 15: multiplexer,
16: truncation processing unit, 21: code amount limit value setting unit for the first pass, 22: block code amount integration and code amount occurrence frequency distribution creation unit, 23: block code amount upper limit value calculation and total code amount calculation unit, 24 ... Block code amount limiter, 31.

Continuation of the front page (56) References JP-A-2-222386 (JP, A) JP-A-3-24889 (JP, A) Keiji Nemoto (1 other person), "Code amount control method of DCT coding method" , IEICE Autumn National Conference, Proceedings of the Institute of Electronics, Information and Communication Engineers, August 15, 1989, D-45, Volume 6, p. 6-45 (58) Field surveyed (Int.Cl. ⁷ , DB name) H04N ^7/ 24-7/68 H04N 1/41-1/419 H03M 7/30 JICST file (JOIS)

Claims (1)

(57) [Claims] An input image is divided into blocks each including a plurality of pixels, and an allocated code amount and a quantization scale generated based on statistics of the blocks and the entire input image are used to control a code amount of the input image. In an image encoding device that orthogonally transforms each block using, and quantizes and encodes the output transform coefficients, an upper limit value setting that sets an upper limit value of a quantization scale according to a set code amount of an input image. An image encoding apparatus comprising: