JPH06311369A - Picture coder - Google Patents

Abstract

PURPOSE:To improve the coding efficiency by applying various coding to each block and selecting a code with least code quantity and providing an output of the result. CONSTITUTION:A block division section 11 divides a picture signal into blocks of predetermined picture elements and is inputted to an orthogonal transformation section 12, in which orthogonal transformation is executed to each block. A quantization section 13 uses a quantization table 14 to quantize an output of the transformation section 12 and the result is fed to Huffman coding sections 151-153. The coding sections 151-153 apply zigzag, longitudinal and lateral scanning to rearrange a quantization output of 2-dimension arrangement into linear arrangement and a Huffman coding table 16 is used for coding. Then a code selection section 17 counts the code quantity coded by the coding sections 151-153 and selects the code from the coding section generating a least code quantity and the selected code is outputted to a multiplexer section 18. Thus, the coding efficiency is improved.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image coding apparatus for coding an image signal.

[0002]

2. Description of the Related Art Conventionally, as shown in FIG. 5, an image coding apparatus which performs coding by an orthogonal transform coding method typified by a color still image international standard coding method divides an input image signal into blocks. The unit 1 divides the block into, for example, 8 × 8 pixels, and the orthogonal transform unit 2 performs orthogonal transform such as discrete cosine transform for each block.

In the output from the orthogonal transform unit 2, a low frequency component appears in the upper left part of the block and a high frequency component appears in the lower right part of the block. Then, the quantization unit 3 uses the quantization table 4 to quantize the output from the orthogonal transform unit 2. This quantization is performed by quantizing each output in the block with the quantization step size at the same position in the quantization table 4.

The output from the quantizing unit 3 is encoded by the Huffman encoding unit 5 using the Huffman encoding table 6 and output as encoded data.

The Huffman coding unit 5 has 0, 1, shown in FIG.
Zigzag scan is performed to rearrange the two-dimensional array of blocks into a one-dimensional array as indicated by 2, 3, 4, ...
The combination of the quantized output of 0 and 0 run is encoded as an event by the Huffman code.

For example, when the image signal of the block of 8 × 8 pixels shown in FIG. 8 is orthogonally transformed by the orthogonal transformation unit 2, a transformed output as shown in FIG. 9 is obtained. When this is quantized by the quantization table 4 shown in FIG. 6, a quantized output as shown in FIG. 10 is obtained.

The DC component "51" at the upper left of this quantized output block is Huffman-encoded for the difference value from the DC component of the preceding block. For the AC component, a non-zero quantized output and a run length of 0 following the non-zero quantized output are obtained according to the zigzag scan of FIG. 7, and the Huffman encoding table 6
To be encoded.

After encoding all the coefficients in the block, EOB (4 bits) indicating the end of the block is added and output as encoded data.

For example, when the Huffman coding table of FIG. 4 (only the bit length of each event is shown) is used, the number of bits of the AC component is 33 bits.

[0010]

When an orthogonal transformation is applied to an image signal, it is transformed into a signal corresponding to a frequency component, and the transformed coefficient has a low frequency component in the upper left portion of the block and a high frequency component in the lower right portion of the block. . The image signal has a high correlation between adjacent pixels and a large amount of low-frequency component power. Therefore, when the flat part of the image is converted, power is concentrated on the upper left part of the block.

Since the edge portion of the image has a high frequency component,
After conversion, a large amount of power is generated on the high frequency component side depending on the direction of the edge. The correspondence between the image signal area and the converted signal area is shown in FIGS. 11 and 12.
In addition, the shaded portion indicates the portion where the electric power is large.

That is, in the flat image signal area as shown in FIG. 11A, the electric power is concentrated in the upper left portion as shown in FIG. 12A, and the image signal as shown in FIG. When there is a vertical edge in the area, power is concentrated in the horizontal direction in the converted signal area as shown in FIG.
When there is a horizontal edge in the image signal area as shown in (c), power is concentrated vertically in the converted signal area as shown in (c) of FIG. 12, and as shown in (d) of FIG. When the image signal area has an oblique edge, (d) of FIG.
In the converted signal area, the electric power is concentrated diagonally as shown in FIG.

As described above, the distribution of transform coefficients has the property of changing according to the statistical properties of the image.

However, conventionally, when encoding a quantized output obtained by quantizing an orthogonal transform output, zigzag scanning is performed in all blocks. Therefore, encoding a flat portion of an image enables efficient encoding. In the case of the edge part, there is a problem that the coding efficiency is reduced.

Therefore, the present invention intends to provide an image coding apparatus capable of improving coding efficiency by performing various kinds of coding for each block and selecting and outputting the code having the smallest code amount. Is.

[0016]

The invention according to claim 1 is
A block division unit that divides an image signal into blocks including a plurality of pixels, an orthogonal transformation unit that performs orthogonal transformation of each block divided by the block division unit, and an orthogonal transformation output for each block from the orthogonal transformation unit is quantized. A quantizer for encoding, a plurality of encoders for encoding the quantized output from the quantizer by different encoding processes, and an encoder for outputting the smallest code amount from each encoder. A code selection unit for selecting is provided.

[0017]

In the invention having such a structure, the image signal is orthogonally transformed for each block, quantized and coded. At this time, various coding processes are performed by the plurality of coding units, and the code selection unit selects the coding unit that outputs the smallest code amount. In this way, the code with the smallest code amount is output.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings.

As shown in FIG. 1, the image signal is supplied to a block division unit 11 to be divided into blocks of, for example, 8 × 8 pixels, and each divided block is input to an orthogonal transformation unit 12 to be discrete for each block. Orthogonal transformation such as cosine transformation is performed.

The orthogonal transform output from the orthogonal transform unit 12 is supplied to the quantizing unit 13, and the quantizing unit 13 uses the quantizing table 14 as shown in FIG. 6 to output the output from the orthogonal transform unit 12. Quantize. This quantization is performed by quantizing each output in the block with the quantization step size at the same position in the quantization table 14.

The quantized output from the quantizer 13 is supplied to the first, second and third Huffman encoders 151, 152, 1
5 3 respectively, and each Huffman coding unit 151, 1
The Huffman coding table 16 is used for coding at 52 and 153.

The first Huffman encoder 151 is shown in FIG.
Zigzag scanning is performed to rearrange the quantized output of the two-dimensional array into a one-dimensional array as indicated by 0, 1, 2, 3, 4, ... The Huffman coding table 16 shown in FIG. 4 is used for coding.

The second Huffman coding unit 152 is shown in FIG.
As shown by 0, 1, 2, 3, 4, ..., The vertical scan is performed to rearrange the quantized output of the two-dimensional array into a one-dimensional array, and the combination of the non-zero quantized output and the zero run is combined. Is an event and is encoded by the Huffman encoding table 16 shown in FIG.

The third Huffman coding unit 153 shown in FIG.
Then, the horizontal scan is performed as indicated by 0, 1, 2, 3, 4, ... to rearrange the quantized output of the two-dimensional array into a one-dimensional array, and the combination of the non-zero quantized output and the zero run is combined. Is an event and is encoded by the Huffman encoding table 16 shown in FIG.

Each of the Huffman coding units 151, 152,
The code amount coded by 153 is counted by the code selecting unit 17, the code from the coding unit that generates the smallest code amount is selected, and the selected code is output to the multiplexing unit 18.

The multiplexer 18 adds a code after the selected information bit indicating the Huffman code and outputs the coded data.

In the embodiment having such a configuration, for example, when the image signal of the block of 8 × 8 pixels shown in FIG. 8 is orthogonally transformed by the orthogonal transformation section 12, a transformed output as shown in FIG. 9 is obtained. Is quantized by the quantization table 4 shown in FIG. 6, a quantized output as shown in FIG. 10 is obtained.

The quantized output is converted into a Huffman coding table 16 as shown in FIG. 4 by the first Huffman coding unit 151.
When encoded using, "0, 2, 2, 0, 0, 0,
0,0,2,1,0,0,0,0,0,0,0,0,
0, 1 ”, and the number of bits of the AC component is 33 bits.

That is, the first "0, 2" is 0, the run is 1,
If the value is 2, it will be 5 bits from the Huffman encoding table,
The next "2" is 0, the run is 0, the value is 2, and the Huffman coding table has 2 bits, and the next "0, 0, 0, 0, 0,
2 "is 0 run 5 and the value is 2 and becomes 11 bits from the Huffman coding table. Next" 1 "is 0 run 0 and the value is 1
2 bits from the Huffman coding table, and the last "0,0,0,0,0,0,0,0,0,1" has 0 runs, the value 1 and 9 bits from the Huffman coding table Becomes Then, 4 bits of EOB are added to this, 33
Become a bit.

When the second Huffman encoder 152 encodes this quantized output using the Huffman encoding table 16, it becomes "2,2,2,1,1", and the number of bits of the AC component is It will be 14 bits.

That is, "2" is 0 when the run is 0 and the value is 2 when the run is 2
It becomes a bit and "1" becomes 0 when the run is 0 and the value is 1 and becomes 2 bits. Therefore, the total becomes 10 bits.
Bits are added to make 14 bits.

When the third Huffman coding unit 153 codes this quantized output using the Huffman coding table 16, "0,0,0,0,0,0,0,2,0,
0,0,0,0,0,0,2,0,0,0,0,0,
0,0,2,0,0,0,0,0,0,0,1,0,
0,0,0,0,0,0,1 ", and the bit number of the AC component is 56 bits.

That is, "0,0,0,0,0,0,0,
"2" has 7 runs and a value of 2 and is 12 bits, and "0,0,0,0,0,0,0,1" has 7 runs and a value of 1 and has 8 bits. It becomes 52 bits, and 4 bits of EOB are added to this to become 56 bits.

The code amount of each of the Huffman coding units 151 to 153 is counted by the code selection unit 17, and the code amount of the coding unit 152 which generated the smallest code amount, that is, the code of 14 bits is selected. Thus the multiplexing unit 1
Reference numeral 8 adds a code after the selected information bit indicating the Huffman code, and outputs the coded data.

In this way, each Huffman coding unit 151-1
Since the code from the coding unit having the smallest code amount when 5 3 is coded is selected and output as the coded data, the coding efficiency can be improved. That is, it is possible to output encoded data having a higher compression rate.

Further, since the Huffman coding units 151 to 153 differ only in the order of rearranging from the two-dimensional array to the one-dimensional array, the device scale can be slightly increased and easily realized.

In the above embodiment, three types of Huffman coding units are used, but the present invention is not necessarily limited to this. Even if two types of Huffman coding units are used, there are four or more types. A Huffman coding unit may be used.

[0038]

As described above, according to the present invention, it is possible to improve coding efficiency by performing various kinds of coding for each block and selecting and outputting the code having the smallest code amount.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention.

FIG. 2 is a diagram showing a scanning order when rearranging from a two-dimensional array to a one-dimensional array in the second Huffman encoding unit of the embodiment.

FIG. 3 is a diagram showing a scanning order when rearranging from a two-dimensional array to a one-dimensional array in a third Huffman coding unit of the embodiment.

FIG. 4 is a diagram showing an example of a Huffman encoding table according to the same embodiment.

FIG. 5 is a block diagram showing a conventional example.

FIG. 6 is a diagram showing an example of a quantization table.

FIG. 7 is a diagram showing a scanning order when rearranging from a two-dimensional array to a one-dimensional array in the conventional and first Huffman encoding units.

FIG. 8 is a diagram showing an example of an image signal.

FIG. 9 is a diagram showing an example of an orthogonal transform output.

FIG. 10 is a diagram showing an example of a quantized output.

FIG. 11 is a diagram showing an example of an image signal area.

12 is a diagram showing an example of a converted signal area corresponding to FIG.

[Explanation of symbols]

 11 ... Block division unit 12 ... Orthogonal transformation unit 13 ... Quantization unit 151-153 ... Huffman coding unit 17 ... Code selection unit

Claims (1)

[Claims] 1. A block division unit that divides an image signal into blocks composed of a plurality of pixels, an orthogonal transformation unit that performs orthogonal transformation of each block divided by this block division unit, and a block for each block from the orthogonal transformation unit. A quantizer that quantizes the orthogonal transform output, a plurality of encoders that encode the quantized output from this quantizer by different encoding processes, and the smallest amount of code output from each encoder An image coding apparatus, comprising: a code selecting unit that selects a coding unit that performs the image coding.