JP2910416B2 - Orthogonal transform 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 orthogonal transform encoding apparatus used for effectively increasing a compression ratio when a video signal is encoded with high efficiency.

[0002]

2. Description of the Related Art A band compression technique (or a high-efficiency code) for reducing a data amount of a video signal for long-time recording is used for configuring a recording / playback apparatus (VTR, video disk player, etc.) for recording / reproducing a video signal. Technology) is used. One of the band compression techniques is orthogonal transform coding. Orthogonal transform coding blocks video signals,
This is a method of performing frequency decomposition on a block-by-block basis to reduce the amount of data to be transmitted (recorded) for higher frequency components, and utilizes a visual characteristic that deterioration is more difficult to detect for higher frequency components. .

FIG. 4 is a block diagram of a conventional orthogonal transform coding apparatus for performing the above-described orthogonal transform coding, and its operation will be described.

In FIG. 4, reference numeral 1 denotes an input terminal for inputting a block signal obtained by blocking a video signal, 2 denotes an orthogonal transformer for orthogonally transforming the block signal, and 3 denotes an orthogonal transform coefficient obtained from the orthogonal transformer 2. In order to perform encoding, a reordering unit for reordering the arrangement of the transform coefficients, 4 is a reordering unit, 3 is an encoder for encoding the orthogonal transform coefficients output from the output unit, and 5 is an output terminal of the orthogonal transform encoding device. Output the coded data.
Further, the encoder 4 includes a quantizer 41 for quantizing the orthogonal transform coefficients at a certain step width, and a quantization selector 42 for selecting a quantizer having a step width suitable for keeping a desired data amount. And an encoder 43 for encoding the quantized data output from the quantizer 41.

[0005] The encoder 43 uses variable-length coding in which a code word with a higher frequency of occurrence is assigned a shorter (small data amount) code word with respect to the frequency of occurrence of the encoded code word.
Therefore, the quantization selector 42 selects the quantizer 41 having the optimal step width by calculating the data amount as a result of the variable length coding.

FIG. 5 shows the output of the orthogonal transformer 2 (A) and the output of the rearranger 3 (B) for explaining the operation of the rearranger 3, and shows the arrangement of the transform coefficients in units of one block. Is shown. The coefficient sequence shown in FIG. 5 is such that the orthogonal transform coding apparatus of the conventional example adopts a two-dimensional orthogonal transform technique in the horizontal and vertical directions, and the block has a block size of 4 pixels horizontally and 4 pixels vertically. Things. Therefore, in FIG. 5A, one block is composed of 16 coefficients in the horizontal direction and 4 coefficients in the vertical direction. In the block shown in FIG. 5A, the frequency component represented by each coefficient corresponds to a horizontal low band as leftward, and corresponds to a vertical low band as upward. The configuration of the orthogonal transformer 2 for performing the two-dimensional orthogonal transformation is such that the orthogonal transformation in one of the horizontal and vertical directions is performed and then the orthogonal transformation in the other direction is performed.
The coefficients are arranged as shown in FIG.

However, in order to perform encoding for two-dimensional orthogonal transform, the coefficients shown in FIG. 5B and a two-dimensional frequency band called a zigzag scan as shown in FIG. It is suitable to line up from the upper left) to the higher range (lower right in the figure). In other words, the lower frequency component including the DC component has a greater effect on vision, and the lower frequency component is a more important component. Therefore, in the encoder 4, it is only necessary to gradually reduce the step width from the beginning of the coefficient array until the data amount becomes a desired data amount, starting from a state where quantization is performed with the coarsest step width.

[0008]

However, the above configuration has the following problems. In terms of the data amount by encoding, there are blocks having a large data amount and blocks containing large amplitude frequency components, and blocks having a small data amount include blocks containing only small amplitude frequency components. If a similar step width is assigned to these two types of blocks to reduce the data amount,
Most of the latter coefficients of blocks containing only small-amplitude frequency components are deleted (coefficients are assigned to 0). Although this is a block with a flat screen, important information is lost and block noise occurs (for example, noticeable in the background vegetation and sky). In comparison, the block containing the large amplitude of the frequency components of the former, the frequency component not be more deleted in step width <br/> for amplitude, thus affected by the quantization error visually latter Smaller than a block.

However, since the optimal quantization performed in the encoder 4 of the conventional orthogonal transform encoder is performed for each block, it is difficult to control the selection of the quantizer between blocks (data amount control). In the block including only the small-amplitude frequency component, the quantization selection operation is performed in the same manner as the block including the large-amplitude frequency component,
This deteriorates the image quality of a flat portion such as the background.

A block having a large DC component and a small AC component (for example, a white background having a high luminance level) is difficult to detect even if visually deteriorated, but focuses on the AC component only. Therefore, the quantization is performed by reducing the step width. That is, quantization and encoding are performed while including the visual redundancy. Therefore, efficient compression cannot be performed by the conventional method.

The present invention has been made in view of the above-mentioned problems of the prior art, and provides an orthogonal transform coding apparatus which facilitates control of data amount between blocks, increases compression efficiency, and has a very small increase in circuit scale. The purpose is to:

[0012]

SUMMARY OF THE INVENTION The present invention provides orthogonal transform means for orthogonally transforming a block signal, and a coefficient and a sequence for encoding the transform coefficients orthogonally transformed by the orthogonal transform means for each block. Rearranging means for rearranging, classifying means for classifying transform coefficients into predetermined classes according to a plurality of threshold values, and detecting, for each block, a class which is considered to have the maximum value of the transform coefficient among the classes obtained by the classifying means And a plurality of thresholds of the maximum value class detecting means and the classifying means are configured to be different between a DC component and an AC component of a transform coefficient.

[0013]

According to this configuration, since the result obtained by classifying the transform coefficients corresponding to the maximum amplitude of the frequency components included in each block is obtained, the amplitude of the frequency component for each block can be easily determined at the time of the subsequent encoding. The data amount can be controlled without deleting important frequency components even with a small amplitude. or,
By performing the classification including the DC component, visual redundancy can be effectively removed, so that efficient compression can be performed.

[0014]

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram of an orthogonal transform coding apparatus according to an embodiment of the present invention, and FIGS. 2 and 3 are characteristic diagrams of transform coefficients and classes.

In FIG. 1, reference numeral 1 denotes an input terminal.
Similarly, each block has a block size of 4 horizontal pixels and 4 vertical pixels. Reference numeral 2 denotes an orthogonal transformer for performing a horizontal / vertical two-dimensional orthogonal transformation on an input block of 4 × 4 pixels. Reference numeral 3 denotes a sequence of transform coefficients output from the orthogonal transformer 2,
Is a rearranger for rearranging two-dimensional frequencies so as to line up from low to high in order to perform quantization and encoding. Reference numeral 5 denotes an output terminal for outputting the orthogonal transform coefficient encoded by the encoder 4. A classifier 60 having a class detector 61 and a comparator 62 and a maximum value class detector 70 are newly provided.

2 and FIG. 3, FIG. 2 shows an AC component (A
C) Maximum value class detection by FIG. 3, FIG. 3 shows direct current component (DC)
5 shows the progress of the maximum value class detection by means of. The horizontal axis indicates the order of the conversion coefficients, which corresponds to the order of the conversion coefficients in FIG. The left vertical axis represents the magnitude of the amplitude of the direct current component (DC) and the alternating current component (AC), and is the absolute value of the conversion coefficient. On the right vertical axis, the AC components are P1 and P
2, three types of thresholds P3 and four types of classes. The four types of classes are 0,
1, 2, and 3. The DC components are one type of threshold of P4 and two types of classes 0 and 3. Referring to FIG. 2, the classifier 60 receives the orthogonal transform coefficients of the conventional example in FIG. 5 (A) as input and converts the transform coefficients having small absolute values according to the magnitude of the absolute value of each orthogonal transform coefficient. 0, 1, 2,
Classify into any of the three classes. In the classifier 60, the conversion coefficient A is input to the class detector 61 and the comparator 62. The comparator 62 detects the absolute value of the DC component and the class detector 61 detects the absolute value of the AC component and the class. As shown in FIG. 2, since the absolute value of the DC component is equal to or smaller than the threshold value P4, the comparator 62 sends a "LOW" signal to the class detector 61. When the class detector 61 receives the signal of "LOW" from the comparator 62, it sends the data of class 0 to the maximum value class detector 70, and thereafter, the class detector 61 converts the conversion coefficient A
The second and subsequent absolute values are detected using the thresholds P1, P2, and P3, converted into each class, and sent to the maximum value class detector 70. The progress is shown in FIG.

Next, the maximum value class detector 70 updates the maximum class candidate by comparing the maximum class candidate with the input class based on the class C in FIG. The initial value is the DC component class value 0 or 3. If the maximum class candidate is 3, the detection of the maximum class candidate is stopped and the class is held for one block as shown in the fourth column of FIG. The retained class is used by the quantization selector 42 in the encoder 4 to select a quantizer having a different step width. When the absolute value of the DC component exceeds the threshold value P4 as shown in FIG. 3, a signal of "HI" is sent from the comparator 62 to the class detector 61. Class detector 6
1 receives the signal of "HI" from the comparator 62, sends out the data of class 3 to the maximum value class detector 70, and thereafter performs the same processing as described above.

According to the present embodiment, based on the class having the maximum value of the transform coefficient for each block obtained by the classifier 60 and the maximum value class detector 70, the quantization selector 42 As compared with the block including the block, the ratio of deleting the transform coefficient of the block including only the small amplitude frequency component can be reduced to prevent the deterioration of the image quality in the flat portion.
On the other hand, on a white background or the like having a high luminance level and a small amplitude of an AC component, the step width is coarsened, so that efficient quantization can be performed.

If a class for a large transform coefficient is assigned to a large number as in this embodiment, it is only necessary to find the maximum value of the number of assignments. In addition, in this embodiment, the range that the input class can take is 2 bits for classes 0 to 3, and the range that the orthogonal transform coefficient can take is generally 8 bits or more. And a register for storing the maximum value class candidate itself are small, which is very effective.

Furthermore, by fixing the input signal of the classifier to a predetermined level when the maximum value class is held, current consumption is reduced.

In this embodiment, the threshold of the absolute value of the DC component is set to one type and the class is set to two types. However, it is naturally possible to increase the threshold and the class and finely quantize.

In the embodiment of the present invention, the block signal to be orthogonally transformed has a block size of 4 × 4 pixels, but may have a block size of 8 × 8 pixels or 16 × 16 pixels. Furthermore, instead of horizontal and vertical two-dimensional orthogonal transformation,
A case of a dimensional orthogonal transform may be used. As described above, as the block size or the number of dimensions becomes larger and the number of transform coefficients constituting the block becomes larger, the detection of the maximum value class corresponding to the maximum value of the transform coefficient is performed in the same manner as in the present embodiment. Executing in parallel with the changer 3 is more effective because a delay device or the like is not newly detected for the maximum value class detection.

[0024]

As described above, according to the present invention,
Since the result of classifying the maximum value of the transform coefficient corresponding to the maximum amplitude of the frequency components included in each block is obtained, the amplitude of the frequency component for each block can be easily understood at the time of the subsequent encoding, and even if the amplitude is small, Since the data amount can be controlled without deleting important frequency components, it is possible to perform efficient orthogonal transform coding without deteriorating image quality, which is very effective. Furthermore, since the class of the frequency component is obtained in parallel with the reordering unit, no new memory is required for the class detection, which is advantageous in terms of circuit scale.

[Brief description of the drawings]

FIG. 1 is a block diagram of an orthogonal transform encoding apparatus according to an embodiment of the present invention.

FIG. 2 is a characteristic diagram of an absolute value of a conversion coefficient for describing the embodiment.

FIG. 3 is a characteristic diagram of an absolute value of a conversion coefficient for describing the embodiment.

FIG. 4 is a block diagram of a conventional orthogonal transform coding apparatus.

FIG. 5 is a diagram showing an arrangement of transform coefficients of a block for explaining the operation of the conventional example.

[Explanation of symbols]

 2 Orthogonal Transformer 3 Rearranger 60 Classifier 61 Class Detector 62 Comparator 70 Maximum Value Class Detector

Claims (1)

(57) [Claims] An orthogonal transform means for orthogonally transforming a block signal, a coefficient for encoding the transform coefficients orthogonally transformed by the orthogonal transform means for each block, and a rearranging means for rearranging the transform coefficients; Classifying means for classifying transform coefficients into predetermined classes according to a plurality of threshold values; and maximum value class detection for detecting, for each block, a class which is considered to have the maximum value of the transform coefficient among the classes obtained by the classifying means. Means, and a plurality of threshold values of said classifying means are configured to be different between a DC component and an AC component of said transform coefficient.