JP3169019B2 - Image encoder - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to an image encoder, and more particularly, to an image encoder that can be applied to a variable-rate moving image encoder that performs image quality control.

2. Description of the Related Art Conventionally, as a variable-rate video encoder for controlling image quality, there is one shown in FIG. 2 (Ogata et al., “Generation of Burst Video Coding Information,” IEICE Technical Committee, IE87-83, pp53-69).

In FIG. 2, an image signal input from an input terminal 11 is converted into an encoding processing format by a pre-processing unit 12, and then supplied to a subtracter 13 as a subtracted input. The prediction signal Sp for the image signal Sin after the format conversion is given to the arithmetic unit 13 as a subtraction input, and the subtractor 13 obtains a prediction error signal e. The prediction error signal e is converted into a variable quantizer 14. Given to. Variable quantizer 14
After quantizing according to the quantization characteristics such as the quantization step size and the number of quantization bits given from the quantization characteristic selection unit 27 described later, the local reproduction prediction error signal ek is obtained by performing inverse quantization in particular. Entropy encoder 15 for prediction error information
Give to. The entropy encoder 15 performs Huffman coding, run-length coding or modified Huffman coding on the reproduction prediction error signal ek, and outputs
To a multiplexing unit (not shown), a packet assembling unit, and the like.

Also, the local reproduction prediction error signal ek from the variable quantizer 14
Is supplied to the adder 17. The adder 17 is also supplied with the above-described prediction signal Sp of the image signal.
Outputs a local reproduction image signal Sink. The local reproduced image signal Sink is supplied to the one-pixel delay circuit 18 and the memory
It is provided to a frame delay circuit 19. One pixel delay circuit
The one-frame delay circuit 19 forms a prediction signal of an image signal (inter-frame prediction) by delaying by one frame, while the one-frame delay circuit 19 forms a prediction signal of an image signal (inter-frame prediction) by delaying by one frame. These predicted signals are 2
It is provided to a switch circuit 20 and a prediction selection unit 21 having an input / output configuration. The prediction selecting unit 21 controls the switch circuit 20 so as to compare these and select a prediction signal determined to have a smaller prediction error. In this way, the switch circuit
The prediction signal (Sp) selected in 20 is provided to the above-described subtractor 13 in the next frame.

In this conventional example, motion compensation is performed for inter-frame prediction. The image signal Sin that has passed through the pre-processing unit 12 is provided to a motion vector detection circuit 22, where a motion amount MV is detected. This motion amount MV is given to the one-frame delay circuit 19,
The one-frame delay circuit 19 controls writing or reading of the reproduced image signal Sink to perform motion compensation on the prediction signal.
Note that the motion amount MV is coded through the DPCM coder 23, the entropy coder 24, and the output terminal 25 sequentially, and is then provided to a multiplexing unit, a packet assembling unit (not shown) and the like. The entropy encoder 24 performs Huffman coding or run-length coding.

This conventional example is further provided with a configuration for at least compensating for S / N regardless of the image content. The image signal Sin and the locally reproduced image signal Sink are provided to the S / N calculator 26. The S / N calculation unit 26 calculates the S / N for each block obtained by dividing one image vertically and horizontally based on these signals, and supplies the calculated S / N to the quantization characteristic selection unit 27. The quantization characteristic selection unit 27 selects a quantization characteristic for each block and supplies the selected quantization characteristic to the variable quantizer 14 such that the S / N of each block becomes substantially equal to a constant value. The quantization characteristic information is subjected to, for example, fixed-length encoding by the entropy encoder 28, and supplied to an unillustrated multiplexing unit, packet assembling unit, and the like via the output terminal 29.

[Problems to be Solved by the Invention] As described above, according to the conventional video encoder, it is possible to perform encoding so as to minimize S / N regardless of the input image content.

By the way, an image appeals to human vision. Therefore, when coding an image, it is desirable to consider human visual characteristics. However, the S / N as the image quality evaluation value in the conventional image encoder is not always proportional to the human subjective evaluation value, and therefore, the conventional image encoder considers human visual characteristics. I can't say that.

The present invention has been made in view of the above points,
It is an object of the present invention to provide an image encoder that performs encoding so that image quality can be controlled in consideration of human visual characteristics.

[Means for Solving the Problem] To solve the problem, the image encoder according to the present invention includes the following components.

That is, a frequency separation unit that separates an input image signal into a plurality of frequency components (including at least a DC component) by a filter, and performs encoding having predetermined quantization characteristics in accordance with each of the separated frequency components. Using a plurality of encoders and the characteristics of each separated frequency component and visual frequency characteristic information, distribute the specified allowable total distortion amount to each frequency component, and calculate the A quantization characteristic selection unit for selecting a predetermined quantization characteristic to be given to each encoder.

 In addition, the encoding method in each encoder does not matter.

[Operation] In the present invention, the frequency separation unit separates an input image signal into a plurality of frequency components and supplies the plurality of frequency components to an encoder corresponding to each frequency component. Each encoder encodes the input frequency component signal according to the quantization characteristic provided by the quantization characteristic selection unit.

The present invention is characterized in a method of selecting a quantization characteristic in a quantization characteristic selection unit. The quantization characteristic selection unit uses the visual frequency characteristic information to distribute the specified allowable total distortion amount to each frequency component, and provides the quantization characteristic to each encoder based on the distributed distortion amount. To select.

In this way, image quality can be controlled in consideration of visual characteristics.

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

FIG. 1 shows the structure of this embodiment. In FIG. 1, an image signal S input from an input terminal 30 is converted to a QMF (Quadrature Mi
rror Filter). This QMF bank 31
As a result, the input image signal S is separated into n subband signals Si (i is 1 to n) having the same bandwidth. here,
It is assumed that the center frequency ωi of each subband signal Si has a relationship represented by the following equation: ω1 <ω2 <ω3 <... <ωn (1)

The sub-band signal S1 including the lowest frequency component is provided to the DC component calculator 32 and the subtractor 33. The other subband signals S2 to Sn are respectively associated with the corresponding encoders 342 to 3
4n. The center frequency ωi of each subband signal Si is provided to the quantization characteristic selection unit 35.

The DC component calculator 32 calculates the DC component S0 from the subband signal S1 including the lowest frequency component. For example, the DC component S0 is Calculate according to Here, S1k is a sample value in the subband signal S1, and M is the number of samples used for calculation.

The DC component thus obtained by the DC component calculation unit 32
S0 is provided to the encoder 340 and the subtractor 33.

The subtracter 33 subtracts the DC component S0 from the subband signal S1 from the QMF bank 31, and supplies the encoder 341 with the lowest subband signal S1a from which the DC component has been removed.

Each of the encoders 34j (j is 0 to n) is provided from the quantization characteristic selection unit 35. For example, encoding is performed based on a quantization characteristic including a quantization bit number Rj and a step size Δj. The coded signal Ij thus obtained is
The data is output to a multiplexing unit (not shown), a packet assembling unit, and the like via a corresponding output terminal 36j.

Here, the DC component S0 is derived from the lowest sub-band signal S1.
The DC component S0 is separately encoded and transmitted because the DC component S0 is the most important component for the image signal, and the distortion amount D0 with respect to the DC component S0 is minimized as described later. This is because it can be made smaller. Also, the DC component S0 is removed from the lowest sub-band signal S1 and then coded and transmitted.The DC component S0 is separately coded and transmitted in consideration of a predetermined distortion amount. This is because different distortion amounts can be assigned to the subband signal S1a from which the DC component S0 has been removed. Since the sub-band signal S1 has the largest level, the DC component S0 is obtained from the sub-band signal S1.

The quantization characteristic selection unit 35 performs software-based (in accordance with a predetermined algorithm) a quantization characteristic R for each encoder 34j.
j and Δj, but functionally, the distortion distribution ratio determining unit 35a, the distribution distortion amount calculating unit 35b, and the quantization characteristic calculating unit 35
The quantization characteristics Rj and Δj for each encoder 34j are determined in consideration of the visual spatial frequency characteristics.

The distortion distribution ratio determining unit 35a calculates the allowable total average distortion amount D given through the input terminal 39 for each subband signal.
S1a, S2, ..., the amount of distortion D1, D2, ..., Dn (D1
Is used to determine the distortion distribution ratios W1, W2,..., Wn for distribution to the subband signal S1a.

Originally, it is necessary to distribute the total average distortion amount D also to the DC component S0. However, as described above, since the DC component S0 is the most important component for the image signal, this distortion amount is extremely low. Therefore, in this embodiment, the distortion distribution ratio W0 or the distortion amount with respect to the DC component S0 is required in this embodiment.
D0 is taken from the input terminal 40 and set.

Therefore, the distortion distribution ratio determining unit 35a determines the distortion distribution ratios W1, W2,..., Wn for the sub-band signals S1a, S2,.

The distortion distribution ratio determining unit 35a has a built-in table in which, for example, a frequency is associated with a relative sensitivity value for that frequency on a contrast sensitivity curve indicating spatial frequency characteristics of vision, and When the center frequency ωi is given, a relative sensitivity value to the center frequency ωi is obtained and multiplied by a predetermined value to obtain a distortion distribution ratio Wi. Each strain distribution ratio
Wi is provided to the distribution distortion amount calculation unit 35b.

In addition, each distortion distribution ratio Wi has the following formula: There is a relationship shown in

The distribution distortion calculating unit 35b executes the calculation represented by the following equation Di = Wi · D (4) to calculate each subband signal S1 (S1a), S2 from the total average distortion D and each distortion distribution ratio Wi. , ..., Sn
, Dn are obtained. DC component
If information on distortion is set for S0 in the form of a distortion distribution ratio W0, equation (4) is also performed for this distortion distribution ratio W0.

From equations (3) and (4), the following equation is obtained. Holds, and the total average distortion amount D is distributed.

Each of the distribution distortion amounts Di (in some cases, including the distortion amount D0) obtained by the distribution distortion amount calculation unit 35b in this manner is provided to the quantization characteristic calculation unit 35c. The quantization characteristic calculator 35c calculates the DC component S0 and the sub-band signals S1a and S1 as follows.
,..., Sn, the quantization characteristics Rj (j is 0, 1 (1a),
2 to n), Δj is determined. In addition, the following determination method,
This method ignores overload distortion that occurs when the input to the quantizer exceeds the quantization range.

When the range of the input signal x to the quantizer in the encoder (340 to 34n) is from -xmax to xmax, the change range 2xmax of the input signal is determined by the number of quantization steps 2Rj determined by the number of quantization bits ^Rj. Then, the step size becomes Δj, and the following expression Δj = xmax / 2 ^Rj-1 (6) is established. Further, the distribution distortion amount Dj has a functional relationship with the size of the step size. When the step size for setting the distortion amount to the specified distribution distortion amount Dj is represented by ΔOPTj, the following equation is obtained: ΔOPTj = (12) ^1/2 · Dj ... (7) holds.

Therefore, the maximum step size Δj satisfying the following equation Δj ≦ ΔOPTj (8) and the quantization bit number Rj corresponding to the maximum step size Δj are determined. In practice, the optimum step size Δj and the optimum number of quantization bits Rj are determined by variously changing the number of quantization bits Rj. Therefore, although illustration of signal lines is omitted, a signal input to a quantizer in each encoder 34j is given to a quantization characteristic calculator 35c.

Each of the quantization characteristics Δj and Rj determined in this way is given to the corresponding encoder 34j and used for encoding.

Further, the determined quantization characteristics Δj and Rj are encoded by the encoder 37.
, Which itself is encoded, and the encoded signal J
Is output to a multiplexing unit (not shown), a packet assembling unit, and the like via a corresponding output terminal 38. Transmission by this encoding,
It is transmitted for proper decoding on the decoder side.

Therefore, according to the above-described embodiment, an encoded image signal is obtained in consideration of the spatial frequency characteristics of the human eye, and image quality control in proportion to the subjective evaluation value of the human becomes possible.

Although an image encoder that divides an image signal into subband signals and encodes each of the divided image signals has conventionally existed,
The conventional one does not consider the visual characteristics and does not have an S / N control configuration like the conventional encoder described above. Such an image decoder corresponding to the conventional image encoder can be applied as it is as an image decoder corresponding to the image encoder of the embodiment.

In the above-described embodiment, the encoding method in each encoder 34j is not mentioned, but the encoding method is not limited to a specific encoding method, and various encoding methods such as inter-frame encoding and intra-frame encoding may be used. An encoding method can be applied.

Further, in the above-described embodiment, the number of quantization bits and the step size are shown as the quantization characteristics that are changed according to the visual characteristics. However, other parameters that switch between linear and non-linear may be used.

Further, in the above embodiment, the band division filter
Although the QMF bank is shown, another band division filter may be used, and a two-dimensional filter may be applied unlike the embodiment.

In the above-described embodiment, the contrast sensitivity curve is used as the human visual characteristic. However, the quantization characteristic may be changed according to another visual characteristic (a subjective evaluation method) such as a luminance characteristic. Alternatively, the quantization characteristics may be changed in consideration of the visual characteristics of.

Also, without inputting the signal input to the quantizer in each encoder 34j to the quantization characteristic calculation unit 35c, the range of such a signal is set in advance to calculate the quantization characteristic. Is also good.

[Effects of the Invention] As described above, according to the present invention, an image signal is frequency-separated, and a quantization characteristic for each frequency component is determined and encoded in consideration of the spatial frequency characteristic of the human eye. As a result, an encoded image signal is obtained in consideration of the spatial frequency characteristics of the human eye, and image quality control in proportion to the subjective evaluation value of the human becomes possible.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of an image encoder according to the present invention, and FIG. 2 is a block diagram showing a conventional image encoder. 31 QMF bank, 32 DC component calculator, 33 Subtractor, 340 to 34n Encoder for image signal component, 35
... quantization characteristic selection unit, 35a ... distortion distribution ratio determination unit, 35b ...
Distribution distortion calculation unit, 35c... Quantization characteristic calculation unit, 37... Encoder for quantization characteristics.

────────────────────────────────────────────────── ─── Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI H04N 11/04 G06F 15/70 410 (72) Inventor Mitsuo Tsujikado 1-7-12 Toranomon, Minato-ku, Tokyo Oki Electric Industry (72) Inventor Kenichiro Hosoda 1-7-12 Toranomon, Minato-ku, Tokyo Oki Electric Industry Co., Ltd. (56) References Jiro Kato and Yasuhiko Yasuda, "6-1 Variable Considering Visual Characteristics Quality Control in Rate Coding ”, Image Symposium 4th Symposium Material (PCSJ89), 1989, 93-94

Claims (1)

(57) [Claims] 1. A frequency separating means for separating an input image signal into a plurality of frequency components by a filter, and a plurality of codes for performing coding having predetermined quantization characteristics in accordance with each of the separated frequency components. Using the characteristics of the separated frequency components and visual frequency characteristic information, distributes the specified allowable total distortion amount to each frequency component, and calculates each of the divided distortion amounts from the distributed distortion amount. An image encoder comprising: a quantization characteristic selection unit that selects each of the predetermined quantization characteristics to be applied to the encoder.