JPH0420088A - Inter-frame coder - Google Patents

Abstract

PURPOSE:To prevent deterioration in picture quality and increase in a code quantity regardless of quantity of movement by controlling the frequency characteristic and the quantizing characteristic of an error signal corresponding to the moving quantity of a picture. CONSTITUTION:A moving quantity calculation circuit 76 calculates movement from a moving vector inputted from a movement detection circuit 69 and the frequency band characteristic of a filter circuit 71 is controlled depending on the moving quantity. Moreover, the quantization characteristic in a quantization circuit 74 is outputted from the moving quantity calculation circuit 76 and controlled depending on the moving quantity inputted via a switch 79. Since the frequency characteristic and the quantization characteristic of an error signal are controlled corresponding to the moving quantity of the picture in this way, deterioration in picture quality and increase in a code quantity are prevented regardless of quantity of movement.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a coding device for a moving picture signal, and particularly to an interframe coding device for a pre-11 error of a moving picture signal.

[Prior Art] FIG. 10 is a block diagram showing the configuration of an example of a conventional interframe encoding device. As shown in FIG. 11, a reference image is generated every N frames (N=4 in this example). 31.35, and the images 32 to 34 between the reference images 31 and 35 are encoded using intra-frame encoding, that is, the correlation of images within the frame.

When the reference image 31 (35) is being input, the switches 61 and 72 are switched to the contact a side. As a result, the reference image is transferred to the orthogonal transform circuit 73 via the switches 61 and 72.
input, it is orthogonally transformed.

The orthogonal transform coefficients output from the orthogonal transform circuit 73 are input to a quantization circuit 74 and quantized.

The quantization circuit 74 generates encoded values of orthogonal transform coefficients and quantization control information, and outputs them to the encoding circuit 75. The encoding circuit 75 encodes the input encoded value and quantization control information, and outputs the encoded value to a circuit (not shown) via the mixer 78.

The switch 62 is switched every four frames to alternately store input reference images in frame memories (FM) 66 and 67.

On the other hand, when the interframe images 32, 33, and 34 are being input, the switches 61, 72 are switched to the contact side. As a result, the inter-frame images 32 to 34 can be sequentially stored in the frame memories 63 to 65.

The motion detection circuit 69 divides the interframe image data 32 (33, 34) in the frame memory 63 into blocks of a predetermined size (for example, 8×8 pixels), and stores each block in the frame memory 66, 67. Detects the motion vector from the stored reference plane ($31°35) immediately before or after.

The forward/backward motion compensation circuit 68 predicts the predicted image Pa from the immediately previous reference image 31, the predicted image Pb from the immediately subsequent reference image 35, and the prediction of the blocks of the interframe images 32 (33, 34) sequentially output from the frame memory 63. A preliminary image Pc is generated from the average value of each corresponding pixel of images Pa and pb. Then, the predicted image with the highest correlation among them is selected for each block and output to the :g calculator 70.

Furthermore, information regarding the selected predicted image is output to the encoding circuit 77 as selection mode data.

The subtracter 70 receives the interframe image 32 (33, 34) inputted from the frame memory 63 and the forward/backward motion compensation circuit 68.
Calculate the difference (prediction error) with the predicted image that can be input more easily,
It outputs to the filter circuit 71 (filter means). After the filter 71 limits the input signal to a fixed band characteristic,
Output.

The output of the filter circuit 71 is input to an orthogonal transformation circuit 73 via a switch 72, and is orthogonally transformed.

The orthogonal transform coefficients generated by the orthogonal transform are sent to the quantization circuit 7.
4, it will be quantized. Although the quantization characteristic in the quantization circuit 74 is fixed, the inter-reference image is processed with a quantization characteristic coarser than that of the reference image.

The quantization value and quantization control information generated in the quantization circuit 74 are input to the encoding circuit 75, encoded, and then supplied to the mixer 78.

The encoding circuit 77 encodes the motion vector output from the motion detection circuit 69 and the selection mode output from the forward/backward motion compensation circuit 68, and outputs it to the mixing equation 78. The mixer 78 mixes the data that can be input from the encoding circuits 75 and 77, and outputs the mixed data to a circuit (not shown).

In this way, in encoding each reference inter-image,
It is made independent from other reference inter-images and eliminates the accumulation of coding errors. The filter circuit 71 limits the band of the prediction error, and the quantization circuit 74 performs coarse quantization, thereby reducing deterioration in image quality and significantly reducing the amount of code.

[Problems to be Solved by the Invention] However, in the conventional device, the filter circuit 7
Since the band-limiting characteristics of 1 are fixed, there are problems in terms of visual characteristics, such as a decrease in resolution being easily detected and blurring being noticeable in images with small movement.

Furthermore, in images with large movements, quantization errors are difficult to detect due to visual characteristics, and prediction errors are generally large, but in images with small movements, quantization errors are easily detected. Therefore, since the quantization characteristic of the quantization circuit 74 is fixed, the amount of code increases for images with large motion, and the quantization error becomes noticeable for images with small motion.

The present invention was developed in response to this situation, and is intended to improve image quality and reduce the amount of code, regardless of the magnitude of movement.

[Means for Solving the Problems 1] The interframe encoding device of the present invention sets reference images at intervals of a plurality of frames, and with respect to images between the reference images, the reference image immediately before or after the reference image, or the reference image immediately after the reference image, or In an interframe encoding device for a moving image signal that encodes a prediction error with a predicted θ image obtained using a decoded image of encoded data, a motion amount calculation means that calculates the amount of motion of an image on a time axis. and filter means for controlling the frequency band of the prediction error in accordance with the output of the motion estimation means;
The present invention is characterized by comprising a quantization means for controlling the error or the quantization characteristic of the orthogonal transform coefficient obtained by orthogonally transforming the error in accordance with the output of the motion amount calculation means.

1 Effect] In the interframe encoding device having the above configuration, the frequency characteristics and quantization characteristics of the error signal are controlled in accordance with the amount of motion of the image. Therefore, regardless of the magnitude of motion, deterioration of image quality and increase in code amount can be prevented.

[Embodiment] FIG. 1 is a block diagram showing the configuration of an embodiment of an interframe encoding device of the present invention.

This embodiment also performs the same interframe encoding process as in FIG. 11.

When the reference plane 631 (35) is being input, the switches 61, 72, and 79 are switched to the contact a side. Thereby, the reference image is input to the orthogonal transformation circuit 73 via the switches 61 and 72, and is orthogonally transformed. Orthogonal transformation circuit 73
The orthogonal transform coefficients output by are input to the quantization circuit 74,
Can be quantized.

At this time, since the motion amount O is set via the switch 79, the quantization circuit 74 quantizes the reference image as a still image.

The quantization circuit 74 generates encoded values of orthogonal transform coefficients and quantization control information, and outputs them to the encoding circuit 75. The encoding circuit 75 encodes the input encoded value and quantization control information, and outputs the encoded value to a circuit (not shown) via the mixer 78.

The switch 62 is switched every four frames to alternately store input reference images in frame memories (FM) 66 and 67.

On the other hand, when the interframe images 32, 33, and 34 have been inputted, the switches 61, 72, and 79 are switched to the contact side. As a result, the interframe images 32 to 34 are
The frames are sequentially stored in the frame memories 63 to 65.

The motion detection circuit 69 divides the interframe image data 32 (33, 34) in the frame memory 63 into blocks of a predetermined size (for example, 8×8 pixels), and stores each block in the frame memory 66, 67. A motion vector from the reference image 31.degree. 35 stored immediately before or after is detected.

The forward/backward motion compensation circuit 68 calculates the predicted image Pa from the immediately preceding reference image 31, the predicted image pb from the immediately following reference image 35, and the prediction of the block of the interframe image 32 (33, 34) read from the frame memory 63. image (Pre-Kl+image (gJPc,
are generated respectively. Then, the most correlated predicted image is selected for each block and output to the subtracter 70. Furthermore, information regarding the selected predicted image is output to the encoding circuit 77 as selection mode data.

The subtracter 70 receives the interframe image 32 (33, 34) inputted from the frame memory 63 and the forward/backward motion compensation circuit 68.
Calculate the difference (prediction error) from the predicted image input by
It outputs to the filter circuit 71 (filter means).

The motion amount calculation circuit 76 (motion amount calculation means) calculates the amount of motion from the motion vector input from the motion detection circuit 69 (for example, calculates the sum of the absolute values of the horizontal and vertical components of the motion and the vector as the amount of motion). do).

The frequency band characteristics of the filter circuit 71 are controlled in accordance with this amount of movement.

The output of the filter circuit 71 is input to an orthogonal transformation circuit 73 via a switch 71, and is orthogonally transformed.

The orthogonal transform coefficients generated by the orthogonal transform are sent to the quantization circuit 7.
4 and is quantized. The quantization characteristic in the quantization circuit 74 is controlled in accordance with the amount of motion output from the motion amount calculation circuit 76 and inputted via the switch 79.

The quantization value and quantization control information generated in the quantization circuit 74 are input to an encoding circuit 75 and, after being encoded, are supplied to a mixer 78.

The encoding circuit 77 encodes the motion vector output from the motion detection circuit 69 and the selection mode output from the forward/backward motion compensation circuit 68 and outputs it to the mixer 78 . The mixer 78 mixes the data input from the encoding circuits 75 and 77 and outputs the mixed data to a circuit (not shown).

FIG. 2 is a block diagram showing the configuration of one embodiment of the filter circuit 71.

The amount of motion output from the amount of motion calculating circuit 76 is input to the filter mixture value determining circuit 81 and converted into a filter mixture value f. For example, the filter mixture value determining circuit 81
It is composed of a ROM having input/output characteristics as shown in the figure.

That is, within a range until the amount of motion reaches a predetermined value, the filter mixture value f changes in direct proportion to the amount of motion, and beyond that range, it becomes a constant value (maximum value m).

The filter mixture value f is input to a subtracter 83, and the difference between it and the maximum value m is calculated. The output (m-f) of the subtracter 83 is input to a multiplier 84 and multiplied by the prediction error input from the subtracter 70.

The prediction error is also limited to a predetermined frequency band by the low-pass filter circuit 82J and then input to the multiplier 85. The multiplier 85 multiplies the low-frequency component of the prediction error extracted by the low-pass filter 82 (this filter mixed value f). The adder 86 adds the outputs of the multipliers 84 and 85 and outputs the result to the divider 87.

The divider 87 divides the data from the adder 86 by the maximum value m and outputs it to the orthogonal transform circuit 73.

FIG. 4 is a block diagram showing the configuration of one embodiment of the quantization circuit 74.

The quantization step determination circuit 92 converts the motion amount output from the motion amount output circuit 76 into a corresponding quantization step g, and outputs it to the quantizer 91. The quantization step determination circuit 92 is constituted by, for example, a ROM having input/output characteristics as shown in FIG. That is, when the amount of motion is O, the quantization step g is g. After that, the quantization step g changes in proportion to the amount of motion (as the amount of motion increases, the quantization step becomes coarser).

FIG. 6 is a block diagram showing the configuration of another embodiment of the quantization circuit 74.

The amount of motion output from the amount of motion calculation circuit 76 is input to the threshold control circuit 113, and is converted into a corresponding threshold control value thf. The threshold control circuit 113 has input/output characteristics as shown in FIG. 7, for example. That is, in the range until the amount of movement becomes ■1, the threshold control value thf gradually decreases in accordance with the amount of movement, and when it becomes more than ■1, the threshold control value thf decreases to a constant value (1/2).
becomes.

The quantization step control circuit 112 includes a motion amount calculation circuit 76
The amount of motion outputted from is converted into a corresponding quantization step g and outputted to the quantizer 111 and the multiplier 115, and quantization control information is outputted to the encoding circuit 75. The quantization step control circuit 112 is also constructed from a ROM.

The multiplier 115 multiplies the threshold control value thf by the quantization step g and outputs the result to the threshold circuit 114.

The threshold circuit 114 compares the absolute value of the orthogonal transform coefficient input from the orthogonal transform circuit 73 and the output of the multiplier 115, and if the absolute value of the orthogonal transform coefficient is smaller than the output of the orthogonal transform coefficient subtracter 115, the orthogonal transform coefficient is changed. 0, otherwise,
The orthogonal transform coefficients inputted from the orthogonal transform circuit 7 are output to the quantizer 111 as they are.

The quantizer 111 quantizes the input orthogonal transform coefficients in accordance with the quantization step g, and outputs the quantized coefficients to the encoding circuit 75.

FIGS. 8 and 9 show other interframe encoding processing to which the present invention is applicable.

In FIG. 8, a reference image is set every N frames, and intraframe encoding is performed every MXN frames. The reference image from the intra-frame encoded image to the next intra-frame encoded image is the same as the previous reference image or the predicted image predicted from the decoded image of the encoded data. is encoded.

In FIG. 9, a reference image is set every N frames, and intraframe encoding is performed every MXN frames.

The reference images from the intra-frame encoded t: image to the next intra-frame encoded reference image are the reference images of the immediately preceding and immediately following intra-frame encoded images, or the decoded image of the encoded data. The pre-II error with the pre-processed predicted image is encoded.

[Effects of the Invention] As described above, according to the interframe encoding device of the present invention, the frequency characteristics and quantization characteristics of the error signal are controlled in accordance with the amount of motion of the image, so that the magnitude of the motion can be controlled. Regardless, it is possible to prevent deterioration in image quality and increase in code amount.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of an embodiment of the interframe encoding device of the present invention, FIG. 2 is a block diagram showing the configuration of an embodiment of the filter circuit of FIG. 1, and FIG. FIG. 4 is a block diagram showing the configuration of an embodiment of the quantization circuit of FIG. 1, and FIG.
The figure is an input/output characteristic diagram of the quantization step determination circuit in Figure 4,
FIG. 6 is a block diagram showing the configuration of another embodiment of the quantization circuit shown in FIG. 1, FIG. 7 is an input/output characteristic diagram of the threshold control circuit shown in FIG. 6, and FIGS. A first diagram illustrating the operation of interframe encoding processing to which the invention is applicable.
FIG. 0 is a block diagram showing the configuration of an example of a conventional interframe encoding circuit, and FIG. 11 is a diagram illustrating the operation of the interframe encoding process to explain the operation of the example of FIG. 10. 41 to 45.63 to 67...Frame memory, 7
1-...Filter circuit (filter means), 74...Quantization circuit (quantization means), 76...Motion calculation circuit (motion calculation means). Patent applicant: Victor Japan Co., Ltd.

Claims (1)

[Claims] Prediction obtained by setting reference images at a plurality of frame intervals, and using the reference image immediately before or after the reference image, or the decoded image of the encoded data, regarding the image between the reference images. An interframe encoding device for a moving picture signal that encodes a prediction error with an image, comprising: a motion amount calculation unit that calculates a motion amount of the image on a time axis; and a frequency band of the prediction error that is determined by the motion amount calculation means. filter means for controlling in response to the output of the means; and the prediction error;
or quantization means for controlling the quantization characteristics of orthogonal transform coefficients obtained by orthogonally transforming the coefficients in accordance with the output of the motion amount calculation means.