JP3175905B2 - Motion vector detection method and apparatus - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and apparatus for detecting a motion vector of an image signal such as a television signal.

[0002]

2. Description of the Related Art In a high-efficiency coding of an image signal of a television or the like, a method of performing motion compensation using a past or future image signal is known. A motion vector used for motion compensation is searched for between the past or future image signal and the encoding target frame. At this time, the position where the motion compensation prediction error such as the sum of squared differences between the corresponding pixels between the small blocks is minimized is determined as the motion vector. Motion compensation is performed using the determined motion vector, the motion compensation prediction error signal is encoded, and the motion vector is also encoded to compress the information amount.

FIG. 3 is a block diagram for explaining an example of a conventional motion vector determination method. As an image coding method, it is assumed that standard motion compensation and discrete cosine transform are used.

First, an image input from an input terminal 101 is divided by a small block dividing unit 102 into small blocks of M × N (M and N are integers of 1 or more). Small block image signal 103 and 1 stored in frame memory 107
A motion vector detection unit 104 obtains a motion vector 105 that minimizes the motion compensation prediction error signal from the image signal before the frame, and the motion compensation unit 106 performs motion compensation based on the motion vector 105, and a subtractor 108 After the motion-compensated prediction error signal is obtained in (1), the discrete cosine transform unit 109 performs a discrete cosine transform, which is one of orthogonal transforms, on the motion-compensated predicted error signal.

In the quantization step calculator 114,
The quantization step 116 is set based on the buffer memory occupancy 115. Set quantization step 116
, The discrete cosine transform coefficient is quantized by the quantization unit 110 and is variable-length coded by the variable-length coding unit 117.

The difference between the motion vector of the previous block stored in the motion vector memory 118 and the motion vector of the motion vector 105 is calculated by the subtractor 119. Is done. Generally, since the difference between motion vectors is small, a short code is assigned to a small difference.

[0007] The signal variable-length coded by the variable-length coding section 117 is input to the buffer memory 120 and output to the coded data output terminal 121 at a constant bit rate. The quantized discrete cosine transform coefficient is inversely quantized by an inverse quantization unit 111 and an inverse discrete cosine transform unit 112
After the inverse discrete cosine transform, the data of the motion-compensated reference frame is added to the data of the reference frame by the adder 113 and stored in the frame memory 107 for motion compensation prediction of the next frame.

According to such a method, since the motion of the moving image signal can be compensated, the prediction efficiency can be increased, and as a result, the image quality of the encoded image can be improved.

[0009]

In the conventional method described above, as shown in FIG. 4, a motion vector in which a motion compensation prediction error such as a sum of squares of differences between corresponding pixels between small blocks is minimized is always calculated. When detecting and transmitting a motion vector, a difference (FIG. 5) from a predicted value calculated from a motion vector of a neighboring small block coded before the current block is calculated. Because it is small,
A short code is assigned to the one with a small difference and encoded. On the other hand, orthogonal transform coefficients such as discrete cosine transform are quantized by a predetermined quantization step, and then are encoded. Then, the code amount of the motion vector and the orthogonal transform coefficient occupies almost the entire code amount. Here, the relationship between the motion compensation prediction error signal and the quantization step will be considered.

First, when the quantization is performed to some extent, the smaller the motion compensation prediction error is, the smaller the code amount of the orthogonal transform coefficient after quantization is. Therefore, it is necessary to detect the motion vector having the minimum motion compensation prediction error. There is.

On the other hand, when quantization is coarse, even if a motion vector having a large motion compensation prediction error is detected, the number of codes of the orthogonal transform coefficient after quantization is rarely increased. Therefore, in such a case, a motion vector is selected such that the difference between the prediction value calculated from the motion vector of the neighboring small block encoded before the current block and the motion vector of the current block is small. As a result, the total amount of generated information may decrease. However, according to the conventional method, since the quantization step is not considered in the detection of the motion vector, the motion vector that minimizes the motion compensation prediction error is always detected. As a result, the difference between the motion vectors may be large, and there is a problem that the total amount of generated information does not decrease.

An object of the present invention is to select a motion vector detecting method and apparatus capable of reducing the amount of information generated by a motion vector.

[0013]

A motion vector detecting method according to the present invention divides image data of a current frame into small blocks with respect to a digital image signal such as a television signal.
In a motion vector detecting method for obtaining a motion vector from an image signal of a past or future reference frame stored in a frame memory for each encoding target small block, quantization applied to quantization of the encoding target small block An allowable range T is determined according to the size of the step, and a plurality of motion vectors whose prediction error with respect to the small block of the reference frame is smaller than the minimum prediction error plus the allowable range T are obtained. Among the vectors, a vector having the smallest difference from the motion vector of the small block near the small block to be encoded which has already been encoded is set as the motion vector of the small block to be encoded.

In the above motion vector detecting method, the quantization step is Q, and the size of the small block is M × N (M, N
Is an integer of 1 or more), the allowable range T is T = Q · (M
× N).

According to the motion vector detecting device of the present invention, an input terminal for inputting an image, an output terminal for outputting encoded data, and a small block for dividing the image input from the input terminal into small blocks A dividing unit, a frame memory for storing an image signal of one frame before, a buffer memory for inputting a variable-length coded signal and outputting to the output terminal at a constant bit rate, and a buffer memory occupancy of the buffer memory A quantization step calculation unit for setting a quantization step based on the prediction error allowable range setting for setting an allowable range of a prediction error signal at the time of motion search of a small block to be encoded based on the quantization step Unit, the image signal divided into small blocks by the small block dividing unit and the image signal of one frame before stored in the frame memory. A motion vector detecting unit that detects one or more motion vectors in which the prediction error index of the motion compensated prediction error signal is smaller than the minimum prediction error index plus the allowable range; A motion vector memory to be stored, and a minimum difference for detecting a minimum value of a difference between the motion vector within the allowable range obtained by the motion vector within the allowable range and the motion vector of the previous block stored in the motion vector memory. A motion vector detecting unit, and a motion compensating unit that performs motion compensation on a previous frame motion vector stored in the frame memory based on the minimum difference motion vector obtained by the minimum difference motion vector detecting unit. Subtracting the image signal divided into small blocks by the small block division unit and the output of the motion compensation unit, A first subtractor that outputs a difference signal; and a second subtractor that calculates a difference between a minimum difference motion vector detected by the minimum difference motion vector detection unit and a previous block of motion vectors stored in the motion vector memory. (2), an orthogonal transform unit for performing orthogonal transform on the motion compensated prediction error signal, and a quantizing unit for quantizing orthogonal transform coefficients output from the orthogonal transform unit using the quantization step. A variable-length coding unit that performs variable-length coding on a difference between the orthogonal transform coefficient quantized by the quantization unit and the motion vector calculated by the second subtractor, and stores the difference in the buffer memory; An inverse quantization unit that inversely quantizes the orthogonal transform coefficient, an inverse discrete cosine transform unit that performs an inverse discrete cosine transform of the inverse quantized orthogonal transform coefficient, and an inverse discrete cosine transformed orthogonal transform coefficient. There is an adder for adding the data of the reference frame motion-compensated by the motion compensation unit and storing the data in the frame memory.

[0016]

First, an image signal is divided into small blocks, and an allowable range T of a motion compensation prediction error is set based on a quantization step of a small block to be coded. Next, as shown in FIG. 1A, a motion vector whose motion compensation prediction error falls within a range obtained by adding the allowable range T of the motion compensation prediction error previously set to the minimum value of the motion compensation prediction error is past or past. Detect from future image signals. Subsequently, among the motion vectors falling within the allowable range, a motion vector that minimizes the difference between the predicted value of the motion vector and the motion vector of the small block to be coded is detected (FIG. 1B).

According to the method described above, the amount of information generated from a motion vector can be reduced without increasing the amount of information generated due to an increase in a motion compensation prediction error.

[0018]

Next, embodiments of the present invention will be described with reference to the drawings.

FIG. 2 is a block diagram of a motion vector detecting device according to one embodiment of the present invention. Blocks with the same reference numbers as those in FIG. 3 indicate the same blocks. In the present embodiment, the quantization step value is used as the allowable range of the prediction error, and the sum of the absolute differences is used as the index of the prediction error in the motion search. It is assumed that motion compensation discrete cosine transform is used as an image coding method.

The motion vector detecting apparatus according to the present embodiment further includes a prediction error allowable range setting unit 122 and replaces the motion vector detecting unit 104 with the motion vector detecting unit 1 within the allowable range.
24 and a minimum difference motion vector detection unit 126.

First, the image input from the input terminal 101 is divided into small blocks of M × N (M and N are integers of 1 or more) in a small block dividing section 102. In the quantization step calculator 114, the buffer memory occupancy 1
15, ie, the buffer memory occupancy 115
Is large, the quantization step Q (116) is finely set when the buffer memory occupancy 115 is small. Based on the quantization step Q (116), the prediction error allowable range setting unit 122 sets the allowable range T (1) of the prediction error signal at the time of the motion search of the small block to be encoded.
23) is set. The prediction error allowable range T (123) is
Since the permissible range per pixel is the value of the quantization step Q and the sum of absolute differences is used as an index of the magnitude of the motion compensation prediction error, the permissible error per pixel is Q, so T = Q ·
(M × N).

Next, the image signal 1 divided into small blocks
03 and the image signal of the previous frame stored in the frame memory 107 from the motion vector detector 12 within the allowable range.
4, the sum of absolute differences AD of the motion compensated prediction error signals
One or more motion vectors satisfying AD _min ≤ AD ≤ T are detected.
Here, AD _min is the minimum difference absolute value sum of the motion compensation prediction error signal of the small block to be encoded.

The motion vector within the obtained prediction error allowable range

[0024]

[Outside 1] And the motion vector of the previous block stored in the motion vector memory 118

[0025]

[Outside 2] From the minimum difference motion vector detection unit 126

[0026]

(Equation 1) Is detected. A motion compensation unit 106 performs motion compensation based on the motion vector 105, and a subtractor 108 calculates a motion compensation prediction error signal. Cosine transform is performed.

Further, using the quantization step Q (116) set in the quantization step calculation section 114,
The discrete cosine transform coefficients are quantized by a quantization unit 110,
The variable length coding unit 117 performs variable length coding.

The minimum difference motion vector detecting section 126
Also, the difference between the motion vector 105 detected in step (1) and the motion vector of the previous block stored in the motion vector memory 118 is calculated by the subtractor 119, and is variable-length coded by the variable-length coding unit 117. . At this time, a shorter code is assigned as the difference between the motion vectors is smaller. Each signal subjected to variable-length encoding by the variable-length encoding unit 117 is
The data is input to the buffer memory 120 and output to the encoded data output terminal 121 at a constant bit rate. The quantized discrete cosine transform coefficients are inversely quantized by an inverse quantizer 111, inverse discrete cosine transformed by an inverse discrete cosine transform unit 112, and subjected to motion compensated reference frame data and an adder 113. The sum is added and stored in the frame memory 107 for motion compensation prediction of the next frame.

With such a configuration, the code amount of the motion vector can be reduced without increasing the code amount of the quantized discrete cosine transform coefficient.

In the embodiment described above, the sum of absolute differences is used as an index of the magnitude of the motion compensation prediction error.
Although the value of the quantization step is used as it is for setting the allowable range of the motion compensation prediction error signal, when another index is used, it can be dealt with by appropriately changing the setting method of the allowable range.

[0031]

As described above, according to the present invention, the amount of information generated by a motion vector can be reduced. Therefore, when coding is performed at a constant bit rate, the amount of information allocated to a prediction error signal or the like can be increased. Yes, and as a result,
There is an effect that the image quality can be improved based on a fixed amount of information.

[Brief description of the drawings]

FIG. 1 is a diagram showing a method for determining a motion vector according to the present invention;
FIG. 7A is a diagram illustrating detection of a motion vector by setting an allowable range T, and FIG. 7B is a diagram illustrating determination of a motion vector based on a difference between the motion vector and a predicted value.

FIG. 2 is a block diagram showing a motion vector detecting device according to one embodiment of the present invention.

FIG. 3 is a diagram illustrating a configuration example of conventional quantization control based on image signal dispersion.

FIG. 4 is an explanatory diagram showing a conventional motion vector detection method.

FIG. 5 is a diagram illustrating a difference between motion vectors.

[Explanation of symbols]

Reference Signs List 101 input terminal 102 small block division unit 103 small block image signal 104 motion vector detection unit 105 motion vector 106 motion compensation unit 107 frame memory 108 subtractor 109 discrete cosine transformation unit 110 quantization unit 111 inverse quantization unit 112 inverse discrete cosine Transformer 113 Adder 114 Quantization step calculator 115 Buffer memory occupancy 116, Q quantization step 117 Variable length encoder 118 Motion vector memory 119 Subtractor 120 Buffer memory 121 Output terminal 122 Prediction error allowable range setting unit 123, T prediction error allowable range of small block to be encoded 124 motion vector detector within allowable range 125, Vi motion vector whose prediction error signal is within allowable range 126 minimum difference motion vector detector

Claims (3)

(57) [Claims] An image data of a current frame is divided into small blocks with respect to a digital image signal such as a television signal, and a past or future reference frame stored in a frame memory is stored for each small block to be encoded. In the motion vector detection method for obtaining a motion vector from an image signal, an allowable range T is determined by the size of a quantization step applied to quantization of the encoding target small block, and a prediction error between the reference frame and the small block is: A plurality of motion vectors that are smaller than the minimum prediction error plus the allowable range T are obtained, and among those motion vectors, the motion vector of the small block near the already-encoded small block to be coded is calculated. A motion vector of a small block to be coded is determined as a motion vector having a minimum difference from the motion vector. Method. 2. Assuming that the quantization step is Q and the size of a small block is M × N (M and N are integers of 1 or more), the allowable range T is T = Q · (M × N). Item 7. The motion vector detecting method according to Item 1. 3. An input terminal for inputting an image, an output terminal for outputting coded data, a small block dividing unit for dividing an image input from the input terminal into small blocks, and one frame before. A frame memory for storing the image signal of the above, a buffer memory for inputting the variable-length-encoded signal and outputting it to the output terminal at a constant bit rate, and a quantization step based on the buffer memory occupancy of the buffer memory. A quantization step calculation unit that sets a prediction error allowable range setting unit that sets an allowable range of a prediction error signal during motion search of a small block to be encoded based on the quantization step; The motion compensation prediction error signal is obtained from the image signal divided into small blocks by the section and the image signal of one frame before stored in the frame memory. A motion vector detecting unit that detects one or more motion vectors whose measurement error index is smaller than the minimum prediction error index plus the allowable range; a motion vector memory that stores a motion vector one block before; A minimum difference motion vector detection unit that detects a minimum value of a difference between a motion vector within an allowable range obtained by the motion vector detection unit within the allowable range and a motion vector one block before stored in the motion vector memory; A motion compensator that performs motion compensation on the motion vector of one frame before stored in the frame memory based on the minimum difference motion vector obtained by the minimum difference motion vector detector; Subtracting the output of the motion compensation unit from the image signal divided into small blocks and outputting a motion compensation prediction error signal. Subtracter and, second calculating a difference between said minimum differential motion vector detection unit 1 block previous motion vector, wherein a minimum difference motion vector detected are stored in the motion vector memory at
A subtractor, an orthogonal transformation unit that performs orthogonal transformation on the motion compensation prediction error signal, and a quantization unit that quantizes orthogonal transformation coefficients output from the orthogonal transformation unit using the quantization step. The orthogonal transform coefficient quantized by the quantization unit;
A variable-length encoding unit that encodes the difference between the motion vectors calculated by the subtractor in the variable length encoding unit and stores the difference in the buffer memory; an inverse quantization unit that inversely quantizes the quantized orthogonal transform coefficients; An inverse discrete cosine transform unit for performing an inverse discrete cosine transform of the obtained orthogonal transform coefficient, an orthogonal transform coefficient subjected to the inverse discrete cosine transform, and data of the reference frame motion-compensated by the motion compensation unit, and stored in the frame memory. A motion vector detecting device having an adder.