JPH057372A - Motion vector detector - Google Patents

Abstract

(57) [Abstract] [Purpose] It has a mode for obtaining a motion vector by using a color component and a mode for obtaining a motion vector by using a luminance component. Enable to detect a certain motion vector. [Structure] The current image data is input to the evaluation circuit 21, and the evaluation circuit 21 determines from the data distribution of the current image block data that
It is determined whether the block matching is performed by the luminance component (luminance component mode) or the color component (color component mode), and the result is sent to the arithmetic circuit 22. The arithmetic circuit 22
Then, block data which is considered to match the current image block and the previous frame is input, and the inter-block error is calculated for the two block data based on the result of the evaluation circuit 21 by a predetermined evaluation interval mathematical expression. The motion vector is detected by selecting the smallest one and the vector at that time from the error data of many block matchings calculated for one current block.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion vector detecting device for detecting a motion vector of a subject between frames.

[0002]

2. Description of the Related Art A block matching type is known as a motion vector detecting method.

The block matching type motion vector detection method is (8 × 8) between the previous frame and the current frame.
Or, take pixels such as (16 × 16) to make one block,
Matching is performed for each block, and a motion vector is given to each block of the current frame.

Next, the basic idea of this motion vector detection method will be described in detail.

A block of horizontal m pixels and vertical n pixels whose upper left corner pixel position is (i, j) on the current frame is B
_ij represents (m, n) and represents a block in the same position on the previous frame P _ij (m, n) and. At this time, the evaluation function of the error between B and P is

[0006]

[Outer 1] When x and y are changed from ± 0 to s (s is a parameter of the search range), the movement is performed with the displacements x and y from i and j such that the smallest value of f is obtained. Vector.

Generally, a color moving image is composed of a frame in which luminance data (Y) and color difference data (Cr, Cb, etc.) are divided, and a large amount of structural information is used to search for a moving vector. The included luminance data (Y) is used.
In addition, the motion of the entire screen is calculated by finding a new vector centered on the vector obtained in the immediately preceding block and the surrounding blocks, and obtaining the correlation between the vector generated from each block and the block position. A method for estimating is known.

[0008]

However, an image having actual meaning (an image having information as a picture such as a person, a landscape, a test pattern, etc., not just a set of RGB luminance data) has many colors. In general, a set of points of the same hue means one grouped object.

In the block matching using the luminance data, which is generally performed, the calculation error value can be locally minimized, but it is difficult to detect the same object in the image. is there.

Therefore, when trying to obtain the motion of the entire screen from individual vectors, or when trying to use the vectors generated in the previous block and the processed surrounding blocks for vector search in the current block, a large error is caused. Will be invited.

In view of the above point, the present invention has an object to provide a motion vector detecting device capable of reliably extracting the same object as the object in the current screen and the previous screen and detecting a reliable motion vector. To do.

[0012]

A motion vector detecting apparatus according to the present invention detects color data, which is one of color image information, in a motion vector detecting apparatus for detecting a motion vector of a subject between adjacent frames in block units. It is characterized by having a first mode in which a motion vector is detected by using it and a second mode in which a motion vector is detected by using luminance data which is one of color image information.

[0013]

According to the above invention, a first mode in which a motion vector is detected by color data that is one of color image information and a second mode in which a motion vector is detected by luminance data that is one of color image information are provided. By having these two modes, it is possible to reliably search the same object as the object in the current screen from the previous screen and detect a reliable motion vector.

[0014]

Embodiments of the present invention will be described in detail below with reference to FIGS. 1 and 2.

FIG. 1 is a block diagram of a moving picture compression coding apparatus using a vector detection circuit according to the present embodiment, and FIG. 2 is a block diagram showing a configuration of the vector detection circuit.

In FIG. 1, 1 is a vector detection circuit which will be described in detail later, 2 is a frame memory for storing a previous frame, 3 is a subtraction circuit, 4 is a buffer memory, 5 is a quantization circuit, and 6 is lossless encoding. Circuit, 7 is the quantization circuit 3
Is an inverse quantization circuit which operates in the completely opposite manner, and 8 is an adder circuit.

The compression operation of the moving picture compression coding apparatus configured as described above will be described in detail.

Through the signal lines 101 and 102, the image data of the current frame (lightness L ^* , chromaticity a ^* , b ^{* is} 8 bi each.
(expressed by t) is input to the vector detection circuit 1 for each block in a form in which one frame is divided into a plurality of blocks.

On the other hand, the block address signal is input to the frame memory 2 via the signal line 103.

Image data in the vector search range around the block is read from the frame memory 2 in accordance with the block address signal and sent to the vector detection circuit 1 via the signal line 104.

The vector detection circuit 1 outputs to the signal lines 105 and 106 the block data evaluated to be the optimum from the predetermined search range.

Next, the current image block delayed for a predetermined time by the buffer memory 4 is input to the subtraction circuit 3 via the signal line 107, and at the same time, the value of the optimum block is input to the subtraction circuit 3 via the signal line 105. Is entered.

The subtraction circuit 3 subtracts the values of the input current image block and the input optimum block, and the difference value data is input to the quantization circuit 5 and quantized.

The quantized data is input to the lossless encoding circuit 6 and the inverse quantization circuit 7.

The lossless encoding circuit 6 assigns a lossless code to the quantized data, and the lossless encoded data is sent to the signal line 10.
8 is output.

The inverse quantization circuit 7 reproduces the difference value, and the difference value and the optimum block data are input to the addition circuit 8 via the signal line 109 and added. The added data is stored in the frame memory 2.

Next, the vector detection circuit 1 shown in FIG.
Will be described in detail with reference to FIG.

First, the previous frame image data in the search area is input to the memory 20 from the frame memory 2 (see FIG. 1) via the signal line 104. Current image data is signal line 1
It is input to the evaluation circuit 21 via 02.

In the evaluation circuit 21, the block matching is performed by the lightness (L) according to the data distribution of the original block data.
^* ) Brightness block matching mode, chromaticity (a ^* ,
It is determined (evaluated) which of the chromaticity block matching modes to be performed in b ^* ), and the result is sent to the arithmetic circuit 22 via the signal line 201. An evaluation method for determining whether the block matching is performed in the lightness mode or the chromaticity mode will be described later.

The arithmetic circuit 22 uses MPX (Multi) for the vector to the target block to be matched from 202.
Plexor) 23.

The MPX 23 generates an address signal according to the input vector, the address signal is transmitted to the memory 20 via a signal line 203, and is read from the memory 20 by the address data via a signal line 204. The output block data is received, and the block data is input to the arithmetic circuit 22 via the signal line 205.

The arithmetic circuit 22 receiving the two block data calculates the inter-block error by the following evaluation function formula.

In case of L ^* mode

[0034]

[Outside 2] In case of a ^* and b ^* modes

[0035]

[Outside 3] (Subscripts are the same as in the conventional example, those with dashes indicate blocks on the previous frame).

The error value between blocks and the vector data of the matched block are output from the arithmetic circuit 22 to the signal line 2.
It is input to the comparison circuit 24 via 06.

The comparator circuit 24 selects the smallest one and the vector at that time from a large number of block matching error data calculated for one current block. When the block matching for one current block is completed, the arithmetic circuit 22 outputs the vector output circuit 25.
To the end via the signal line 208. In response to this, the vector output circuit 25 reads the vector data stored in the comparison circuit 24, outputs the vector data to the outside via the signal line 106, and reads the block data to the MPX 23 via the signal line 209. Input the signal. The MPX 23 reads the best matching block data from the memory 20 according to the read signal and outputs it to the outside through the signal line 105.

Next, a method of evaluating whether the block matching in the evaluation circuit 21 shown in FIG. 2 is performed with lightness data (lightness mode) or chromaticity data (chromaticity mode) will be described.

In the input block, L ^* = 255 is white in lightness (L ^* ) and a ^* in chromaticity (a ^* , b ^* ) ^.
= 127 (same for b ^* ) is colorless in the axial direction, and a ^* =
It is assumed that 0, a ^* = 255 (same for b ^* ) has the highest saturation in both directions of the axis. The block is composed of (m × n) pixels, and the block position in the frame is (i,
j), the pixel position in the block is represented by (p, q). Input L ^* , a ^* , b ^* blocks are referred to as L _ij , A _ij , and B _ij, and in order to describe the pixel position in them, for example, L _ij (p, q) Like this.

In the evaluation circuit 21, A is calculated according to the following equation.
For the pixels in each of the (a ^* block) and B (b ^* block), the power with reference to the data value 127 is calculated, and the values are taken as PA and PB.

[0041]

[Outside 4]

[0042]

[Outside 5] The value of PA + PB is compared with a predetermined threshold value, which will be described later, and if it is larger than that, it is determined that this block uses block matching by the chromaticity block, and "1" is sent via the signal line 201 of FIG. Is input to the arithmetic circuit 22. On the other hand, if it is less than or equal to the threshold value, it is determined that this block is to be subjected to block matching by the brightness block,
“0” is input to the arithmetic circuit 22 from the signal line 201 of FIG.

That is, the arithmetic circuit 22 automatically switches the block matching mode by the output of "1" or "0" input through the signal line 201.

Next, the setting of the threshold will be described in detail.

In general, the saturation of the color appearing in the natural image is not so high, and it generally appears in a half or less of the dynamic range of each of the a ^* and b ^* axes.

From this, 1 / the dynamic range
Colors of 4 or less are considered to have weak colors, and those of greater than 1/4 are judged to have strong colors. Also, if in the image area having the same chromaticity is block, all pixels within the block, the same a ^*, show a b ^* value, a portion at the edge portion of the region pixels only, the a ^*, It shows the b ^* value, and the rest is the a ^* and b ^* values of the background.

From this, there is a block
^When the power is calculated on the assumption that it is filled with the ^* and b ^* values, it seems appropriate to set about 1/2 of the power value as the threshold value. In this embodiment, the threshold value for PA + PB is 1024. About xmxn is used.

In this embodiment, the lightness data (L ^* ) and the chromaticity data (a ^* , b ^* ) are used, but the luminance color difference data (Y, C).
It is obvious that it can be realized by using r, Cb, Y, I, Q, L, U, V, etc.).

[0049]

As described above, according to the present invention, in the motion vector detecting device, the first mode in which the motion vector is detected by using the color data which is one of the color image information, and the one of the color image information are detected. By having the second mode in which the motion vector is detected using the luminance data, it is possible to detect the correct motion vector even when the brightness of the background is close to that of the target object or when the motion of the target object is large. .. Therefore, accurate estimation can be performed when estimating the motion of the entire screen from the motion vector of each block, or when estimating the motion vector of a new block from the past or surrounding motion vectors. It will be possible.

[Brief description of drawings]

FIG. 1 is a block diagram of a motion vector detection device according to an embodiment.

2 is a block diagram of a vector detection circuit used in the apparatus of FIG. 1. FIG.

[Explanation of symbols]

 1 Vector Detection Circuit 2 Frame Memory 3 Subtraction Circuit 4 Buffer Memory 5 Quantization Circuit 6 Reversible Encoding Circuit 7 Inverse Quantization Circuit 8 Addition Circuit 21 Evaluation Circuit 22 Arithmetic Circuit 24 Comparison Circuit 25 Vector Output Circuit

Claims (1)

Claim: What is claimed is: 1. A motion vector detecting apparatus for detecting a motion vector of a subject between adjacent frames in block units. A first mode for detecting a motion vector using a color component, and a luminance. And a second mode for detecting a motion vector using a component.