JPS6135756B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a motion detection device that compares video signals of previous and current frames and detects movement of each point in an image from the luminance difference. Here, the term "frame" refers to either one screen in progressive scanning, one frame in interlaced scanning, for example, one of the even or odd fields of a standard television system, or a frame in which both fields are combined.

The cause of the brightness difference between two frames is
There are two causes: video movement and system noise. In general, changes in brightness due to noise are smaller than changes in brightness due to video, so a threshold hold level is set to distinguish between noise and video movement, and when the change in video brightness is within this range, it is determined to be noise. When it exceeds , it can be determined that there is movement in the video.

Therefore, using this motion detection device, the coefficient K _i is calculated from the luminance difference, and the video signal A _i of the previous frame and the video signal B _i of the current frame are calculated.
From this, K _i A _i + (1-K _i ) B _i is calculated, and the result of the calculation is output as a video signal at that point, thereby reducing the noise component and smoothing the motion of the video. A removal device has been proposed.

This noise removal device sets the threshold level to, for example, 1.1% of the maximum luminance, and when the luminance change between frames is within this range, sets K _i to, for example, "1/2", and the obtained 1/2 ( A _i + B _i ) is output as the video signal of the current frame at that point, and when it exceeds 1.1%, it is determined that the image has moved, and the input video signal is output as is. It is.

In this way, when the change in the video signal is below the threshold level, it is judged as noise and
Since the previous frame and the current frame are weighted and added, noise components are reduced.

However, with this device, the threshold hold level is set to a constant value regardless of the brightness, so
When improving the S/N ratio in the low-brightness area of a video signal, the movement of the image in minute parts becomes unnatural, or the minute parts of the image (for example, hair, hair, etc.)
The disadvantage is that the grass (grass, etc.) is crushed.

This point will be explained with reference to FIG. Figure 1 shows a state in which the image moves and the brightness changes by 10% between frames t _i and t _{i +1} at two brightness levels: maximum brightness "1" and low brightness "0.1". It is. All of these are brightness changes due to the movement of the image, but at the maximum brightness of "1", the brightness change |Δ| is |Δ|=1.00−0.90≫0.011, which is considerably larger than the discrimination level, so it is due to the movement of the image. It is judged as.

However, in a low brightness area, |Δ|=0.1(1.00−0.90)<0.011, which means that it is determined as noise and is added to the previous frame. As a result, the movement of minute images, such as hair, becomes unnatural. In addition, a collection of low-brightness minute parts, such as hair, is judged as not moving even if the whole part moves slightly, and is added together, so the overall brightness is averaged and the image becomes distorted. .

SUMMARY OF THE INVENTION In view of the above drawbacks, it is an object of the present invention to provide a motion detection device that can eliminate discrimination errors in low brightness areas.

By expanding the low brightness level, the device of the present invention greatly amplifies the brightness in the low brightness area, increasing the absolute value of the difference, and reduces it in the high brightness area, thereby effectively reducing the threshold. This is characterized in that the width of the hold level is changed depending on the brightness. As a result, the width of the threshold hold level becomes narrower in low brightness areas,
Even a slight change in brightness can be determined as motion in the image. Therefore, the image of minute parts will not be distorted or the movement of these parts will not be erased.

Here, when obtaining a video signal from a movie film, we will consider the movement of minute parts such as grass and noise due to the graininess of the video film.

Let the brightness of an arbitrary point in the previous frame be S, and let the brightness of the same point in the current frame be S+ΔS.
In the conventional device, since the absolute value of the difference |Δ| is directly calculated, |Δ| becomes ΔS. In scenes where minute changes are modulated on a relatively sticky and flat screen, such as grass, distant greenery, hair, etc., if the average brightness is S and the modulation rate is ΔM, then ΔS=
It can be defined as SΔM. Using this definition, |Δ| is expressed as follows.

|Δ|=SΔM (1) Further, in movie film, noise occurs due to graininess. Graininess is expressed by dispersion (RMS), or σ _D , and the peak value of the RMS value of a typical film has concentration dependence as shown in Figure 2.
Set to 0.03. The graininess between two frames is σ′
Then, σ′=√2σ _D because sampling is from independent populations with equal variance. Therefore, S=10 ^-D = T S+ΔS=10 ^-D+ 〓'=T・10〓'. Here, D is the density and T is the transmittance.

Therefore, the brightness change due to graininess can be expressed as follows.

|Δ|=T(1-10〓')≒T/0.43σ' (2) When the above equations (1) and (2) are plotted, curves A and B shown in FIG. 3 are obtained. Here, curve A represents motion, and curve B represents grain. In this Figure 3, we assume that the motion detection limit is 5% (Webner's law), that is, ΔM = 0.05, and the S/N ratio is
The target is around 46dB (200/1). If the threshold hold level TH is 0.005, then TH=ΔS=S・ΔM=S×0.05=0.005.

Therefore, S=0.1, that is, density D=1.0. As a result, conventional devices cannot detect motion when the concentration exceeds 1.0.
The screen will have small details crushed.

On the other hand, in the present invention, the low luminance area is expanded. Therefore, for example, in the case of logarithmic transformation, |Δ|=logS+ΔS/S =0.43ΔS/S.

In detailed scenes such as lawns, as mentioned above, Since it can be defined as ΔS=S ΔM, |Δ|=0.43 ΔM…(3) becomes.

Also, the brightness changes due to the graininess of movie film are |Δ|=D−(D+σ′)=σ′…(4) becomes.

When these equations (3) and (4) are plotted, curves A' and B' shown in FIG. 4 are obtained. As can be seen from FIG. 4, when logarithmic transformation is performed, it becomes constant regardless of brightness in a fine scene. Therefore, if the threshold level TH is set to TH=0.43 ΔM=0.43×0.05≈0.02, motion in all images can be detected regardless of brightness.

Furthermore, the graininess of the motion picture film is considerably improved since the proportion of the shaded area is larger than that of the conventional apparatus shown in FIG.

Embodiments of the present invention will be described in detail below with reference to the drawings.

In FIG. 5, a video signal (picture signal) sent from a television camera or telecine projection device is converted into a digital signal for each point on a horizontal scanning line by an A/D converter 10, and is calculated by an arithmetic circuit 11. Ru. The result of this calculation is stored in the frame storage device 12, converted into an analog signal by the D/A converter 13, and output as a video signal for each point.

The frame storage circuit 12 stores a video signal A _i at each point of one frame. This frame storage circuit 12 includes a write/read control circuit 14
The video signal of the previous frame stored at a predetermined address is read out, and the video signal of the current frame calculated by the calculation circuit 11 is written.

A clock signal generating circuit 15 generates a predetermined number of clock signals based on the horizontal synchronization signal of the video signal.
This clock signal is counted by the address designation counter 16, and its output signal is sent to the write/read control circuit 14 to designate the address of the frame storage circuit 12.

Therefore, when the video signal of the current frame is sent, this video signal is converted into a digital signal B _i by the A/D converter 10. In synchronization with the video signal B _i of each point of this frame, the video signal A _i of the previous frame of the same point (address) stored in the frame storage circuit 12 is read out. The read video signal A _i is input to a first black expansion nonlinear conversion table 17 that expands the low luminance level portion. As this black expansion nonlinear conversion table 17, a logarithmic conversion table or the like is used. FIG. 6 is a graph showing this output characteristic.

This first black expansion nonlinear conversion table 17 is composed of an address conversion decoder and a P-ROM or RAM, and the address of the P-ROM is specified according to the input level, and the data written to this address is read out. .

Similarly, the level of the low luminance area of the video signal B _i of the current frame, which has been digitally converted by the A/D converter, is increased by the second black expansion nonlinear conversion table 18.

The video signal A _i ' of the previous frame whose level has been converted in this way and the video signal B _i ' of the current frame being input are input to the absolute value calculation circuit 19,
The absolute value of the difference |Δ _i | is calculated. In this way, the absolute value |Δ _i | is calculated for each point and input to the coefficient generation circuit 20.

The coefficient generation circuit 20 quantizes the absolute value |Δ _i | into at least two or more, and outputs a signal corresponding to the coefficient K _i from “0 to 1”. This coefficient K _i
can be set to any value by operating the changeover switch from the outside.

In Figure 7, a certain judgment level a is used as a standard, and when it exceeds this level, it is judged that the image at that point has changed, and the coefficient K is set to ``0'', and when it is within this level, the coefficient K is set to ``1/2''. An example of the coefficient generator 20 is shown.

It can be easily inferred that in order to smooth the motion of the image, it is desirable that the coefficient K be multi-level or continuous. FIG. 8 shows an embodiment in which the coefficient K is continuous.

The signal of the coefficient K _i set by comparing the luminance signals A _i and B _i of the previous and current frames at the same point is sent to the arithmetic circuit 11, where K _i A _i +(1− K _i )B _i is calculated.

If the coefficient K is set to 0 when the image is moving and 1/2 when it is stationary as shown in Figure 7, it becomes 1/2 (A _i + B _i ), which is the same as the previous value. The video signals A _{i and} B _i of the frame and the current frame are added, and when the frame is stationary, B _i is obtained, and the video signal A _i of the previous frame is cleared.

In this way, the above calculation is performed for each point of the frame, and the calculation result is sent to the D/A converter 13 as a video signal of the current frame, and is converted back to an analog signal and output. At the same time, it is stored at a predetermined address in the frame storage circuit 12 and used for calculation of the next frame.

As described above, since the video signals of the previous frame and the current frame are weighted and added, the rate of change of noise, which has a small change in brightness and is included within the discrimination level, becomes small. Therefore, there is an effect of suppressing noise components to a small level. If the brightness change is large, it is determined that there is a change in the video, and the input video signal B
Since _i occupies a larger proportion, it does not interfere with the movement of the image.

FIG. 9 shows a noise removal device for color television. For color TV, the first
and second comb-shaped films 21 and 22 are provided to extract only the luminance signals A _Yi and B _Yi from the video signal A _i of the previous frame and the video signal B _i of the current frame, and then perform black extension nonlinear conversion. and compare them. This is because a 3.5 MHz subcarrier color signal is superimposed on the video signal, so if logarithmic conversion is performed as it is, a logarithmic conversion error will result.

The first comb filter 21 is provided with a chroma inverter, and the chroma signal is phase-converted. A video signal consisting of the phase-converted chroma signal and luminance signal is sent to the arithmetic circuit 11. Since noise due to the graininess of movie film is concentrated in the low frequency range, providing a low-pass filter can further improve the graininess.

FIG. 10 shows an embodiment in which first and second low-pass filters 23 and 24 are inserted to remove low frequency components of the video signal, and then black extension nonlinear conversion is performed.

Since the present invention having the above configuration greatly amplifies the video signal in the low brightness area, it is possible to obtain an effect equivalent to narrowing the width of the threshold level in the low brightness area. There are no mistakes, and there is no possibility that minute parts of the image will be distorted or movements will become unnatural. Furthermore, noise caused by the graininess of movie film can also be improved.

[Brief explanation of the drawing]

Fig. 1 is a graph showing inter-frame brightness changes in high-brightness and low-brightness areas, Fig. 2 is a graph showing graininess of movie film, and Fig. 3 is a graph showing noise and threshold levels of conventional equipment. FIG. 4 is a graph showing the noise and threshold level of the device of the present invention, FIG. 5 is a block diagram showing the noise removal device for black and white television of the present invention, FIG. 8 and 8 are characteristic diagrams of the coefficient generation circuit, FIG. 9 is a block diagram showing a noise removal device for color television, and FIG.
FIG. 0 is a block diagram showing an embodiment provided with a low-pass filter.

Claims (1)

[Claims] 1. The video signal of each point of the previous frame stored in the storage means is read out and compared with the video signal of each point of the current frame being input, and the image of each point is determined based on the luminance difference. A motion detection device for detecting a motion of
A motion detection device characterized in that video signals of the previous and current frames, which have lower luminance levels, are expanded and then the two are compared. 2. The motion detection device according to claim 1, wherein the black extension nonlinear conversion means is a logarithmic conversion table in which an output signal is set in advance for an input signal.