JPH10257485A - Detection circuit for repetitive image and image coder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To conduct excellent coding processing by detecting a repeated field in an excellent way so as to minimize visual deterioration thereby reducing redundancy sufficiently. SOLUTION: A moving/still image discrimination circuit 154 discriminates that a field is not a repeated field when an integral vector amount is more than a prescribed threshold value. In the case that the integral vector amount is not more than the threshold value, the circuit 154 discriminates the field to be a repeat field for the time being and references the presence of a motion recognition signal by a very small motion detection circuit 152. When a motion recognition signal is in existence, it is possible that a maximum level of noise is higher than an average level of the video signal and an impulse noise is in existence or the video signal is locally superimposed. Thus, the discrimination above is corrected and the field is processed not to be the repeated field. In the case that the integral vector amount is not more than a threshold value and no motion recognition signal is in existence, the field is discriminated to be the repeated field.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a repetitive image detecting circuit for detecting a repetitive image (repeat field) included in a video signal converted from a motion picture film, and an image coding apparatus.

[0002]

2. Description of the Related Art When recording video recorded as a movie film as a video signal, conversion processing is performed by a device such as telecine. For example, when converting a film moving image recorded at 24 frames / sec into a video signal of 59.94 fields / sec of the NTSC system, a 2-3 pull-down system is used. In this system, motion picture film is played at 23.976 frames / second. Then, each frame is captured and recorded while changing the number of times successively from three times to two times to three times.
It is converted into a video signal of 59.94 fields per second.

[0003] Here, the video signal of the NTSC system is composed of interlaces as is well known. Therefore, the fields corresponding to the frames recorded twice do not have the same image signal because the vertical position of each pixel of the image is different between the odd field and the even field. However, in the case of a field corresponding to a frame recorded three times, the first and last two of the three fields are used.
One field is composed of the same image (hereinafter, this is referred to as a “repeat field”).

[0004] On the other hand, when a video signal is recorded on a storage medium or transmitted by digital broadcasting, the redundancy of an image is determined by using high-efficiency coding due to restrictions on the recording capacity of the medium and the transmission band of broadcasting. Processing to reduce the degree has been performed. In this case, if the high-efficiency encoding is directly performed on the video signal obtained by converting the film material, the encoding is performed on each of the images of the two fields corresponding to the repeat fields. I will invite you.

[0005] In order to solve such a problem, in the standard of moving picture coding standardized by the International Standards Organization (ISO), a video signal is returned to a film frequency by removing a repeat field. Encoding (MPEG1) or encoding using information indicating a repeat field (MPEG2) prevents a decrease in encoding efficiency.

By the way, in order to apply such a standard coding method, it is necessary to detect a repeat field from a video signal at the time of coding. When 2-3 pull-down is performed continuously over the entire video source, the repeat field can be determined by setting the timing twice or three times first. However, when a movie film is converted to a video signal, the video signal is generally edited after being converted, subtitles and other subtitles are superimposed, and effects such as fade-in and fade-out are applied. Also, in some cases, video originally shot and created as a video signal, for example, a credit of a production company or a title telop of a movie is mixed.

[0007] In such a processed video source, pure video signals are mixed, and a repeat field does not always exist. For this reason, it becomes necessary to adaptively detect the repeat field.

As a conventional repeat field detection circuit, there is one described in Japanese Patent Application Laid-Open No. 3-250881, which has a configuration shown in FIG. In the figure, a video signal including a video signal obtained from a movie film (hereinafter, referred to as a “telecine video signal”) and a general video signal (hereinafter, referred to as a “general video signal”) are output from an inter-frame motion detection circuit 900. Is input to In the inter-frame motion detection circuit 900, the delay circuit 902 and the subtraction circuit 9
In step 04, a difference value between the same pixels in one frame of the video signal is calculated. When the video signal is a signal of the NTSC system, since the color signal is polarity-inverted and multiplexed for each frame, only the movement of the luminance signal is performed by the LPF (low-pass filter) 906 so that no error occurs in the motion detection. Separated.

The pixel difference value of the separated frame difference is
The threshold value set by the threshold value setting circuit 908 is compared with the threshold value set by the comparison circuit 910, and it is determined whether the image is a still image or a still image. The binarized comparison result is supplied to the hold circuit 912. The hold circuit 912 is controlled by the field control circuit 914 so as to be reset for each field, and the input is held for one field period. The hold value is a 5-field period detection circuit 9
16. When a telecine video signal is included in the input video signal, since a repeat field is included in a 5-field cycle, if this is detected, it is a telecine image.

On the other hand, Japanese Patent Application Laid-Open No. 4-72966 discloses that
A repeat field detection circuit as shown in FIG. 9 is disclosed. In this example, the block control circuit 918 performs a motion determination for each block, and the result is stored in the block memory 920. The 5-field period detection circuit 916 detects the periodicity for each block. Then, the determination of the telecine image is performed based on the detection result of the periodicity of the motion for each block. For example, the upper, lower, left, and right edges of the screen are places where the supermarket seems to be imposed, so blocks corresponding to these areas are excluded from the detection target, and the repeat field or telecine detection is performed using only the center block of the screen. Is performed.

[0011]

However, the above background art has the following disadvantages. (1) In consideration of noise mixed when a video signal is converted and generated from a movie film, an LPF is used for motion detection, or a threshold value equal to or higher than a noise level is set. For this reason, in the case of an image in which most of the screen is in a stationary state but a minute part is moving, the difference value due to the movement is reduced by cutting the high-frequency signal, or the difference due to the movement is reduced. When the value becomes equal to or lower than the noise level, there is a possibility that the field is erroneously determined to be a repeat field.

(2) The difference value between the frames accompanying the motion is
It also depends on the resolution of the input video signal which changes due to the characteristics of the camera and film, the image synthesis processing, the band limitation in the signal recording process, and the like. For this reason, there is a case where an image having the same motion can be determined as both a still image and a moving image.

(3) Since the noise level differs for each video source, if an appropriate noise level cannot be set, stable repeat field detection cannot be performed.

Then, the repeat field is erroneously determined,
If a moving image is erroneously encoded as a still image, temporally discontinuous motion will be reproduced during decoding, and visually significant deterioration will occur. Conversely, when a still image is erroneously encoded as a moving image, the same field is encoded twice, and the redundancy cannot be reduced sufficiently.

The present invention focuses on the above points,
Provided is a repetitive image detection circuit and an image encoding device capable of satisfactorily detecting a repeat field, minimizing visual degradation, sufficiently reducing redundancy, and performing favorable encoding processing. That is its purpose.

[0016]

In order to achieve the above object, the present invention provides a repetitive image detecting circuit for detecting an image repeated in an image signal by detecting a degree of movement between two images to be compared. Motion detection means (16)
A difference measuring means (18) for measuring a difference between the two images to be compared, a maximum difference measuring means (20) for measuring a local maximum value of the difference between the two images to be compared, Image discriminating means (22) for repeatedly discriminating an image based on the output of the means.

According to the main mode, the image discriminating means corrects the judgment result based on the detection result by the motion detecting means based on the measurement results of the difference measuring means and the maximum difference measuring means. And According to another aspect, the image discriminating means includes a minute moving part detecting means (152) for detecting a minute movement based on the measurement results of the difference measuring means and the maximum difference measuring means; While repeatedly judging an image based on the detection result, and in the case where it is judged that the image is a repetitive image, and when a minute motion is detected by the minute moving portion detecting means, it is determined that the image is a repetitive image. Determining means (154) for correcting the determination.

According to the present invention, there is provided an image coding apparatus, comprising: a signal removing unit (206) for removing a portion corresponding to a repetitive image from an image signal based on an output of the repetitive image detecting circuit described above. Encoding means (300) for performing an encoding process based on an image signal from which a portion corresponding to a repeated image has been removed.

According to the present invention, an image to be compared, for example, a motion amount between frames is detected to determine whether the image is still or not. Then, at this time, the difference between the images and the local maximum difference are measured, and the determination is corrected. Thereby, the criterion of the motion is changed according to the noise level of the image, and the erroneous detection caused by overlooking the fluctuation of the noise level or minute motion is reduced. The above and other objects, features, and advantages of the present invention will be apparent from the following detailed description and the accompanying drawings.

[0020]

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

(1) Overall Configuration FIG. 1 shows the configuration of a repeat field detection circuit according to this embodiment. As shown in the figure, the following circuits are included. One frame delay 10 that delays one frame corresponding to a video signal including a telecine video signal in a general video signal. Between an input video signal and a delayed video signal, between frames, that is, between even fields of continuous frames. Alternatively, a motion vector detection circuit 12 for calculating a motion amount between odd fields in units of a block. For an input video signal and a delayed video signal, an inter-frame calculation for obtaining a sum of absolute differences between pixels spatially at the same position between frames in units of a block. Difference circuit 14

A vector value accumulating circuit 16 for obtaining the sum of the fields of the motion amount based on the outputs of the motion vector detecting circuit 12 and the inter-frame difference circuit 14, based on the output of the inter-frame difference circuit 14, Difference value integration circuit 1 for calculating the sum
8. A block difference maximum value detection circuit 20 for obtaining a maximum value of the absolute sum of inter-frame differences in block units based on the output of the inter-frame difference circuit 14, the vector value integration circuit 16, the difference value integration circuit 18,
Based on each output of the block difference maximum value detection circuit 20,
Repeat field determination circuit 22 for determining a repeat field

(2) One Frame Delay 10 Each circuit will be described below in order. First, the one-frame delay 10 will be described. For example, in the case of the NTSC system, if one horizontal scanning cycle is 1H, one frame is 52
5H. Therefore, for the input video signal, 5
By giving a delay of 25H, a delay equivalent to one frame can be obtained.

(3) Motion vector detection circuit 12 FIG. 2 shows an example, in which a block forming circuit 120, a block matching circuit 122, an accumulator circuit 12
4, a difference value comparison circuit 126 and a register 128. The input video signal and the video signal delayed by the one-frame delay 10 are input to the blocking circuit 120. The blocking circuit 120 cuts out each video signal image into two-dimensional blocks. The size of the two-dimensional block is, for example, N × M (N and M are natural numbers) for an input video signal, and (N + X) × (M + Y) for a delayed video signal.
(X and Y indicate the range in which the motion vector of each dimension is searched). For example, a two-dimensional block of about 16 × 16 pixels is used.

In the cut-out block, the matching start position of the delayed video signal moves for each value of the vector to be searched.
The absolute difference value between each pixel of the M block is obtained. The absolute difference value obtained for each pixel is integrated by the accumulator circuit 124. When the integration is completed for the N × M block sections, the integrated value (hereinafter referred to as “MAD”)
Is stored in the register 128. The MAD stored in the register 128 is compared with the minimum value of the MAD for the vector already calculated in the difference value comparison circuit 126. If the MAD is smaller than that, the vector is selected as the most correct vector. And the vector value register 12
8A and the MAD register 128B store the vector value and MAD
Are respectively stored.

When the MAD calculation and magnitude comparison for all the searched vectors are completed, the vector value finally selected and stored in the register 128 and its MAD are output to the vector value integrating circuit 16.

(3) Inter-frame difference circuit 14 As shown in FIG. 3, the inter-frame difference circuit 14 is composed of a blocking circuit 130, a block matching circuit 132, and an accumulator circuit 134. The configuration is almost the same as that of the vector detection circuit 12. However, in the case of the inter-frame difference circuit 14, N × M blocks that are spatially at the same position in the delayed video signal are cut out by the blocking circuit 130 from the N × M blocks of the input video signal. Thereafter, the block matching circuit 132 and the accumulator circuit 134 having the same function as the motion vector detection circuit 12 determine the N × M block MAD. Required MAD
Are output to the vector value integration circuit 16, the difference value integration circuit 18, and the block difference maximum value detection circuit 20, respectively.

(4) Vector Value Integration Circuit 16 As shown in FIG. 4, the vector value integration circuit 16 includes a block motion determination circuit 140 and a vector amount integration circuit 142. The block motion determination circuit 140 includes a MAD value “MV · MAD” for the motion vector output from the motion vector detection circuit 12 and an inter-frame difference MAD value “NOMV · MAD” output from the inter-frame difference circuit 14.
Is entered. These integrated values MV / MAD and NOMV / MAD are
The data is input to the block motion determination circuit 140 in synchronization with each of the N × M blocks extracted by the blocking circuits 120 and 130.

The block motion determination circuit 140 determines the motion of the block under the following conditions.
That is, when MV · MAD + α ≧ NOMV · MAD, static judgment is performed. When MV · MAD + α <NOMV · MAD, dynamic judgment is performed. May be determined. in this way,
In the block motion determination circuit 140, motion determination is performed in consideration of an average noise level. The determination result is output to the vector amount integrating circuit 142 as a vector valid flag.

The motion amount judgment value determined for each block by the above judgment and the motion vector value for the block supplied from the motion vector detection circuit 12 are input to the vector amount accumulation circuit 142. Vector quantity integration circuit 1
At 42, the vector amount for one field is integrated using these inputs. The “vector value” is a value in the horizontal and vertical directions on a two-dimensional image, and the “vector amount” is a value of a distance (vector length) on the two-dimensional image.

The vector amount integrating circuit 142 is provided with a register for integration. In the case of a stillness determination, the integration process is not performed, and only the block of the motion determination is performed. The vector amount to be integrated is calculated from the vector value. As an example, assuming that the horizontal direction of the vector value is "x" and the vertical direction is "y", the vector amount L is calculated by the following equation. L = | x | + (β × | y |) In this equation, | a | indicates the absolute value of a. “Β” is a coefficient determined based on the pixel ratio in the horizontal and vertical directions. In this way, the vector amount integrated for the motion determination block for one field is output to the repeat field determination circuit 22. The integrated vector amount is the sum of the vector amounts of the motion determination blocks in the field screen, and indicates the magnitude of the motion of the image between frames, and serves as a reference for determining the motion between frames.

(5) Difference Value Integration Circuit 18 The difference value integration circuit 18 is a circuit that integrates the value of the inter-frame MAD for each block received from the inter-frame difference circuit 14 over one field. Difference value integration circuit 18
Is provided with a register for integration as in the case of the vector value integration circuit 16, and the MAD of all blocks is integrated. Then, the integrated value is output to the repeat field determination circuit 22.
Is output to Note that the difference integrated value is the sum of the difference values of spatially identical locations of images between frames, and indicates a noise level when the screen is assumed to be stationary.

(6) Block difference maximum value detection circuit 20 The block difference maximum value detection circuit 20 is a circuit that searches the maximum value of the inter-frame MAD for each block received from the inter-frame difference circuit 14 over one field. is there. The block difference maximum value detection circuit 20 is provided with a register for storing the maximum value of the inter-frame MAD, and the MAD input for each block is compared with the maximum value so far and written into the register. By repeating this process sequentially, the maximum value in one field is finally obtained.
Note that the maximum difference value is the maximum value of the total sum of differences for each block in the entire screen, and indicates a local peak level of noise when the screen is assumed to be stationary.

(7) Repeat Field Judgment Circuit 22 The repeat field judgment circuit 22 is composed of a block MAD threshold value setting circuit 150, a small moving part detection circuit 152, and a movement judgment circuit 154, as shown in FIG.
The repeat field determination circuit 22 includes a vector amount integration value for one field of the vector value integration circuit 16, an inter-frame difference integration value of the difference value integration circuit 18,
The block MAD maximum value of the block difference maximum value detection circuit 20 is input. Then, based on these inputs, it is determined whether or not it is a repeat field.

First, the inter-frame difference integrated value of the above-described difference value integrating circuit 18 is calculated by the block MAD threshold value setting circuit 150.
Supplied to In the block MAD threshold setting circuit 150,
A threshold value for determining that the movement is slight is calculated based on the inter-frame difference integrated value. The calculation is performed based on the following equation, for example. Assuming that the inter-frame difference integrated value is {MAD} and the number of blocks in one field is B, the threshold value TH is represented by TH = {({MAD × γ) / B}. Here, “γ” needs to be a number of 1 or more, and generally a value of about 3 to 10 is used.

Next, the threshold value T obtained as described above is obtained.
H is input to the minute movement detecting circuit 152. The minute movement detecting circuit 152 includes a block difference maximum value detecting circuit 20.
The maximum value of the block MAD is also input. The threshold value TH is
This block MAD is compared with the maximum value. And the threshold T
If the block MAD maximum value is larger than H, it is determined that a motion exists. That is, the “block MAD maximum value” is considered to be significantly larger than the “block MAD average expressed as the average value of the difference due to the motion level and the motion of the entire field screen”, and the block in which the difference due to the motion rather than noise appears Is determined to exist.

In other words, the threshold value TH corresponds to an average noise level of the entire screen. Also block MA
The D maximum corresponds to the local noise maximum level. If the image is stationary, it is considered that the maximum level of the noise does not deviate from the average level. Therefore, when the maximum level exceeds the average level, it can be determined that the image is a moving image. This is output from the minute moving portion detection circuit 152 as a motion recognition signal, and sent to the motion judgment circuit 154.

The motion determination circuit 154 also receives the one-field integrated vector amount of the vector value integrating circuit 16. The motion judgment circuit 154 determines the presence or absence of a final repeat field based on the one-field integrated vector amount and the motion recognition signal. This determination is performed, for example, in the following procedure.

First, if the integrated vector amount is equal to or larger than a predetermined threshold value ε, it is determined that there is a motion, in other words, the field of the input video signal is not a repeat field. The value of “ε” may be an arbitrary number, but a value considerably larger than 0, for example, about 10 to 50 is appropriate in consideration of variation in motion due to a noise level.

If the integrated vector amount is equal to or smaller than the threshold value ε, it is determined that the field is a repeat field, and the presence or absence of a motion recognition signal is referred to. If a motion recognition signal has been sent, it is determined that a minute motion has been detected, and the above-described judgment is corrected, and the motion is treated as a field having motion. The presence of the motion recognition signal means that the maximum level of the noise is higher than the average level as described above. Such cases may include the presence of impulsive noise or the local imposition of a video signal.

If the integrated vector amount is equal to or less than the threshold value ε and no motion recognition signal is sent, it is determined that the field is a repeat field. The above determination result is sent to an image encoding device (not shown) as a repeat field signal, and the corresponding field is encoded according to this signal.

FIG. 6 is a graph showing an area determined as a repeat field by the motion determination circuit 154 in this way. In the figure, the horizontal axis is the block MAD maximum value indicating the local peak level of noise, and the vertical axis is the one-field integrated vector amount indicating the degree of motion. If the one-field integrated vector amount is equal to or larger than the threshold value ε, it is not determined that the field is a repeat field. Even if the maximum value of the block MAD is equal to or greater than the threshold value TH, even if the maximum value is equal to or greater than the threshold value TH, there is a possibility of superimposition or the like. Therefore, what is determined as a repeat field is a hatched area in the figure. Note that the threshold value TH corresponding to the average noise level of the entire screen fluctuates according to the magnitude of the inter-frame difference integrated value, that is, the degree of noise as described above (see the arrow in the figure).

As described above, according to this embodiment, when the vector amount is clearly larger than the noise level, it is determined to be a moving image. For this reason, the motion recognition signal is effective only when the still area occupies most of the entire screen, and it is possible to recognize the motion of a minute portion that deviates from the noise level of the entire screen such as superimpose. Therefore, visual deterioration due to erroneous determination of the repeat field is favorably reduced.

The image encoding process using the output of the repeat field determination circuit 22 as described above includes, for example, a "frame number inverse conversion device" described in Japanese Patent Laid-Open No. Hei 6-233182. The method of obtaining the edit point and the pattern of the frame number reverse conversion from the order of appearance, and omitting the image of the same field and coding can be applied as it is. In the case of MPEG, in MPEG1, a video image of 59.94 fields is converted into an image of 23.976 frames from the inverse conversion pattern and is encoded in a non-interlace manner. In MPEG2, information indicating a repeat field is added from the inverse conversion pattern, and encoding is performed without the same field.

FIG. 7 shows an example of an image coding apparatus including the repeat field detection circuit and the MPEG coding apparatus of the above embodiment. In the figure, a video signal is supplied to a repeat field detection circuit 200 and a frame memory 202, respectively. The output of the repeat field detection circuit 200 is supplied to a pattern detection circuit 204, where a pattern corresponding to the repeat field is detected. Outputs of the frame memory 202 and the pattern detection circuit 204 are supplied to a video signal capture circuit 206. In the video signal capture circuit 206, the video signal excluding the video signal corresponding to the pattern detected by the pattern detection circuit 204 and the video signal excluding the portion corresponding to the repeat field are stored in the frame memory 20.
2 from the output signal. As a result, the repeat field is removed. And the remaining video signal
From the video signal capture circuit 206 to the image encoding device 3
00 is supplied.

The image coding apparatus 300 includes a subtractor 302,
DCT circuit 304, quantizer 306, variable length encoder 3
08, inverse quantization / IDCT circuit 310, adder 312,
It includes a frame memory 314 and a motion compensation prediction circuit 316, and has a well-known configuration. This allows
Encoding processing according to the MPEG standard is performed, and encoded data is obtained.

As described above, 2 from movie film
−3 For a video source containing a video signal such as telecine which has been pulled down, the difference in resolution of the input image,
Good coding processing can be performed by reducing errors in repeat field detection due to fluctuations in noise level, minute movements, etc., minimizing visual degradation and sufficiently reducing redundancy.

Although the motion vector detecting circuit 12 is used in the above embodiment, a configuration without using the motion vector detecting circuit 12 may be adopted.
That is, the inter-frame difference circuit 14, the difference value integration circuit 1
8. The repeat field may be determined using the block difference maximum value detection circuit 20 and the repeat field determination circuit 22. In this case, the minute moving portion detection circuit 152
The repeat field is determined based on the output of. According to this, erroneous determination due to a difference in noise and resolution due to not obtaining a motion vector slightly increases,
An erroneous determination due to the movement of a minute portion can be prevented in the same manner as in the above embodiment, and there is an advantage that the circuit configuration can be simplified as compared with the above embodiment.

As a preferred application example of the present invention, there is a frame frequency conversion apparatus which performs good repeat field detection on a video source and adaptively converts a frame frequency.

[0050]

As described above, according to the present invention,
Reduces false detections due to differences in resolution of input images, fluctuations in noise levels, and slight movements, and enables repeated images to be detected satisfactorily, minimizing visual degradation and sufficiently reducing redundancy. Thus, there is an effect that it is possible to perform a good encoding process.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a repeat field detection circuit according to an embodiment.

FIG. 2 is a block diagram showing a configuration of a motion vector detection circuit of the embodiment.

FIG. 3 is a block diagram showing a configuration of an inter-frame difference circuit of the embodiment.

FIG. 4 is a block diagram showing a configuration of a vector value integration circuit of the embodiment.

FIG. 5 is a block diagram showing a configuration of a repeat field determination circuit of the embodiment.

FIG. 6 is a graph showing a repeat field determination method of the embodiment.

FIG. 7 is a block diagram showing a configuration of an image encoding device using the repeat field detection circuit of the embodiment.

FIG. 8 is a block diagram showing a configuration of a conventional repeat field detection circuit.

FIG. 9 is a block diagram showing a configuration of a conventional repeat field detection circuit.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 10 ... 1 frame delay 12 ... Motion vector detection circuit 14 ... Inter-frame difference circuit 16 ... Vector value integration circuit 18 ... Difference value integration circuit 20 ... Block difference maximum value detection circuit 22 ... Repeated field judgment circuit 152 ... Small motion part detection circuit 154: motion determination circuit 206: video signal capture circuit 300: image encoding device

Claims (4)

[Claims] 1. A repetitive image detecting circuit for detecting an image repeated in an image signal, a motion detecting means for detecting a degree of motion between two images to be compared, and a difference between the two images to be compared. A difference measuring unit for measuring, a maximum difference measuring unit for measuring a local maximum value of a difference between two images to be compared, an image determining unit for repeatedly determining an image based on an output of each of these units, A repetitive image detection circuit comprising: 2. The apparatus according to claim 1, wherein said image discriminating means corrects a judgment result based on a detection result obtained by said motion detecting means based on the measurement results of said difference measuring means and said maximum difference measuring means. A repetitive image detection circuit as described. 3. The image discriminating means includes: a fine moving part detecting means for detecting a minute movement based on the measurement results of the difference measuring means and the maximum difference measuring means; and a detection result of the motion detecting means. In the case where a repeated image is determined, and it is determined that the image is a repeated image, and a small movement is detected by the small moving portion detecting means, the determination that the repeated image is determined is corrected. 3. The repetitive image detection circuit according to claim 2, further comprising: 4. A signal removing means for removing a portion corresponding to a repetitive image from an image signal based on an output of the repetitive image detection circuit according to any one of claims 1, 2 and 3, and thereby a repetitive image An encoding unit for performing an encoding process based on the image signal from which a corresponding portion has been removed, the image encoding device comprising: