JPH0738689B2 - Noise reduction device - Google Patents

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a noise reduction device that reduces noise in a digital image signal by intra-frame processing using a low-pass filter.

[Conventional technology]

Conventionally, noise reduction devices that reduce noise in digital image signals reduce noise by inter-frame processing using a low-pass filter and noise by intra-frame processing using a low-pass filter. It is roughly divided into those that reduce.

By the way, when the noise is reduced by the inter-frame processing, the noise is reduced by the time averaging processing using the inter-frame correlation of the image signal, and therefore the effective noise reduction is applied to the image signal of the still image portion. However, with respect to the image signal of the moving image portion, the level change portion where the level changes more than the reference level, that is, the outline portion (hereinafter referred to as the edge portion) is apt to be blurred, and the circuit scale is required because the frame memory is required. Grows larger.

Japanese Patent Application Laid-Open No. 55-37018 (hereinafter referred to as "first publication") discloses a device for reducing noise by inter-frame processing, in which a motion vector detection circuit and a motion vector correction circuit It is described that the motion is detected and the motion is corrected to suppress the blur of the edge portion and reduce the noise.

However, in the case of the device described in the first publication, the motion is detected based on the difference level of the image signals between the frames, so that the motion correction is performed especially when the difference level becomes large due to noise. Therefore, it is not possible to perform good noise reduction, and in addition to the frame memory, a motion vector detection circuit, a motion vector correction circuit, etc. are required, and the circuit scale becomes large.

Further, Japanese Patent Laid-Open Publication No. 57-193180 (hereinafter referred to as "second publication") discloses in order to solve the above-mentioned inconvenience of the first publication.
It is described that a motion-corrected prediction screen signal that is corrected according to the motion of an input screen signal (image signal) and a non-motion-corrected prediction screen signal that is not corrected are used for noise reduction. In this case, however, the circuit scale is further increased because a switching determination circuit for switching both prediction screen signals is additionally formed in the first publication.

That is, when noise is reduced by inter-frame processing, at least a frame memory is required, the circuit scale is increased, and when noise is reduced by suppressing blurring at the edge portion, the noise reduction method described in the first and second publications is used. Thus, the circuit scale becomes larger.

On the other hand, when noise is reduced by the intra-frame processing, a line memory or the like may be used instead of the frame memory. Therefore, the circuit scale becomes smaller than that when the inter-frame processing reduces the noise, and Since the processing is performed, the blurring of the edge portion has an advantage over the case where the blurring is reduced by the interframe processing.

Then, in the conventional noise reduction device that reduces noise by the intra-frame processing, even if the noise reduction effect on the edge portion is lower than the noise reduction effect on the flat portion, the noise in the edge portion is reduced based on human visual characteristics. Focusing on the inconspicuousness, there is a method of switching the band characteristic of the low-pass filter between the edge portion and the flat portion where the level does not change in order to suppress the blurring of the edge portion and reduce the noise.

However, in order to switch the band characteristic, for example, a non-linear circuit element is required in the low-pass filter, which causes a problem that the circuit scale becomes large.

Therefore, in Japanese Patent Laid-Open No. 57-193180 (hereinafter referred to as "the third publication"), a two-dimensional low-pass filter having a band characteristic that reduces noise in a flat portion is used, and a component removed by the filter is used. It is described that the signal of is added to the output signal of the filter at the time of detecting the edge portion to suppress the blur of the edge portion and reduce the noise. In this case, the low-pass filter having the predetermined band characteristic is reduced to 1 Since only one is used, the circuit scale does not increase.

[Problems to be solved by the invention]

By the way, in the case of the third publication, an image signal is input to a two-dimensional low-pass filter and horizontal and vertical edge detection units, and the current pixel being input is framed by using the input pixel and neighboring pixels. Since the filter processing is performed inside the filter, regardless of the edge detection, as described below, the output signal of the filter is susceptible to blurring due to the influence of neighboring pixels on the pixels of the edge portion, and blur suppression is effective. However, there is a problem that the image quality is deteriorated.

For example, in the case where the filter processing is performed by using the pixels of three lines by the low-pass filter, the pixels of the double circle mark α of the n-th line ln of FIGS. 9A and 9B are displayed on the filter. It is assumed that α'is input.

By the way, FIGS. 9 (a) and 9 (b) show a digital image signal formed by the process of the sub-sampling method, in which the non-shaded areas indicate the high-level screen portion of the white background, and the shaded areas indicate the black background. Shows the low-level screen portion of.

Further, circled pixels β, γ, δ and σ in FIG. 9A indicate neighboring pixels when the pixel α is filtered, and circled pixels β ′ and γ ′ in FIG. 9B. , Δ ′, σ ′ indicate neighboring pixels when filtering the pixel α ′, and the pixel β,
γ, β ′, γ ′ are located on the n−1th line ln ₋₁ , and pixels δ, σ, δ ′, σ ′ are located on the n + 1th line ln ₊₁ .

Therefore, the pixels α and α'become pixels in the edge portion, respectively.

Then, the pixel α is filtered using five pixels of the pixel α and the neighboring pixels β, γ, δ, σ, and the pixel α ′ is processed by the pixel α ′ and the neighboring pixels β ′, γ ′, δ ′, σ. 5'and 5'are used for filtering.

By the way, while the pixels β and σ are the level of the white background to which the pixel α belongs, the pixels γ and δ are the levels of the black background, and similarly, the pixels β ′ and δ ′ are the levels of the white background to which the pixel α ′ belongs. , Pixels γ ′, σ ′ are black background levels.

Therefore, the pixels α and α ′ after the filter processing are the pixels δ and γ of the black level regardless of the detection of the edge portion.
The level is lowered under the influence of γ'and σ ', which causes blurring in the edge part, and even if the removed component is added to the output signal of the filter, sufficient suppression cannot be performed, and the image quality deteriorates. May lead to

[Means for solving problems]

The present invention has been made with the above points in mind,
In a noise reduction device that reduces noise in a digital image signal by performing in-frame filter processing using a low-pass filter, an input image signal is calculated for each pixel of the input image signal from a level difference between pixels in a two-dimensional neighborhood of the pixel. Level change detecting means for detecting a level change in the horizontal and vertical directions of the frame screen formed by the above, and at least one of the horizontal and vertical directions by the detecting means.
Edge detection means that outputs an edge detection signal assuming that the edge portion of the image is detected each time a level change in the direction is detected, and the reference level for detection by the level change detection means each time the edge detection signal is output. Identification means for identifying, as a threshold value, a pixel of a level opposite to the level to which the detection pixel belongs among the pixels in the two-dimensional vicinity of the detection pixel; and a pixel of the opposite level of each pixel in the two-dimensional vicinity of the detection pixel. When the level change is not detected by the replacement processing means for replacing any of the detected pixels in the level to which the detected pixel belongs with each pixel in the two-dimensional neighborhood of the detected pixel into the pixel of the level to which the detected pixel belongs The input image signal is subjected to filter processing, and when a level change is detected, the replacement processing means performs filter processing on the image signal after the replacement processing. A noise reducing apparatus characterized in that a pass filter.

[Work]

Therefore, when the above-mentioned edge portion is detected by the detection of the level change of the level change detection means, the replacement processing means detects the detected pixel by the identification result of the identification means based on the edge detection signal of the edge detection means. Of the pixels in the two-dimensional neighborhood, the pixel of the level opposite to the level to which the detected pixel belongs,
The pixels in the two-dimensional neighborhood are replaced with any of the pixels at the level to which the detected pixel belongs, and the pixels in the two-dimensional neighborhood are made into pixels at the level to which the detected pixel belongs.

Then, the image signal after the replacement processing by the replacement processing means is supplied to the two-dimensional low-pass filter instead of the input image signal, and the noise reduction filter processing is performed.

At this time, the image signal formed by the replacement becomes a signal in which the pixels δ and γ in FIG. 9A are replaced with the pixels β and σ, respectively, so that the characteristics of the low-pass filter are flattened. Based on the intra-frame processing, by setting the predetermined band characteristic that sufficiently reduces the noise of the part,
Noise can be reduced by suppressing blurring at the edges as much as possible.

〔Example〕

Next, the present invention will be described in detail with reference to FIGS. 1 to 8 showing one embodiment thereof.

FIG. 1 shows the overall configuration of the apparatus when applied to a digital image signal formed by sub-sampling processing. In FIG. 1, (1) shows that each pixel of the digital image signal is sequentially input. The image input terminal (2) is a neighboring pixel extraction unit connected to the input terminal (1), delays the input image signal of the input terminal (1), and
The current pixels corresponding to the pixels α and α ′ in (b) and the neighboring pixels β, γ, δ, σ, β ′, in (a) and (b) of FIG.
A signal of 5 pixels with four nearest two-dimensional neighboring pixels corresponding to γ ′, δ ′, σ ′ is output each time one pixel is input to the input terminal (1).

(3) is an edge detection section connected to the extraction section (2), which is usually composed of an isolated point noise removal block (4) and a detection block (5), and is input based on the input signal of 5 pixels. An edge detection signal is output each time a level change in the horizontal and vertical directions of the frame screen formed by the image signal is detected and the level change is detected in at least one of the horizontal and vertical directions.

(6) is an input side switching device which is switched according to the edge detection signal of the extraction section (2), and outputs the output signal of the extraction section (2) to the flat side contact (a) when the edge detection signal is not output. Then, when the edge detection signal is output, the output signal of the extraction unit (2) is output to the edge contact (b).

(7) is an edge processing unit to which the signal of the contact (b) of the switch (6) is input, which is composed of a replacement control circuit (8) and a replacement circuit (9), Of these signals, the signals of four neighboring pixels are replaced as described later.
Reference numeral (10) is an output side switching device which is interlocked with the switching device (6) in accordance with the edge detection signal of the extraction unit (2), and the contact (a) 'of the flat side contact (a)' While the signal of a) is input, the processing unit (7) is connected to the edge contact (b) '.
The output signal is input to switch the contacts (a) 'and (b)' when the edge detection signal is not output and when the edge detection signal is output.

(11) is a two-dimensional low-pass filter to which the output signal of the switching device (10) is input, and is set to a predetermined band characteristic that surely reduces noise in the flat portion of the input image signal. The filter processed signal is sequentially output pixel by pixel to the image output terminal (12).

The extraction unit (2) and the detection unit (3) form level change detection means and edge detection means, and the switching units (6) and (10) and the processing unit (7) identify and replace them. Processing means are formed.

Next, the detailed configuration and operation of each unit in FIG. 1 will be described.

First, the extraction unit (2) is configured as shown in FIG. 2, and in the same drawing, (2a) and (2b) are two line memories connected in cascade to the input terminal (1), respectively The signal of the pixel one line before the selected pixel is output. (2c),
(2d) is two 1s connected in series to the line memory (2b)
It is a clock delay device and outputs the signal of the pixel one clock before the input pixel. Reference numerals (2e) and (2f) are two one-clock delay units connected in series to the input terminal (1), and output the signals of the pixels one clock before the input pixel.

(Oa) is an output terminal that outputs the signal Sa of the pixel A that is input to the memory (2a), (Ob) is an output terminal that outputs the signal Sb of the pixel B that is input to the memory (2b), and (Oc) is Delay device (2
The output terminal for outputting the signal Sc of the pixel C output from f), and (Od) is the signal Sd of the surface element D input to the delay device (2c).
And (Oe) is an output terminal for outputting the signal Se of the pixel E output from the delay device (2d).

By the way, the input image signal of the input terminal (1) is formed by the sampling process of the sub-sampling method, and as shown by the circles in FIG. 3, the pixel array of each line ln _-1 , ln, ln ₊₁ is , And the array is shifted by one clock (one pixel) from the pixel array of the adjacent line, and the same array is provided every two lines.

Therefore, as shown in FIG. 3, the pixels of the output signals of the output terminals (Oa) to (Oe) are the current pixels corresponding to, for example, the pixel α of FIG. When it becomes a detection pixel, the pixels E, D, C, and A are the pixels β, γ, and
There are four two-dimensional neighboring pixels corresponding to δ and σ, that is, the four pixels closest to the pixel B.

Next, the configuration and operation of the detection unit (3) will be described.

First, the operating principle of the detection unit (3) will be described.

For example, when detecting a level change for the pixel B in FIG. 3, the level difference between the two pixels on the same line, that is, the pixel A, is determined based on the four neighboring pixels A, C, D, E which are two-dimensionally adjacent to the pixel B. , C level difference Le ₁ and pixel D, E level difference Le ₂
And the level difference between two pixels at the same position on the lines above and below the pixel B, that is, the level difference Ll ₁ between the pixels A and D and the pixel C,
The level difference Ll ₂ between E and the level difference Le ₁ ,
Le _2, Ll _1, Ll of the absolute value _2, as compared to the reference level each of the level differences le _1, le _2, Ll _1, set corresponding to the Ll ₂ levels change, larger level difference than the reference level If there is even one, a level change above the reference level, that is, an edge portion is detected, and an edge detection signal is output.

By the way, the detection based on the above-described principle operation is performed by the block (5), but in order to prevent erroneous detection of the block (5) due to isolated point noise or the like, the detection unit (3) is used.
In general, the block (4) is provided in front of the block (5).

The block (5) is configured as shown in FIG. 4, and when the detection pixel is B, the input terminals (Ia), (Ic),
Outputs from (2) to (Id) and (Ie) via output terminals (Oa) ', (Oc)', (Od) ', and (Oe)' of block (4) described later. Signals Sa, S of neighboring pixels A, C, D, E
c, Sd, and Se are input.

Further, the signals Sa and Sd from the input terminals (Ia) and (Ic) are input to the subtractor (5a), and the input terminals (Id) and (Ie)
Since the signals Sd and Se of are input to the subtractor (5b) and the signals Sa and Sd of the input terminals (Ia) and (Id) are input to the subtractor (5c), the subtractors (5a) and (5b ) And (5c), the level differences Le ₁ , Le ₂ , and Ll ₁ are calculated.

The output signals of the subtractors (5a) to (5c) are absolute values | Le ₁ |, | Le ₂ |, by the absolute value circuits (5d), (5e), and (5f).
Converted to | Ll ₁ | signal.

Then, the signals of absolute values | Le ₁ |, | Le ₂ |, | Ll ₁ | are input to the non-inverting input terminals (+) of the level comparators (5g), (5h), and (5i), respectively. To the inverting input terminals (-) of the comparators (5g) to (5i), switch registers (5j), (5
k), (5l) reference level signals set respectively,
That is, since the reference level signal for detecting the level change for each of the level differences Le ₁ , Le ₂ , and Ll ₁ is input,
The comparators (5g) to (5i) output high signals when the absolute values of the level differences Le ₁ , Le ₂ , Ll ₁ | Le ₁ |, | Le ₂ |, | Ll ₁ | Become a level.

By the way, the output signal of the absolute value circuit (5f) passes through the two 1-clock delay units (5m) and (5n), and the level comparator (5m
It is also input to the non-inverting input terminal (+) of o).

Then, since the input signals of the input terminals (Ia) and (Id) become the signals Sc and Sd of the pixels C and E, respectively, two clocks before the current clock, the non-inverting input terminal (+) of the comparator (5o) The signal of the absolute value | Ll ₂ | of the level difference Ll ₂ is input to the reference level signal set by the switch register (5p), that is, the level change detection reference for the level difference Ll ₂ . Comparator (5o) because the level signal is input
Outputs a high level when the absolute value | Ll ₂ | of the level difference Ll ₂ is equal to or larger than the reference.

Further, the output signals of the comparators (5g) to (5i) and (5o) are input to the OR gate (5q), and the comparators (5g) to (5g)
When at least one of the output signals of i) and (5o) becomes high level, a high level edge detection signal is output from the OR gate (5q) to the switching devices (6) and (10) through the output terminal (Os). Is output.

By the way, if only the block (5) is provided in the detection unit (3), when the input image signal includes sudden and steep noise such as isolated point noise, the detection unit (3) detects the noise. ) May cause false detection.

Therefore, usually, the block (4) is provided in front of the block (5), and the noise is reduced and removed from the output signal of the extraction unit (2), and then the block (5) performs detection.

The block (4) is constructed as shown in FIG. 5, and the input terminals (Ia) ', (Ib)', (Ic) ', (I
d) ′ and (Ie) ′ have output terminals (O) of the extraction unit (2).
a), (Ob), (Oc), (Od), and (Oe) 5 pixels A, B, C,
The signals Sa, Sb, Sc, Sd, and Se of D and E are input.

Further, the five input switches (SWi) connected to the input terminals (Ia) 'to (Ie)', respectively, and the four output switches (SWo) which will be described later, normally operate the block (4). Therefore, when the isolated point noise is unnecessary to be removed, for example, when the input image signal is sufficiently larger than the noise, the manual switching or the automatic switching interlocks with the disconnecting side contact (x). y).

When the switches (Swi) and (SWo) are held at the contact (x), the signals Sa to Se of the input terminals (Ia) 'to (Ie)' are input to the isolated point removal circuit (4a). The removing circuit (4a) removes noise such as isolated point noise based on the averaging of the levels of the five signals Sa to Se, and removes isolated point noise and the like from the removing circuit (4a). The signal of each pixel is output in time series.

Incidentally, the removal characteristic of the removal circuit (4a) is set to a characteristic of performing an averaging process or the like to the extent that the detection of the block (5) is not hindered.

Furthermore, the output signal of the removal circuit (4a) is converted again into the parallel signal of the above-mentioned 5 pixels A to E, and the 4 neighboring pixels A, C,
Since the D and E parallel signals are supplied to the block (5), the block (4) includes memories (2a) and (2b) of the extraction unit (2).
And two line memories (4b) and (4c) similar to the delay units (2c) to (2f) and four 1-clock delay units (4
d), (4e), (4f), and (4g) are provided, and at this time, the output signal of the removal circuit (4a) becomes the signal Sa of the pixel A, and the delay device (4g), the memory (4c), the delay device The output signal of (4e) becomes the signals Sc, Sd, and Se of the pixels C, D, and E, respectively.

Then, the removal circuit (4a), the delay device (4g), the memory (4
c), the output signals Sa, Sc, Sd, Se of the delay device (4e) are output terminals (Oa) ′, (Oa) through four output switches (SWo).
c) ', (Od)', (Oe) ', and the signals at the output terminals (Oa)', (Oc) ', (Od)', (Oe) 'are input to the block (5). Terminals (Ia), (Ic), (I
It is input to each of d) and (Ie).

When each switch (SWi), (SWo) is switched to the contact (y), each output terminal (Oa), (O) of the extraction unit (2)
The output signals of c), (Od), and (Oe) are input to the input terminals (Ia), (Ic), (Id), and (Ie) of the block (5), respectively.

The block (4) may be omitted when the configuration of the detection unit (3) is simplified.

When a level change is detected by the detection unit (3), the edge detection signal Sf causes the switches (6) and (10) to switch to the contacts (b) and (b) '.

Next, the configuration and operation of the switch (6) and the processing unit (7) will be described.

First, the switching circuit (6) and the control circuit (8) of the processing unit (7) are configured as shown in FIG. 6, and when the edge detection signal Sf is output, five switching circuits (6) are formed. Switch (6
a), (6b), (6c), (6d), and (6e) are switched to the illustrated contact (b), and the output terminals (Oa) to (Oa) of the extraction unit (2) to
The output signal of (Oe) passes through the switches (6a) to (6e), and the subtracter (801) and the 1-clock delay unit (80) of the control circuit (8) are provided.
2), the subtracters (803), (804), and (805), respectively.

Further, the output signal of the delay device (802) is the subtractor (801),
Input to (803), (804), (805).

Therefore, the subtracters (801) and (803) to (805) are provided with the level differences Lba and L between the detection pixel B and the four neighboring pixels A, C, D and E, respectively.
bc, Lbd, Lbe are detected, and at this time, each subtractor (801), (80
3) to (805) output signals are absolute value circuits (806), (80
7), (808), and (809), respectively, so that from the absolute value circuits (806) to (809), the pixel B and each pixel A, C,
The signals of the absolute values of the level differences Lba to Lbe with respect to D and E respectively are output.

Then, the output signals of the absolute value circuits (806) and (807) are input to the non-inverting input terminal (+) and the inverting input terminal (−) of the subtractor (801), respectively, and the absolute value circuit (80
The output signals of 8) and (809) are input to the non-inverting input terminal (+) and the inverting input terminal (-) of the subtractor (811),
The output signals of the absolute value circuits (809) and (807) are the non-inverting input terminal (+) and the inverting input terminal (-) of the subtractor (812).
Input to each, absolute value circuit (808),
The output signal of (806) is input to the non-inverting input terminal (+) and the inverting input terminal (-) of the subtractor (813).

Therefore, the output signals of the subtracters (810) to (813) are the difference signal of the absolute values of the level differences Lba and Lbc, the difference signal of the absolute values of the level differences Lbd and Lbe, and the absolute value of the level differences Lbe and Lbc. Of the absolute value of the level difference Lbd, Lba.

The output signals of the subtracters (810) to (813) are input to the absolute value circuits (814), (815), (816) and (817), respectively, and the absolute value circuits (814) to (817) are also input. Is input to the non-inverting input terminal (+) of each of the amplitude comparators (818), (819), (820), (821), and at this time, the inverting input of each comparator (818) to (821). Since the signals of the set levels of the switch registers (822), (823), (824), and (825) are input to the terminal (-), each comparator (81
8) to (821) are high level when the output signals of the absolute value circuits (814) to (817) are higher than the set levels of the switch registers (822) to (825), respectively.

Further, the output signals of the comparators (818) to (821) are input to one input terminal of each of the AND gates (826), (827), (828), (829), and at this time, the AND gate (82).
Subtractors (810) to (8) to the other input terminals of 6) to (829).
Since the code signal of 13), that is, the code signal that becomes high level when the output signal is positive, is input, the AND gates (826) to (829) use the code signals of the subtracters (810) to (813). The output signal becomes high level only when both and the output signals of the comparators (818) to (821) become high level.

Then, the four first control signals composed of the output signals of the AND gates (826) to (829) are output terminals (Og), (Oh),
Four types of second control signals, which are output to the respective (Oi) and (Oj), and which are the output signals of the comparators (818) to (821), are output terminals (Ok), (Ol), (Om), (On) Output to each.

The output signal of the delay device (802) is output to the output terminal (Ob) '.

That is, when the control circuit (8) detects a level change, any of the four pixels A, C, D, and E in the two-dimensional vicinity of the detection pixel determines the reference level for detection by the detection unit (3). The value (reference) is used to identify whether the detected pixel B is a pixel of a level opposite to the level to which it belongs, and the pixel of the opposite level may be replaced with any of the four pixels A, C, D, E. It is provided to determine whether or not.

The output signals of the comparators (818) to (821) indicate which of the two pixels A and C, D and E, E and A, D and C has a level change, and the AND gate (826). The output signals of () to (829) indicate which of the two pixels A and C, ..., D and C is the pixel of the opposite level.

For example, the output signal of the comparator (818) becomes high level if there is a level change between the two pixels A and C, and if the pixel A has the opposite level at this time, the output signal of the AND gate (826). Becomes high level, and if the pixel C has the opposite level, the output signal of the AND gate (826) becomes low level.

Next, the replacement circuit (9) is configured as shown in FIG. 7, and the output terminals (Oa), (Oc) -through the switches (6a), (6c)-(6e) of the switch (6). The output signal of (Oe) is
Input to each of the contacts (c) of the switches (901), (902), (903), (904), and at this time, the output terminal (O
The output signals of a) and (Od) are also input to the other contacts (d) of the switches (902) and (904), respectively.

Further, the output signal of the switching piece of the switch (902) is input to the other contact (d) of the switch (901) and one contact (e) of the switch (905), and at the same time, the switch (90)
The output signal of the switching piece in 4) is the switch (903), (905).
Are input to the other contacts (d) and (f) of the switch and one contact (e) of the switch (906).

Further, the output signal of the switching piece of the switch (905) is input to the other contact (f) of the switch (906),
The output signal of the switching piece of the switch (901) is the switch (907).
One contact (g), the other contact (h) of the switch (908), and the output signal of the switching piece of the switch (903) is input to one contact (g) of the switch (908). At the same time, the output signal of the switching piece of the switch (908) is input to the other contact (h) of the switch (907).

And the switches (907), (905), (908), (906)
The output signal of the switching piece of becomes the signal of the replacement pixel of each of the pixels A, C, D, E, and the output terminals (Oa) ″, (Oc) ″,
It is output to each of (Od) ″ and (Oe) ″.

By the way, the switch (901) is switched by the level of the signal of the output terminal (Og) and the signal of the output terminal (Ok) via the inverter (909) and the on / off state of the switch (910) for gate. The switch (902) is connected to the output terminal (O
g), (Ok) signal level and gate switch (91
The switch (910), (911) is turned on by the high level of the signal at the output terminal (Ok), and the switch (901), (902) is switched on and off.
Switches to the other contact (d) when low level signals are input from the switches (910) and (911) respectively.

Further, the switch (903) is switched by the level of the signal of the output terminal (Oh) and the signal of the output terminal (Ol) via the inverter (912) and the on / off of the switch (913) for gate. Switch (904) is connected to the output terminal (O
h), (Ol) signal level, and gate switch (9
14) is switched on and off, and at this time the switches (913) and (914) are both turned on by the high level of the signal at the output terminal (Ol), and the switches (903) and (904) are When a low level signal is input from each of the switches (913) and (914), the contact is switched to the other contact (d).

Similarly, the switch (906) is based on the level of the signal at the output terminal (Oi) and the signal at the output terminal (Om) via the inverter (914) and the on / off state of the switch (915) for the gate. The switch (905) is switched to the output terminal (O
i), (Om) signal level, and gate switch (9
16) It is switched by and.

The switch (908) is switched by the level of the signal of the output terminal (Oj) and the signal of the output terminal (On) via the inverter (917) and the on / off state of the switch (918) for the gate. The switch (907) is connected to the output terminal (O
j), (On) signal level and gate switch (9
It can be switched by turning on and off in 09).

Therefore, for example, in the case of replacing the pixel A with the pixel C in the pixels A and C, the signals at the output terminals (Og) and (Ok) become high level, and the switch (901) is switched to the contact (d). , Switch (901) output signal is output terminal (O
c) The output signal, that is, the signal of the pixel C is replaced, and conversely, when the pixel C is replaced by the pixel A, the signal of the output terminal (Og) becomes low level and the signal of the output terminal (Ok) becomes high. The level is changed to the other contact (d) of the switch (902), and the output signal of the switch (902) is replaced with the output signal of the output terminal (Oa), that is, the signal of the pixel A.

And each set of switches (903) and (904), (905)
And (906), (907), and (908) operate in the same manner as described above, the output terminals (Oa) ″, (Oc) ″ through (Oe) ″ are respectively operated when a level change is detected. Output terminal (Oa), (O
Output signals c) to (Oe), that is, signals in which the pixels A, C, D, and E are replaced with any other pixels of the level to which the pixel B belongs are output.

Next, the configurations and operations of the switching device (10) and the filter (11) will be described.

The switch (10) and the filter (11) are configured as shown in FIG. 8, and when the edge detection signal Sf is output, the switch (1
5 switches forming (0) (10a), (10b), (10)
c), (10d) and (10e) are switched to the contact (b) 'shown in the figure, and the signal output from the switch (10) to the filter (11) is the signal of the output terminals (Oa) to (Oe). Output terminal (O
a) ", (Ob) ', (Oc)" to (Oe) "signals are switched. The signal at the output terminal (Ob)' is delayed by one clock by the one-clock delay device (10f). Is output.

The output signals of the switches (10a) to (10e) are converted into coefficient units (11a), (11b), (11c), (11) of the filter (11).
d) and (11e) respectively, and the coefficient units (11a) to (1
1e) to accumulator (11f), switches (10a) to (10
A signal obtained by multiplying the output signal of e) by a predetermined filter coefficient is output, and by the arithmetic processing of the accumulator (11f),
2 detection pixels B from the accumulator (11f) to the output terminal (12)
The video signal subjected to the dimensional filter processing is output.

That is, when the edge detection signal Sf is not output, the filter (11) is connected to the input terminal (1) through the extraction unit (2).
When the detection pixel B and the neighboring pixels A, C, D, and E of the input image signal of are input as they are to the filter processing, and the edge detection signal Sf is output, the reverse of the detection pixel B is detected in the filter (11). The image signal after the replacement processing in which the neighboring pixels of the level are replaced by the neighboring pixels of the level to which the detected pixel belongs is input and filter processing is performed.

Therefore, in the edge part of the field of the input pixel signal, the two-dimensional neighboring pixels input to the filter (11) are always replaced with the pixel of the level to which the detected pixel belongs, and the filter processing is performed to fix the band characteristic. By the in-field processing of the small-scale circuit configuration using the filtered filter (11), the blur of the edge part can be suppressed as much as possible to reduce the noise, and the blur of the image signal of both the moving image and the still image can be reduced. It is possible to significantly suppress the occurrence of noise and perform good noise reduction processing.

In the above embodiment, the digital image signal formed by the sub-sampling method is used as the input image signal. However, the digital image signal formed by various processing other than the sub-sampling method, that is, the pixel array. Can be applied when a digital image signal different from that shown in FIG. 3 is used as the input image signal, and at this time, the respective parts (2), (3),
The configurations of (9), the switches (6), (10) and the filter (11) are modified from the configuration of the embodiment based on the pixel arrangement, selection of neighboring pixels and the like.

〔The invention's effect〕

As described above, according to the noise reducing apparatus of the present invention, when the level change detecting means detects the level change in the horizontal and vertical directions of the field screen formed by the input image signal, the edge portion of the image is detected. As a result, the replacement processing means determines, based on the identification result of the identification means based on the edge detection signal of the edge detection means, a pixel having a level opposite to the level to which the detection pixel belongs among the pixels in the two-dimensional neighborhood of the detection pixel. Of the pixels in the two-dimensional neighborhood, the pixels in the level to which the detected pixel belongs are replaced, and each pixel in the two-dimensional neighborhood is converted into a pixel in the level to which the detected pixel belongs, and the image signal after this replacement processing is the input image signal. Instead of the above, the filter is supplied to a two-dimensional low-pass filter and subjected to the filter processing, so that a filter having a simple configuration using a two-dimensional low-pass filter having a predetermined band characteristic is used. In filter processing within over beam, then minimizing the blurring of edge portions level change is detected is capable of performing good noise reduction processing.

[Brief description of drawings]

1 to 8 show an embodiment of the noise reducing apparatus of the present invention. FIG. 1 is a block diagram, FIG. 2 is a detailed block diagram of a neighboring pixel extracting section, and FIG. 3 is a diagram showing an input image signal. FIG. 4, FIG. 5, FIG. 6, FIG. 7, FIG. 7, and FIG. 8 are pixel pattern diagrams for explanation, which show details of the detection block, the isolated point noise removal block, the replacement control circuit, the replacement circuit, and the low-pass filter. The block diagrams and FIGS. 9A and 9B are pixel pattern diagrams for explaining the operation of the conventional noise reduction device. (1) ... image input terminal, (2) ... neighborhood pixel extraction unit,
(3) ... edge detection section, (6), (10) ... input / output side switching unit, (7) ... edge processing section, (11) ... low-pass filter, (12) ... image Output terminal.

Claims (1)

[Claims] 1. A noise reduction apparatus for reducing noise of a digital image signal by performing in-frame filter processing using a low-pass filter, wherein a level between pixels in a two-dimensional vicinity of each pixel of an input image signal. Level change detection means for detecting a level change in the horizontal and vertical directions of the frame screen formed by the input image signal from the difference, and at least one of the horizontal and vertical directions by the detection means.
Edge detection means for outputting an edge detection signal as detecting an edge portion of an image each time a level change in the direction is detected, and a reference level for detection by the level change detection means each time the edge detection signal is output. A threshold value for identifying a pixel having a level opposite to the level to which the detection pixel belongs among the pixels in the two-dimensional vicinity of the detection pixel; A replacement processing unit that replaces one of the pixels of the level to which the detected pixel belongs and replaces each pixel in the two-dimensional neighborhood of the detected pixel with a pixel of the level to which the detected pixel belongs; When the input image signal is not detected, the input image signal is filtered, and when the level change is detected, the image signal after the replacement processing of the replacement processing means is filtered. Noise reduction apparatus being characterized in that a two-dimensional low-pass filter of the band characteristic.