JP2611598B2 - Image compression device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image compression apparatus for compressing image data, and more particularly, to an improvement in an image compression apparatus for performing pre-processing such as noise removal at a stage prior to compression processing.

[0002]

2. Description of the Related Art In an image data compression apparatus, FIG.
As shown in (1), some pre-processing may be performed by the pre-filter 200 in order to perform the compression processing by the band compression processing apparatus 100 favorably. For example, an image on which noise is superimposed, an image having a frequency band more than necessary, an image having a small horizontal or vertical correlation portion, and the like are all considered to be images having a large amount of information. In such a case, the compression distortion in the compression processing can be reduced by previously removing unnecessary image information amount by the pre-processing.

[0003] As a conventional technique using such a pre-filter 200, there is the following. First, an apparatus disclosed in Japanese Patent Application Laid-Open No. 61-141289 is shown in FIG.
2, a noise elimination circuit 202 is provided in a preceding stage. The noise elimination circuit 202 performs smoothing when the signal difference at the same position between the frames oscillates positively and negatively, and outputs the difference when the signal difference is continuous positively or negatively.

Next, the one disclosed in Japanese Patent Laid-Open No. 62-198269 is schematically shown in FIG.
A high spatial frequency reduction pre-processing circuit 204 is provided before the redundancy suppression coding processing device 104, and a high spatial frequency emphasis post-processing circuit 206 is provided after the decoding processing device 106. The high spatial frequency reduction pre-processing circuit 204 performs pre-processing for reducing high spatial frequency components in the image data, and the high spatial frequency emphasizing post-processing circuit 206 performs high spatial frequency component emphasis processing in the expansion process.

[0005] Next, as disclosed in Japanese Unexamined Patent Publication No. 62-226383, a smoothing circuit 208 is provided before the band compression processing unit 108 as schematically shown in FIG.
Is provided. The smoothing circuit 208 performs smoothing by using an average value when pixel data is separated by a predetermined amount or more between adjacent pixels, and by using a median value when pixel data is not separated by a predetermined value or more.

[0006] Next, as disclosed in Japanese Patent Application Laid-Open No. 63-158972, a noise elimination filter 2 is provided before the compression encoding circuit 110, as schematically shown in FIG.
10 are provided, and a distortion removal filter 212 and an outline emphasis filter 214 are provided at the subsequent stage of the decoding / expansion circuit 112. Overload distortion, quantization distortion, and the like included in the image data decompressed and decoded by the decompression decoding circuit 112 are removed by the distortion removal filter 212, and thereafter, the contour is removed by the contour emphasis filter 214 that acts in reverse of the distortion removal filter 212. Emphasis is made.

Next, as disclosed in Japanese Unexamined Patent Application Publication No. 2-29179, a noise removal circuit 216 is provided at a stage preceding the adaptive variable sampling device 114 as schematically shown in FIG. It has a configuration. Before the sampling process by the adaptive variable sampling device 114,
Noise removal circuit 21 constituted by a non-linear filter
At 6, the noise component is removed.

[0008]

However, in each of these prior arts, the same filter processing such as noise removal is performed on the input image data, so that information such as unnecessary noise can be obtained. The purpose is to leave only necessary video information. However, a filtering error always occurs in any of the techniques, and 100% of the video information is not transmitted. In particular, when the processing by the pre-filter 200 is performed by applying the above-described conventional technology when the amount of information of the input image is small, inconveniences such as deterioration of resolution and image quality occur.

More specifically, band compression of image data uses image correlation or redundancy in any system. Therefore, for example, in the case of an image having a large area having a strong correlation in the horizontal or vertical direction as shown in FIG. 4A, the amount of image information is small as a whole and compression is easy. In addition, compression distortion hardly occurs. However, in the case of an image as shown in FIG. 3B, the amount of image information is large, and distortion tends to occur. Also, when random noise is superimposed on the image shown in FIG. 2A, this noise increases the apparent image information amount. Since the noise is a signal having a weak correlation, the result is similar to the image shown in FIG. That is, the original video signal is distorted (changed) by unnecessary information.
Will occur.

In the above-described band compression technique, such various images are generally compressed to a fixed code length. For example, a fixed amount is set for one frame. Therefore, due to the difference in the amount of input information,
In the case shown in FIG. 10A, the reproduced image and the original image hardly differ from each other, but in the case shown in FIG. 10B, a large distortion occurs and the reproduced image is deteriorated. The present invention
It is an object of the present invention to pay attention to these points and to provide an image compression apparatus capable of improving the image quality by performing a good compression process while eliminating the inconvenience caused by the pre-filter process.

[0011]

To achieve the above object, the present invention provides a compression means for compressing information of an input image,
An image compression apparatus comprising a pre-filter means that performs information amount reduction of the input image, obtained from the compression means compressing
The amount of image information afterward is determined by comparing the time when the pre-filter
By comparing with different comparison values at FF,
Switching the pre-filter means ON and OFF
The prefix is set so that the image information amount has a hysteresis.
The filter control means for controlling the injection of the filter means is provided.

According to another aspect of the present invention, there is provided an image compression apparatus including a compression unit for compressing information of an input image and a pre-filter unit for reducing the amount of information of the input image, wherein the pre-filter unit performs multi-stage reduction of the information amount. Has a plurality of filters performed in, the amount of image information after compression obtained from the compression means,
When each filter selected according to the image information amount is ON
By comparing each with a different comparison value at OFF and
Therefore, the ON / OFF switching of the filter and the image information
Inject the filter so that the amount has hysteresis
It is characterized by comprising filter control means for controlling .

[0013]

According to the present invention, the pre-filter means is prepared, and the information amount of the image is detected by the compression means.
You. Then, the prefix is adaptively adjusted according to the amount of image information.
The filter means is provided before the compression means. That is, the pressure
The amount of image information after reduction is the same as when the pre-filter means is ON.
At the time of OFF, each is compared with a different comparison value. to this
Switching of pre-filter means ON and OFF and image
The amount of information has a hysteresis. Prefill
If filters are provided in multiple stages in the
At least one is selectively input according to the amount of image information. this
Switching system with hysteresis for each filter
Control is performed.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of an image compression apparatus according to the present invention will be described below in detail with reference to the accompanying drawings. <First Embodiment> First, a first embodiment of the present invention will be described with reference to FIGS. FIG. 1 shows a main part of the first embodiment. In FIG. 1, an input terminal 10 for image data to be subjected to compression processing is connected on one side to an input side of a pre-filter 12 and on the other side to one switching input side of a closing switch 14. The output side of the pre-filter 12 is connected to the other switching input side of the closing switch 14.

The output side of the switch 14 is connected to a band compression processor 20. This band compression processing device 2
On the input side of 0, an in-plane redundancy extraction circuit 22 is provided, and the input side is connected to the output side of the closing switch 14. A variable information reduction circuit 24 is connected to the output side of the in-plane redundancy extraction circuit 22, and an encoding circuit 26 is connected to the output side. The output side of the encoding circuit 26 is the output side of the band compression device 20, which is connected to the output terminal 28.

A code amount control circuit 30 is connected to the other input side of the variable information reduction circuit 24 and the encoding circuit 26, respectively. The output side of the in-plane redundancy extraction circuit 22 is connected to the code amount control circuit 30, and the output side of the requantization level is connected to the filter control circuit 32.
The filter control circuit 32 receives comparison values THD1 and THD2, which will be described later, from a memory 34. The output side of the filter control circuit 32 is connected to the control input side of the closing switch 14.

The input terminal 10 of each of the above-mentioned units is adapted to receive, for example, image data of a digital video tape recorder (VTR). Next, the pre-filter 12 may be of any type that can reduce the amount of image information, such as the noise removal filter shown in the above-mentioned prior art, but in this embodiment, a horizontal low-pass filter is used. Have been. That is, when the image data is filtered by the pre-filter 12, high frequency components in the horizontal direction are removed, the frequency band is narrowed, and the amount of image information is reduced.

Next, any method may be used as the band compression processing device 20. However, considering the special reproduction similar to the analog VTR, a method using intra-picture image correlation is advantageous. In consideration of the processing at the time of editing, a method of compressing one screen to a certain fixed length code amount is generally used. This means that the conventional analog VTR has one field = 1 track. Therefore, in this embodiment, one screen of either a field or a frame is
A compression method for controlling the code amount to the track length is used.

The in-plane redundancy extraction circuit 22 of the band compression processing device 20 has a function of extracting redundancy in units of fields or frames. The variable information reduction circuit 24 has a function of making the code amount constant so that the information amount of one screen corresponds to one track. For example, a re-quantization process in orthogonal coding using cosine transform corresponds to this. The encoding circuit 26 has a function of performing a predetermined encoding process on the input data. The code amount control circuit 30 is for controlling the variable information reduction circuit 24 for setting a constant code amount.

Next, the filter control circuit 32 compares the amount of image information obtained from the code amount control circuit 30 with the comparison value input from the memory 34, and controls the switching of the closing switch 14 according to the result. Has a function. In the present embodiment, control is performed by the code amount control circuit 30 so that one screen is equal to one track, that is, the code amount is constant.
Corresponds to the amount of image information.

That is, to keep the code amount constant,
When the amount of input image information is large, it is necessary to reduce a large amount of information. Therefore, the requantization level becomes rough. On the other hand, when the input image information amount is small, the requantization level becomes smaller. The filter control circuit 32 can obtain the amount of image information from the requantization level supplied from the code amount control circuit 30 using such a relationship.

The comparison value TH output from the memory 34
D1 and THD2 are values set in advance from the relationship between the requantization level and the compression distortion, that is, the image information amount experimentally obtained in advance. FIG. 2 shows the relationship between these comparison values THD1 and THD2 and the opening / closing control of the closing switch 14. When the image information amount becomes THD1 or more in a state where the closing switch 14 is OFF, the closing switch 14 is turned ON as indicated by an arrow FA.
Becomes Then, when the image information amount becomes THD2 or less while the switch 14 is ON, the switch 14 is turned OFF as indicated by an arrow FB.

Next, the operation of the first embodiment configured as described above will be described with reference to FIGS. FIG. 3 shows the relationship between the change of the input image and the presence / absence of filter input, and FIG. 4 shows an example of the input image.

First, assume that an image having a relatively small amount of information as shown in FIG. 4A is input (see FIG. 3TA). In this case, the image information amount is relatively small and the comparison value THD1
Switch 14 is OFF,
The input image is directly input to the band compression processing device 20 and compression processing is performed. That is, the above-described processing is sequentially performed on the input image by the in-plane redundancy extraction circuit 22, the variable information reduction circuit 24, and the encoding circuit 26, and the compressed data is output from the output terminal 28.

Assume that the input image changes as shown in FIG. 4B due to, for example, a scene change (T in FIG. 3).
B). Then, this image is input to the band compression processing device 20 at first. Since this image has a large amount of information, the requantization level in the code amount control circuit 30 becomes rough. Based on this, when the filter control circuit 32 determines that the image information amount has exceeded the comparison value THD1, the closing switch 14 is turned on as shown by an arrow FA in FIG.
That is, as shown in FIG. 3, when the input image changes from FIG. 4A to FIG. 4B, the switch control based on the information amount from the code amount control circuit 30 is performed by the feedback control. The turning-on switch 14 is turned ON one screen later. Therefore, for the input image,
First, the processing by the pre-filter 12 is performed.

Next, it is assumed that the input image changes from FIG. 4B to FIG. 4A due to a scene change (see TC in FIG. 3). Then, at first, this image is transmitted to the band compression processing unit 20.
Is input to Since this image has a small amount of information, the requantization level in the code amount control circuit 30 becomes fine. Therefore, when the filter control circuit 32 determines that the image information amount has become smaller than the comparison value THD2, the closing switch 14 is turned off as shown by an arrow FB in FIG. That is, as shown in FIG.
When the state changes from (B) to (A) in the same figure, the switch 14 is similarly turned off with a delay of one screen. Therefore, the input image is directly supplied to the band compression processing device 20 without passing through the pre-filter 12.

As described above, in this embodiment, when the amount of image information is small, the switch 14 is turned off and the pre-filter 12 is not turned on. However, when the amount of image information is large, the closing switch 14 is turned on and the pre-filter 1 is turned on.
2 is input. Hereinafter, each case will be described in more detail.

A, In the case of an image having a small amount of information First, in the case of an image having a small amount of information as shown in FIG. 4A, the quantization noise is small due to the fine requantization level, and the compression distortion is therefore small. Inconspicuous. On the other hand, the deterioration of the spatial resolution due to the reduction of the information amount of the pre-filter 12 becomes more noticeable. Therefore, in the present embodiment, the closing switch 14 is turned off by the filter control circuit 32, and the pre-filter 12 is not turned on.

B, In the case of an image having a large amount of information On the other hand, in the case of an image having a large amount of information as shown in FIG. 4B, since the requantization level is rough, the quantization noise generated at that time is compressed. Appears as distortion. On the other hand, in the image after the information amount is reduced by the pre-filter 12, the image information amount is reduced and the sharpness of the reproduced image is reduced, but the deterioration due to the compression distortion is also reduced. Here, the decrease in sharpness is similar to the out-of-focus state of the camera, and does not seem so unnatural when viewed as image quality deterioration. On the other hand, deterioration due to compression distortion often has an unnatural property as seen in mosquito noise, and the viewer has a strong impression of image quality deterioration. Therefore, even if the spatial resolution is sacrificed by inputting the pre-filter 12, it is considered that the quality as a reproduced image is better when the compression distortion is reduced.
Also, in an image on which noise is superimposed, the noise removal effect of the pre-filter 12 is visually more effective.

As described above, when the amount of image information is large and compression distortion occurs, the effect of reducing the amount of information by the pre-filter 12 is great. Therefore, in the present embodiment, the closing switch 14 is turned on by the filter control circuit 32, and the pre-filter 12 is closed.

The values of the comparison values THD1 and THD2, which are the references for the opening / closing control of the closing switch 14, are experimentally determined in consideration of the above-described compression distortion and spatial resolution. Further, since both are set to different values, the opening and closing of the closing switch 14 and the amount of image information have a hysteresis as shown in FIG. When the pre-filter 12 is inserted, the requantization level in the band compression processing device 20 becomes fine due to the effect of the reduction in the amount of information. Then, the filter control circuit 32 determines that the apparent amount of image information has decreased. However, the turning-off switch 14 is turned off.
It is better to perform the processing when the original information amount of the image is reduced to the extent that the compression distortion is not conspicuous. For such a reason, the comparison value THD2 of the OFF control becomes ON as shown in FIG.
It is set lower than the control comparison value THD1.

Since the opening / closing control of the closing switch 14 by the filter control circuit 32 is performed before the code amount control in the band compression processing device 20, the control is performed only in the form of feedback as described above. Can not. Therefore, in the present embodiment shown in FIG. 1, a time difference of one screen occurs between the change in the information amount of the input image and the opening / closing control of the closing switch 14. However, the amount of image information of two consecutive screens is empirically almost the same, and one screen is short (1/60 sec in a field unit, 1/3 in a frame unit).
Since the compression distortion is considered to be inconspicuous because of 0 sec), this time delay does not visually impair.

Further, in FIG. 2, the hysteresis width (TH
D2−THD1) may be set as appropriate. However, if the input image is too narrow, hunting may occur when the input image fluctuates around the comparative value THD.

As described above, according to this embodiment, when the information amount of the input image is small, the pre-filter 12 is not applied, and only the band compression processing is performed. Therefore, the spatial resolution does not deteriorate due to the reduction in the amount of information of the pre-filter 12. On the other hand, when the information amount of the input image is large, the pre-filter 12 is inserted to reduce the information amount. For this reason, deterioration due to compression distortion in the band compression processing is reduced. Therefore, as a whole, the compression processing is performed favorably while eliminating the inconvenience caused by the pre-filter processing, and the image quality is improved. In addition, since it has a simple configuration in which only a filter control circuit, a memory, and the like are provided, it is easy to implement an IC.

<Second Embodiment> Next, a second embodiment of the present invention will be described with reference to FIGS. In the first embodiment, the input control of the pre-filter 12 is performed according to the information amount of the input image.
The difference in image quality from F is slightly larger. Therefore, in this embodiment, by performing the filter opening and closing operation in multiple stages,
The image quality of all types of images changes smoothly before and after opening and closing the filter.

FIG. 5 shows the main structure of the second embodiment. Note that the same reference numerals are used for components similar to or corresponding to the first embodiment described above. In the figure, the input terminal 10 is connected on one side to the input sides of the first to third pre-filters 50, 52, 54, respectively, and on the other side to the switching input side of the changeover switch 56. The other switch input side of the switch 56 is connected to the first to third switches.
The output sides of the pre-filters 50, 52, 54 are respectively connected.

The output side of the switch 56 is connected to the band compression processor 20. This band compression processing device 2
The output side of the requantization level of the code amount control circuit 30 of 0 (see FIG. 1) is connected to the filter control circuit 58. Each of the comparison values THD1 to THD6 shown in FIG.

The first to third pre-filters 50 and 5
2, 54, so that the filter effect becomes deeper in that order,
That is, the characteristics are set so that the information amount reduction effect is enhanced.

Next, the operation of the second embodiment configured as described above will be described with reference to FIG. The basic operation is the same as that of the first embodiment, and the changeover switch 56 by the filter control circuit 58 is changed according to the information amount of the input image.
Is performed. First, when the information amount of the input image is relatively small, the changeover switch 56 becomes the changeover terminal T0, and no prefilter is turned on. In this state, when the image information amount increases and reaches the comparison value THD1, the changeover switch 56 is switched to the changeover terminal T1, and the first pre-filter 50 is turned on (see arrow FC in FIG. 6).

Next, in this state, when the image information amount further increases and exceeds the comparison value THD3, the changeover switch 56 is changed over to the changeover terminal T2, and the first prefilter 50 is changed over to the second prefilter 52 (see FIG. 4). See arrow FD in FIG. Conversely, when the image information amount decreases and becomes less than or equal to the comparison value THD2, the changeover switch 56 is switched to the changeover terminal T0, and the prefilter is not turned on (see arrow FE in the figure).

When the image information amount further increases and exceeds the comparison value THD5 while the second pre-filter 52 is turned on, the changeover switch 56 is switched to the change-over terminal T3, and the second pre-filter 52 Switching to the three pre-filter 54 (see arrow FF in the figure). Conversely, when the image information amount decreases and becomes equal to or less than the comparison value THD4, the changeover switch 5
6 is switched to the switching terminal T1, and the second pre-filter 5
2 to the first pre-filter 50 (see arrow FG in the figure).

Further, when the image information amount becomes THD6 or less while the third prefilter 54 is turned on, the switch 56 is switched to the switching terminal T2, and the third prefilter 54 is switched to the second prefilter 54. (See arrow FH in the figure)

As described above, according to the present embodiment, a prefilter having a depth corresponding to a large amount of information of an input image is adaptively selected and input. For this reason, the difference in image quality due to the opening and closing of the filter is favorably reduced, and a reproduced image that changes visually very smoothly including hunting is obtained.

<Third Embodiment> Next, a third embodiment of the present invention will be described with reference to FIG. This example is
This is an example in which the present invention is applied to a camera-integrated digital VTR. In the figure, an image signal output side of an image sensor 70 such as a CCD is connected to a camera signal processing section 72, and a pre-filter section 74 is provided on the output side of the camera signal processing section 72. This pre-filter unit 74
The pre-filter 12 and the closing switch 14 shown in FIG. 1 are included, and a band compression processing device 20 is connected to an output side thereof.

The other output side of the camera signal processing section 72 is connected to a brightness detecting section 76.
Are connected to the filter control unit 78 and the motor 80, respectively. The output side of the filter control unit 78 is connected to the pre-filter unit 74. Motor 80
Is for driving the diaphragm 84 of the lens unit 82.
An automatic aperture mechanism is configured by a loop including the camera signal processing unit 72 and the brightness detection unit 76.

Generally, in a video camera, an apparent amount of image information increases due to camera noise (CCD amplifier noise) at the time of photographing a dark part. Therefore, in the present embodiment, the filter control unit 78 controls the input of the pre-filter based on the brightness information output from the brightness detection unit 76 of the automatic aperture circuit of the camera. That is, the filter control unit 7
In 8, when it is determined from the detection output of the brightness detection section 76 that it is during dark area photographing, the pre-filter section 74 turns on the pre-filter. For this reason, the information amount of the image whose information amount has increased due to camera noise is reduced by the pre-filter. As a result, the sharpness of the image tends to be impaired, but the S / N is improved and a visually good image can be obtained.

<Other Embodiments> The present invention is not limited to the above-described embodiments, but includes, for example, the following. (1) As the band compression device, any type such as DPCM and VQ may be used. The same applies to the prefilter. For example, those described as the above-mentioned prior art may be appropriately used. In the above embodiment, band compression is performed within one screen, but this one screen may be a field or a frame. Further, in the second embodiment, a plurality of pre-filters having different filter depths (amounts of information reduction) are used. However, different pre-filters may be used, or a combination of them may be used. Is also good. Furthermore, the switching method of a plurality of pre-filters
In addition to the method of selecting either one, a switching method of increasing or decreasing the number of input filters may be used.

(2) In the above embodiment, the prefilter is not used when the information amount of the input image is small. However, a prefilter having a relatively small filter depth may be used. (3) In the third embodiment, the same effect can be obtained by turning off the aperture (aperture 84) correction of the camera instead of turning on the pre-filter.

(4) In the above embodiment, the input image information amount is obtained from the requantization level. However, this may be obtained by various methods. (5) The present invention can be applied to not only a digital VTR but also various images such as transmission of a video disc or television signal.

[0050]

As described above, according to the image compression apparatus according to the present invention, the input control of the pre-filter is adaptively performed according to the information amount of the image. There is an effect that the image quality can be improved by satisfactorily performing the compression processing while solving the problem.

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing a first embodiment of an image compression apparatus according to the present invention.

FIG. 2 is an explanatory diagram showing an operation of the first embodiment.

FIG. 3 is an explanatory diagram showing an example of filter control in an operation example of the first embodiment.

FIG. 4 is an explanatory diagram showing an example of input images having different amounts of information.

FIG. 5 is a configuration diagram showing a second embodiment of the present invention.

FIG. 6 is an explanatory diagram showing the operation of the second embodiment.

FIG. 7 is a configuration diagram showing a third embodiment of the present invention.

FIG. 8 is an explanatory diagram illustrating an example of a conventional image compression device.

[Explanation of symbols]

Reference Signs List 10: input terminal, 12, 50, 52, 54: pre-filter (pre-filter means), 14: closing switch (filter control means), 20: band compression processing device (band compression means), 22: extraction of in-plane redundancy Circuit, 24 variable information reduction circuit, 26 encoding circuit, 28 output terminal, 30 code amount control circuit, 32, 58 filter control circuit (filter control means), 34, 60 memory (filter control means) , 56... Changeover switch (filter control means), 70
.., Image pickup device, 72, camera signal processing unit, 74, pre-filter unit (pre-filter means), 76, brightness detection unit, 7
8: filter control unit (filter control means), 80: motor, 82: lens, 84: diaphragm (pre-filter means),
FA to FH: arrows.

Claims (2)

(57) [Claims] A compression unit for compressing information of an input image;
And a pre-filter means for reducing the amount of information of the input image, wherein the image information amount after compression obtained from the compression means,
Different comparison values when the pre-filter is ON and OFF
By comparing each of them, the pre-filter means
ON / OFF switching and the amount of image information are hysteresis
So that the pre-filter means is turned on and off.
An image compression device comprising a filter control means . 2. A compression means for compressing information of an input image,
A pre-filter means for reducing the amount of information of the input image, the pre-filter means having a plurality of filters for reducing the amount of information in multiple stages, and a compressed image obtained from the compressing means. The amount of information
When the filter selected according to the image information amount is ON and OF
By comparing each with a different comparison value at F time,
The ON / OFF switching of the filter and the amount of image information
The filter is controlled so as to have hysteresis.
An image compression device comprising: a filter control unit .