JP2002009590A - Digital filter processing system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a digital filter processing system that applies rate revision processing to image or audio information and can reproduce the image or audio information with fidelity even after the filter processing by preserving important information such as edge parts of the original information. SOLUTION: The digital filter processing system having a cross product arithmetic section 110 and a filter processing control section 100 that apply arithmetic processing to a received digital signal, is provided with an edge extract section 120 that extracts edge parts of the received digital signal, an edge pattern extract section 130 that extracts an edge pattern on the basis of combinations of the extracted edge parts, a line decision section 141 that decides a line part on the basis of the edge pattern, and a shift decision section 142 that references filter coefficient revision information stored in a filter processing revision database 143 on the basis of the edge pattern and the line part to revise a filter coefficient for the shift of an interpolated signal position.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a digital filter processing system for performing digital signal rate conversion processing such as pixel density conversion processing or enlargement / reduction conversion processing for image information or compression / expansion conversion processing for audio information.

[0002]

2. Description of the Related Art When performing rate conversion processing such as pixel density conversion or enlargement / reduction conversion processing on image information, interpolation processing for inserting an interpolation signal (interpolated pixel) in the vertical and / or horizontal direction of the image signal is generally performed. Is applied. In such a case, for example, when the interpolation process is performed in the enlargement process of the image information, and even when the gradation is greatly changed between adjacent pixels in the original image information, the original image information Without saving the edge of
Since the interpolation process is performed at the position corresponding to the enlargement ratio,
The image information after the enlargement processing is converted into image information in which a change in gradation is monotonic.

That is, generally, as an interpolation process at the time of rate conversion, a process of approximating and converting information at a position of an interpolation signal as a linear function of information of an adjacent input digital image signal (original pixel) is used. Therefore, the sharpness of the edge portion of the original image information is lost, and the entire image after the enlargement processing becomes a blurred image.

[0004]

In the conventional digital filter processing, as described above, conversion processing such as pixel density conversion processing or enlargement / reduction processing for image information or compression / expansion processing for audio information (hereinafter, referred to as "hereafter"). As a result of performing the interpolation processing in the rate conversion processing), a part of the data of the input digital signal to the digital filter processing unit may be partially lost. Image blur and image unevenness occur in the reproduced image.

[0005] As a measure for reducing such image blur and image unevenness, in the conventional digital filter processing, the output digital signal after the rate conversion processing is subjected to processing such as contour enhancement to solve the problem. However, since information once lost by the rate conversion processing is not subjected to such processing as contour enhancement, there is a disadvantage that an image faithful to an input digital signal cannot be restored.

[0006]

SUMMARY OF THE INVENTION The present invention has been made in view of such a problem, and it is necessary to perform a rate conversion process while storing characteristic signal information in a digital signal input to a digital filter processing unit. Is what you do.

[0007] The digital filter processing system according to the present invention comprises a differential value between adjacent signals of the input digital signal (for example, in the case of image information, a luminance difference value between adjacent image signals (pixels); Means for calculating an intensity difference value between adjacent audio signals (phonemes); means for extracting an edge pattern generated by combining an arbitrary number of the luminance / intensity difference values; a combination of the edge pattern and a filter A filter processing change database prepared in advance corresponding to the type of the processing (rate conversion processing, etc.) for determining the change of the digital filter coefficient, and the digital filter processing is performed by changing the filter coefficient based on the information of the database. And a filter processing control unit to be executed. Specifically, it has the following technical means.

A first technical means in the digital filter processing system according to the present invention is a digital filter processing system having a filter processing means for an input digital signal, wherein the edge extraction means for judging and extracting an edge portion of the input digital signal is provided. Edge pattern extracting means for extracting an edge pattern based on a combination of the plurality of extracted edge parts, and filter coefficient changing means for changing a filter coefficient of the filter processing means based on the extracted edge pattern. It is characterized by having.

A second technical means in the digital filter processing system according to the present invention is the digital technical processing system according to the first technical means, wherein the filter coefficient changing means is used for filter processing based on the plurality of extracted edge patterns. It is characterized by having shift determining means for determining a shift of a position on the time axis of the interpolation signal.

[0010] A third technical means of the digital filter processing system according to the present invention is the digital technical processing system according to the first technical means, wherein the filter coefficient changing means according to the type of the extracted edge pattern and the number of extractions of the type. A line determining unit that determines a line portion of the input digital signal, and a shift determining unit that determines a shift of a position on the time axis of the interpolation signal used for the filtering process based on the edge pattern and the determination result of the line portion. And determining means.

A fourth technical means in the digital filter processing system according to the present invention is a digital filter processing system according to any one of the first to third technical means, wherein the digital filter processing system has the extracted edge pattern and the determination result of the line portion. Based on the determination result of the line portion, the filter coefficient conversion means,
The method is characterized in that the position on the time axis of the interpolation signal located immediately near the edge is shifted to the position on the time axis of the input digital signal having the extracted information on the edge.

A fifth technical means of the digital filter processing system according to the present invention is the digital filter processing system according to any one of the first to fourth technical means, wherein the edge pattern extracting means stores the combination of a plurality of edge portions. It is characterized by having a database.

A sixth technical means of the digital filter processing system according to the present invention is the digital technical processing system according to any one of the third to fifth technical means, wherein the line determining means determines the type of the extracted edge pattern and the number of extractions. And a line determination database for storing

A seventh technical means of the digital filter processing system according to the present invention is the digital filter processing system according to any one of the first to sixth technical means, wherein the filter coefficient changing means is
The type of the filter processing and the edge pattern, and further, when having the determination result of the line portion, a filtering process change database that stores in advance filter coefficient change information according to the determination result of the line portion. It is characterized by having.

[0015]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention, as described above, shows the characteristics of an input digital signal input to a digital filter processing system when the product is subjected to a product-sum operation and a digital filter processing is performed. According to the edge pattern and the content of the filtering process (rate conversion process or the like), the filter coefficient is adaptively changed to perform an appropriate filtering process. An embodiment of a digital filter processing system according to the present invention will be described below with reference to the drawings. Note that the digital processing system according to the present invention is not limited to the case of image information, and can be applied to audio information in the same manner. However, in the following description, the case of image information is taken as an example in the embodiment. Will be described.

FIG. 1 is a functional block diagram of a digital filter processing system according to the present invention. Figure 1
In 100, a filter processing control unit 100 controls various calculation operations of the product-sum calculation unit 110, including designation of a filter coefficient. The product-sum operation unit 110 performs an operation with the input digital signal using an interpolation signal or the like as necessary according to the filter coefficient, performs signal conversion, and generates an output digital signal. An edge extraction unit 120 extracts edge information of the input digital signal. The edge extraction unit 120 has a luminance difference calculation unit 121 and an edge determination unit 122.

The luminance difference calculating section 121 calculates a difference relating to the luminance information of adjacent image signals before and after the input digital signal. The edge determining section 122 calculates the luminance difference value obtained by the luminance difference calculating section 121. By
Determine the type of edge. Reference numeral 130 denotes an edge pattern extraction unit which sequentially accumulates information on the types of edges from the edge determination unit 122 and generates an arbitrary number of combinations of edges (hereinafter referred to as edge patterns). And an edge pattern database 132 for storing the edge pattern.

Reference numeral 140 denotes a filter coefficient changing unit which determines a change of a filter coefficient used for digital filter processing based on the edge pattern extracted by the edge pattern extracting unit 130, and sends a filter coefficient change to the filter processing control unit 100. Instruct. The filter coefficient changing unit 140 includes a line determining unit 141 that determines a line portion based on the type of the edge pattern and the number of occurrences of the type, and information on the edge pattern for determining the line portion (that is, a plurality of extracted edges). A line determination database 14 for storing information on combinations of patterns and the number of times of extraction of the combinations)
4 and a filter storing change information of a filter coefficient according to the type of filter processing based on the edge pattern information obtained by the edge pattern generation unit 131 and the line part determination information obtained by the line determination unit 141. Referring to the processing change database 143, the filter processing control unit 1
Shift determination unit 14 instructing 00 to change the filter coefficient
And 2.

That is, the digital filter processing system according to the present invention includes a first step of extracting an edge pattern which is a characteristic of an input digital signal, a second step of determining a line portion from the edge pattern, and The third step is to change the filter coefficient according to the type of the filter processing based on the edge pattern information and the line portion determination information obtained in the steps and execute the digital filter processing. .

Next, the operation of each circuit part which is a component of the digital filter processing system according to the present invention, that is, each circuit part of the edge extracting unit 120, the edge pattern extracting unit 130 and the filter coefficient changing unit 140 in FIG. Will be described. First, a first step until an edge pattern is extracted from a digital input signal will be described with reference to FIG.

FIG. 2 is a diagram schematically showing an example of the magnitude of the luminance information of the input digital signal input in time series, and judges the edge information of two adjacent input digital signals. FIG.
In FIG. 2, a symbol IN indicates input digital signals Y ₀ , Y ₁ ,
Y ₂ , Y ₃ , Y ₄ , and Y ₅ are shown, and vertical solid lines with white circles indicate the respective luminance information values B of the input digital signal IN series Y ₀ , Y ₁ , Y ₂ , Y ₃ , Y ₄ , and Y _{5. 0} , B ₁ , B ₂ , B ₃ , B ₄ , and B ₅ are shown.

The symbol DB_iBefore and after adjacent to each other
Difference of luminance information value regarding input digital signal IN
5 shows a luminance difference value. Brightness difference value DB_iIs the input
Digital signal IN signal Y₀→ Y₁, Y₁→ Y_Two, Y_Two→
Y_Three, Y_Three→ Y_FourAnd Y_Four→ Y_FiveOf each luminance information value at
Difference, ie, each DB₀= B₁-B₀, DB₁= B
_Two-B₁, DB_Two= B_Three-B_Two, DB_Three= B_Four-B_ThreeAnd DB
_Four= B_Five-B_FourMeans Also, B_SIs the input digital signal
Brightness difference DB for determining the edge portion of signal IN_iThe threshold of
, Brightness difference value DB_iIs B_SIf the edge is over
Is determined to be a copy. For example, in FIG.
In the signal IN sequence, the signal Y₁→ Y_Two, Y_Two→ Y_Three, Y_Four→
Y_FiveBrightness difference value DB at₁= B_Two-B₁, D
B_Two= B _Three-B_Two, DB_Four= B_Five-B_FourIs the threshold B_SBeyond
Signal Y₀→ Y₁, Y_Three→ Y_FourBrightness difference at
Value DB₀= B₁-B₀, DB_Three= B_Four-B_ThreeIs the threshold B_SRange of
The case where it is within the box is shown.

As shown in FIG. 1, the input digital signal IN
Is input to the product-sum operation unit 110 to perform the filtering process, and is also input to the luminance difference operation unit 121 of the edge extraction unit 120 to determine whether the filter coefficient needs to be changed in the filtering process. You.

The luminance difference calculating section 121 receives the input digital signal.
Information value of two signals before and after adjacent in signal IN
Difference value, that is, luminance difference value DB_iIs calculated. Sand
And DB of FIG._iAs shown in the figure, the input digital signal sequence
Y₀, Y₁, Y_Two, Y_Three, Y_Four, Y_FiveBrightness for each of
Difference value DB₀= B₁-B₀, DB₁= B_Two-B ₁, DB_Two= B_Three
-B_Two, DB_Three= B_Four-B_ThreeAnd DB_Four= B_Five-B_FourOperation of
And input to the edge determination unit 122.

The edge determining section 122 receives the input digital signal I
The type of the edge portion of the input digital signal is determined by comparing the brightness difference value DB _i for the N series with a preset threshold value B _S. For example, if the luminance difference value DB _i has increased beyond the threshold value B _S , a code “1” indicating “+” is used as the edge determination value, and conversely, if the luminance difference value DB _i exceeds the threshold value B _S , If decreased "-" to code "2" indicating the edge determination value, and if the luminance difference value DB _i is increased or decreased within the range of the threshold B _S code indicating "no change"" 0 ”is set as an edge determination value. For example, in the input digital signal IN sequence shown in FIG. _{2, -B S ≦ DB 0 ≦} B S, DB 1> B S, -B S> DB 2,
−B _S ≦ DB ₃ ≦ B _S , −B _S > DB ₄
The edge judgment values are “0”, “1”, “2”, “0”,
It becomes “2”. Here, "1" as the edge determination value of the luminance difference value DB _i (i.e. "+") and "2" (or "-") are you distinguish between either edge portion is up, down and This is because the applied filter processing differs depending on whether or not the filter processing is performed.

The edge judgment values are sequentially input to the edge pattern generation unit 131 of the edge pattern extraction unit 130.
The edge pattern generation unit 131 combines an arbitrary number of continuous edge determination values, extracts an edge pattern indicating a feature of the input digital signal, stores the extracted edge pattern in the edge pattern database 132, and changes the edge pattern with a filter coefficient. Output to the unit 140. For example, in FIG. 2, “01202” or “12
An edge pattern such as 02 "is generated.

Next, the second step of determining the line portion will be described. In the line determination unit 141 of the filter coefficient change unit 140, the edge pattern generation unit 13
By counting the type of edge pattern extracted in step 1 and the number of times of extraction of the type, it is determined whether or not the input digital signal sequence generates a line. That is, the edge patterns input from the edge pattern generation unit 131 are sequentially stored in the line determination database 144, and the types of the edge patterns included in the corresponding line and the number of the types of the edge patterns are sequentially counted. Accordingly, if the number of the existing parts is equal to or more than a predetermined threshold value, the line part is determined. For example, in a two-dimensional image signal or the like, it is determined to be an edge pattern for forming an image line. In the present invention, the line determination process may not be performed in some cases.

Next, the operation of the shift judging section 142 according to the present invention in the third step of changing the filter coefficient according to the type of the filter processing and performing the filter processing adaptively will be described. In the shift determining unit 142 of the filter coefficient changing unit 140, the edge pattern generating unit 1
The edge pattern input from 31 and the line generator 14
1. A filter processing change database 1 that stores in advance filter coefficient change information corresponding to the type of filter processing based on the line part determination result input from Step 1.
By referring to 43, a determination is made to "shift" (i.e., "shift") the temporal position of the interpolated pixel (that is, the interpolation signal to be subjected to the interpolation processing) during the filter processing.

For example, in the case of filter processing for performing image reduction processing as rate conversion processing, an output digital signal after filter processing subjected to interpolation processing is output at time intervals according to the reduction ratio. In the conventional digital filter processing, since the interpolation processing is performed without shifting the temporal position of the interpolation pixel (that is, the interpolation signal) at the time of the interpolation processing, the temporal position of the output digital signal is shifted. Is subjected to an interpolation process using the interpolated luminance information value as a linear function of the input digital signal positioned before and after the time of the interpolation pixel. The interpolation pixel means a virtual pixel in the original image information corresponding to each pixel of the image information after the rate conversion processing. Therefore, if the interpolation processing is used, the information of the edge portion (or the line portion) of the original image information is lost, and the sharpness of the original image is lost.

In the present invention, when performing the interpolation processing at the edge portion, the processing for "shifting" the temporal position of the interpolation processing time, that is, the temporal position of the interpolation pixel (the "shifting"processing; (Referred to as a shift process) to store the information of the edge portion of the original image information.

The shift determining unit 142 outputs to the filter processing control unit 100 the fact that an interpolation pixel subjected to such interpolation pixel shift processing is to be generated, so that the filter processing control unit 100 performs appropriate interpolation pixel shift processing. The corresponding filter coefficient is output to the product-sum operation unit 110. Therefore, in the product-sum operation unit 110, the interpolation processing of the input digital signal is performed using the interpolation pixel at the time when the interpolation pixel shift processing based on the designated filter coefficient is performed, and the output corresponding to the interpolation pixel is performed. A digital signal is output.
By performing such interpolation pixel shift processing, adaptive filter processing that preserves the characteristics of the input digital signal according to the content of the filter processing is realized.

Next, a processing procedure for sharpening image information by the above-described interpolation pixel shifting processing will be described in more detail with reference to FIG. As in the conventional digital filter processing, simply performing rate conversion processing of a rational number may cause the loss of important information indicating the edge portion of the original image information, as described above. Unevenness may occur. The present invention saves important information of the edge portion of the original image information by performing a temporal position shifting process of the interpolation pixel, that is, an interpolation pixel shifting process at the edge portion of the original image information when performing the interpolation process. In addition, sharpening of image information is achieved to achieve high image quality.

FIG. 3 shows the principle of the above-described interpolation pixel shifting process in the case of the image information reduction process (ie, the process of thinning out the input digital signal). The temporal arrangement relationship of the interpolation pixels in the interpolation process is shown. It shows. In FIG. 3, a symbol IN indicates an input digital signal, that is, Y ₀ , Y ₁ , Y ₂ , Y ₃ , Y ₄ , Y ₅ . Symbol OUT
₁ is the interpolated pixel before the interpolated pixel shift processing, that is, y ₀ , y
₁ , y ₂ , y ₃ , and y ₄ are positions at equal intervals in time (arrangement at the same interval as the output digital signal after filtering). On the other hand, symbol OUT _2, the interpolation pixel of the processed shifting interpolation pixel i.e. _{y 0, y 1 ', y} 2', indicates y _3, y _4,
The time intervals of the respective interpolated pixels are arranged at different intervals due to the interpolated pixel shifting process.

The vertical solid lines with white circles represent the luminance information values B ₀ , B ₁ , B ₂ , and B ₃ of the input digital signal IN series Y ₀ , Y ₁ , Y ₂ , Y ₃ , Y ₄ , and Y _5, respectively. , B ₄ , and B _5, and a vertical dashed line with a black circle indicates the series of interpolation pixels OUT ₁ y ₀ , y ₁ , y ₂ , y ₃ , and y ₃ during the interpolation processing before the interpolation pixel shift processing.
It shows the luminance information values B ₀ , b ₁ , b ₂ , b ₃ , and B ₅ of each y ₄ . A vertical chain line with a penalty point (x) mark indicates an interpolation pixel OU in the interpolation processing after the interpolation pixel shift processing.
The luminance information values B ₀ , B ₁ , B ₂ , b ₃ , and B ₅ of the T ₂ series y ₀ , y ₁ ′, y ₂ ′, y ₃ , and y ₄ are shown. For example, as shown in FIG. 3, the input digital signal Y ₂ is a luminance information value B ₂
In an ordinary image reduction process, the output digital signal is converted into luminance information values b ₁ and b ₂ at two points of the interpolation pixels y ₁ and y ₂ , and the luminance information value B ₂ is lost, resulting in an image information blurred edge portion is not emphasized.

Therefore, in the present invention, when it is determined that the pixel is an edge portion, as shown by an arrow in FIG. 3, an interpolation pixel shifting process for shifting the position of the interpolation pixel (that is, the time position for performing the interpolation process) is performed. Thus, a measure is taken to approach the pixel position of the input digital signal of the original image information. For example, the temporal position of the interpolation pixel y ₂ in FIG. 3, the position of the input digital signal Y ₂ in the immediately preceding, i.e., in a position (FIG. 3 a transition amount is equivalent to 2 to the left, the transition amount s = 2), the luminance information value of the interpolated pixel y ₂ ′ at the time of the interpolation processing becomes b ₂ → B ₂ , and the input digital signal Y ₂ , which is the edge of the original image, it can be matched to the luminance information value B _2.

Similarly, if the interpolated pixel y ₁ is also subjected to the interpolated pixel shifting process by the transition amount s = 1 based on the edge pattern and the determination result of the line portion, the interpolated pixel at the time of the interpolating process is obtained. The sequences y ₁ and y ₂ are shifted to the positions of y ₁ ′ and y ₂ ′, and their luminance information values are changed to b ₁ → B ₁ , b ₂ → B _2, and the input digital signal at the edge of the original image Y ₁ , Y ₂
With the brightness information values B ₁ , B ₂ of. Therefore, when the interpolation pixel shifting process is performed, the interpolation pixel sequence (y ₀ , y ₁ ′, y ₂ ′, y ₃ , y ₄ ) at the time of the interpolation process is performed.
Is (B ₀ , B ₁ , B ₂ , b ₃ , B ₅ ).
The input digital signal sequence (Y ₀ , Y ₁ , Y ₂ , Y ₃ ,
Y ₄ , Y ₅ ) luminance information values (B ₀ , B ₁ , B ₂ , B ₃ , B ₄ ,
To B _5), can be a more faithful luminance information value and it is possible to input the digital signal of the sharpness is kept high-quality image of the edge portion.

The operation of the interpolation pixel shifting process will be described with reference to FIGS. Here, FIG.
6 show the operation of the interpolation pixel shifting process in the case where the image information is reduced in the rate conversion process, and FIGS. 7 and 8 show the operation of the interpolation pixel shifting process in the case where the image information is enlarged. Show. First, the operation of the interpolation pixel shifting process when performing the image information reduction process will be described with reference to FIGS.

In FIGS. 4 to 6, the symbol IN denotes an input digital signal, that is, Y ₀ , Y ₁ , Y _2i (Y ₂₀ , Y ₂₁ ,
Y _22), shows a _{Y 3, Y 4, Y 5} . Symbol OUT _1, the interpolation for no pixel shift process (or interpolation pixel shifting pretreatment) interpolated pixel or y at interpolation processing in the _0, y _{1,
y 2i (y 20, y 21} , y 22), shows a y _3, y _4. Further, symbol OUT _2, the interpolation pixel at the interpolation process after shifting the interpolation pixel processing i.e. y _0, y _1, y ₂₀ or y _2i '(y
_{21 ', y 22'),} y 3, y 4 shows a displaying by applying a (only interpolation pixels have shifted interpolation pixel dash mark ( ')). Here, the time interval of each interpolation pixel in the OUT ₁ is a layout of the same interval as the output digital signals, but are equally spaced, in the case of OUT ₂ becomes different time intervals by applying a shifting interpolation pixel processing.

The vertical solid line with a white circle indicates the input digital
Signal IN sequence Y₀, Y₁, Y_2i(Y₂₀, Y_{twenty one}, Y_{twenty two}), Y
_Three, Y_Four, Y_FiveEach luminance information value B₀, B₁, B_2i(B
₂₀, B_{twenty one}, B_{twenty two}), B_Three, B_Four, B_FiveIndicates a black circle
The vertical dash-dot line is the complement for interpolation processing before interpolation pixel shift processing.
Pixel OUT₁Series y₀, Y₁, Y_2i(Y₂₀, Y_{twenty one},
y_{Two Two}), Y_Three, Y_FourEach luminance information value B₀, B₁, B_2i
(B₂₀, B_{twenty one}, B_{twenty two}), B_Three, B_FiveIs shown. Also, penalty
The vertical dashed line with (x) mark indicates the complement after the interpolation pixel shift.
Interpolated pixel OUT during interim processing_TwoSeries y₀, Y₁, Y₂₀Or
y_2i′ (Y_{twenty one}', Y_{twenty two}') Y_Three, Y_FourEach brightness information
Value B₀, B₁, B₂₀Or B_{twenty one}(Y_{twenty one}') Or B
_Three(Y_{twenty two}′), B_Three, B_FiveIs shown.

FIG. 4 shows an example in which there is no interpolation pixel shifting processing. FIG. 5 shows an example in which the interpolation pixel shifting processing is performed in a temporally forward direction (that is, by a transition amount s = 2 in the left direction in the drawing). shows an example of a case where the luminance information values B ₂₁ of the input digital signal Y ₂₁ and luminance information value of the interpolation pixel y ₂₁ '. Also,
FIG. 6 shows that the interpolation pixel shifting process is performed in the backward direction with respect to time (that is, the transition amount s = −2 in the right direction in the drawing), and the luminance information value B ₃ of the input digital signal Y ₃ is converted to the interpolation pixel y ₂₂ ′. An example in the case of a luminance information value is shown.

When an edge portion of an image signal is detected,
Any of the above three processes (ie, no interpolation pixel shifting process, leftward shifting, rightward shifting)
Is determined by the shift determination unit 142 with reference to the filter coefficient change information stored in the filter process change database 143 based on the type of the filter process, the edge pattern, and the determination result of the line portion. You.

FIGS. 4 to 6 show Y _2i (i) indicating the edge portion or the vicinity of the edge in each signal of the input digital signal sequence.
= 0, 1, 2) are different only in the luminance information value B _2i (i = 0, 1, 2). FIG. 4 shows the shift determination unit 1.
42 is a case where it is determined that the interpolation pixel shifting process is unnecessary, and FIG. 5 is a case where it is determined that an interpolation process for storing the luminance information value B ₂₁ of the input digital signal Y ₂₁ as an edge portion is performed. FIG. 6 shows a case where it is determined that an interpolation process for preserving the luminance information value B ₃ of the input digital signal Y ₃ as an edge portion is performed.

It is to be noted that the case where the interpolation pixel shift processing is performed only on the interpolation pixel y _2i related to the Y _2i signal is shown, but as described above, the other interpolation pixels (y ₁ , y ₃₎ near the edge portion are processed. And the like), the interpolation pixel shifting process may be performed at the same time. Further, the relationship between the luminance information value of the input digital signal sequence and the interpolation pixel shifting process shown in FIGS. 4 to 6 shows a concept for understanding the present invention,
The present invention is not limited to the examples shown in FIGS. That is, the necessity of the interpolation pixel shift processing and the interpolation pixel shift amount, that is, the transition amount s (direction, value) are determined based on the type of the filter processing, the edge pattern, and the determination result of the line portion, as described above. Processing change database 1
43 is determined by the shift determination unit 142.

Next, a method of calculating the value of the transition amount s in the interpolation pixel shifting process will be described. 4 to 6
Shows the case of image reduction processing, and the reduction ratio at the time of image reduction is given by M / m. Here, M indicates an interval (distance) between adjacent signals (original pixels) of the input digital signal (original pixel), and m indicates an interval (distance) between adjacent interpolated pixels of the interpolated pixel before the interpolation pixel shifting process. . In the examples of FIGS. 4 to 6, M = 4 and m = 5, and the reduction ratio is 4/5 =
It will be 0.8.

On the other hand, a certain interpolation pixel y_jFrom the interpolation pixel y_j
Next to (i.e., the position immediately before)
Force digital signal Y_kThe distance (distance) to (original pixel) is w
When expressed as "eight", the value of the weight is the interpolation image.
Interpolated pixel y before unshift processing _jIs expressed by the following equation (1).
Is done. weight = mod (m × j / M) (1) where j is an interpolation pixel y_jAnd mod is the sum
Shows the operator for finding the remainder in an operation. Therefore,
For example, the interpolation pixel y shown in FIGS._2iOn the immediate left of
Power digital signal (original pixel) Y_2iDistance (distance) to
Since the number j of the interpolation pixel is 2,
From (1), weight = mod (m × j / M) = mod (5 × 2
/ 4) = 2.

Therefore, a shift process is performed on the position in the left direction (that is, the temporally forward direction) by the transition amount s of the amount indicated by weight, and if the interpolation pixel y _j ′ is set, the interpolation pixel y _j is immediately adjacent to the left of the interpolation pixel y _j . Can be shifted to the position of the input digital signal _{Yk at} the position of.

First, there is a change in the luminance information value shown in FIG.
Rate conversion processing of image reduction processing for input digital signals
(That is, the input digital signal Y_{twenty one}of
Brightness information value B_{twenty one}If you want to save), the edge pattern
Filter processing change data based on the
With reference to the database 143, the shift determination unit 142
And the interpolation pixel is shifted to the left (that is,
Is determined in advance. Therefore,
As described above, the output digital signal y_{twenty one}Is w to the left
The shift processing is performed only by the transition amount s indicated by the right, ie, by 2
Is performed. Interpolated pixel y during interpolation processing_{twenty one}Is the input digital
Tal signal Y_{twenty one}Interpolation pixel y at the same position as_{twenty one}'
As a result, the input digital signal Y_{twenty one}Luminance information value B_{twenty one}Is the interpolation image
Element y_{twenty one}The filter processing is performed as the luminance information value of '.
Therefore, the input digital signal Y _{twenty one}Brightness information at the edge of
Report value B_{twenty one}Is filtered in the saved form.

Meanwhile, saving the case (i.e., luminance information value B ₃ of the input digital signal Y _3, which rate conversion processing of the image reduction processing is applied to the input digital signal with a change in the luminance information values shown in FIG. 6 In this case, the shift determination unit 142 refers to the filtering process change database 143 based on the edge pattern and the determination result of the line portion.
In, shifting to the right of the interpolation pixel y ₂₂ (i.e., shifted to the rear direction temporally) a determination is made. Since the transition amount s when the right shift is given by (weight-M), the case of FIG. 6, the transition amount s of the interpolation pixel y ₂₂
Is s = weight−M = mod (m × j / M) −M = m
od (5 × 2/4) −4 = −2.

[0049] For interpolation pixel y _22, only the values shown in the transition amount s = -2, by performing the right direction, that shifting the interpolation pixels temporally shifted in the backward processing is immediately to the right of the interpolation pixel y ₂₂ Input position to the shift processing of the digital signal Y ₃ is performed, the luminance information value B ₃ of the input digital signal Y ₃ filter processing is performed as the luminance information value of the interpolation pixel y ₂₂ 'during the interpolation process. That is, the luminance information value B ₃ of the input digital signal Y ₃ indicating characteristics of the input digital signal sequence is stored, and, by using the luminance information value b ₃ close to the luminance information value B ₄ of the input digital signal Y _4, filter Processing is performed.

It should be noted that the example shown in FIG. 4 is not limited to the case where an edge portion has not been detected. Even if a luminance difference value corresponding to an edge portion of an input digital signal is extracted, the filtering process and the edge and a determination result of the pattern and the line section, upon reference to filter change database 143, a determination of the shift determining portion 142, there is no need to apply the even shift process on one of the interpolation pixel y _j is made The case is also shown.

As described above, by performing appropriate interpolation pixel shift processing according to the input digital signal sequence, the image after the rate conversion processing of image reduction can be performed without causing image blur and image unevenness. A sharp image in which the sharpness of the edge portion is preserved can be obtained. The conditions of the interpolation pixel shift processing are appropriately set based on the determination results of the edge pattern and the line portion. The patterns of the interpolation pixel shift processing are the above three patterns, that is, no shift, left shift, and right shift. Since the transition amount (s) at the time of each interpolation pixel shifting process at the edge portion can be automatically generated, the transition amount (direction, value) of the interpolation pixel shift is directly discarded by an edge pattern or the like. It is also possible to make a selection.

Next, with reference to FIG. 7 and FIG. 8, the operation of the interpolation pixel shifting process at the time of the interpolation process in the case of performing the image information enlargement process will be described. 7 and 8, symbol IN, OUT ₁ and OUT ₂ are similar to the case of FIGS. 4 to 6, respectively input digital signal series Y _i,
An interpolated pixel sequence y _i without interpolated pixel shifting and an interpolated pixel sequence y _j or y _j ′ after interpolated pixel shifting (a dash mark (′) is given only to the interpolated pixels with interpolated pixel shifting) are shown. In addition, a vertical solid line with a white circle, a vertical dashed line with a black circle, and a vertical dashed line with a penalty point (x) mark are shown in FIG.
6, the luminance information value B _{i of} the input digital signal sequence Y _i , the luminance information value b _j of the interpolated pixel y _{j in} the case without the interpolated pixel shift, and the interpolated pixel y _j after the interpolated pixel shift processing. shows 'luminance information value b _j of, b _j' y _j a.

FIGS. 7 and 8 show an example in which the enlargement ratio M / m during the image enlargement processing is 8/5.
Here, FIG. 7 shows the arrangement of the interpolation pixels when the interpolation pixel shift processing is not performed by the filter processing change database 143 based on the edge pattern and the determination result of the line portion. On the other hand, FIG. Based on the determination result of the section, the filtering process conversion database 1
As a result of referring to FIG. 43, the arrangement of the interpolated pixels after performing the interpolated pixel shift processing is shown.

As shown in FIG. 8, the shift judging section 142 judges that a rightward shift process is performed in order to store the luminance information value B ₂ of the input digital signal Y ₂ corresponding to the edge portion of the original image information. If done, the input digital signal Y ₂
Interpolation pixel y ₃ which is located immediately left neighbor of is subject to shifting the interpolation pixel. Transition amount s of the interpolation pixel y ₃ is s = weight-M = mod ( m × j / M) -M = m
od (5 × 3/8) −8 = −1.

Therefore, with respect to the output interpolation pixel y ₃ , interpolation pixel shifting processing is performed rightward, that is, temporally backward by the value indicated by the transition amount s = −1, to obtain an interpolation pixel y ₃ ′. Since pixel y ₃ 'is between該補the same position as the input digital signal Y _2, luminance information value of the input digital signal Y ₂ B
₂ is the luminance information value of the interpolation pixel y ₃ ′. Therefore,
Interpolation pixel shifting result of the processing, the luminance information value B ₂ of the edge portion of the input digital signal Y ₂ is stored, the filter processing is performed.

Further, to the right of the interpolation pixel at the edge portion
When the direction shift processing is performed, the correction near the edge part is performed.
If the same interpolated pixel shifting process is applied to the inter-pixel,
Image information near the edge is symmetrical with the original image information
Image that can be more faithful to the original image
Can be played. That is, near the edge portion
Interpolated pixels nearby (in FIG. 8, corresponding to the edge portion)
Interpolation pixel y_ThreeAnd both interpolated pixels y before and after_Twoand
y_Four) Also applies to the interpolation pixel y corresponding to the edge portion. _ThreeAgainst
Shift by the same amount as the transition amount (s = -1), and_Two'And
Y_Four'. As a result of the interpolation pixel shifting process, the interpolation image
Element y_Two→ y_Two'And y_Four→ y_Four′ No interpolation processing
And the luminance information value of the interpolation pixel is b
_Two→ b_Two'And b_Four→ b_Four′ To move around the edge
That is, the luminance information on the left and right is used as luminance information on symmetrical positions.
And the occurrence of image unevenness can be prevented. Also
However, if interpolation pixel shift processing is not performed,
Lines of several types of edge parts whose center point is shifted are reproduced images
Is displayed on the screen, resulting in a feeling of image unevenness.

[0057]

In the digital filter processing, the time position of the interpolation signal is adaptively adjusted when performing the rate conversion processing such as the pixel density conversion of the image information, the image enlargement / reduction processing, or the audio information compression / expansion processing. By performing interpolation pixel shift processing, important information such as the edge portion of the input digital signal is filtered without loss, so that the edges and contrast of the image and audio information after conversion by the filter processing are emphasized. The effect can be maintained, and for example, it is possible to prevent image blur, image unevenness, and the like of image information after digital filter processing.

Further, by having a database for edge patterns, a database for line determination, or a database for changing filter processing, the processing can be speeded up and the response to various filter processing changes and version upgrades can be expedited. Can be achieved.

[Brief description of the drawings]

FIG. 1 is a functional block diagram of a digital filter processing system according to the present invention.

FIG. 2 is a schematic diagram illustrating a concept of edge portion determination of an input digital signal according to the present invention.

FIG. 3 is a diagram illustrating a concept of an interpolation pixel shifting process according to the present invention.

FIG. 4 is an arrangement diagram of interpolated pixels in a case where no interpolated pixel shift processing is performed at the time of image information reduction processing according to the present invention.

FIG. 5 is an arrangement diagram of interpolated pixels when an interpolated pixel shift process is performed in a left direction during a reduction process of image information according to the present invention.

FIG. 6 is a layout diagram of interpolated pixels when an interpolated pixel shift process is performed in the right direction at the time of image information reduction processing according to the present invention.

FIG. 7 is an arrangement diagram of interpolated pixels in a case where no interpolated pixel shift processing is performed at the time of image information enlargement processing according to the present invention.

FIG. 8 is a layout diagram of interpolated pixels when an interpolated pixel shift process is performed in the right direction at the time of image information enlargement processing according to the present invention.

[Explanation of symbols]

100: filter processing control unit, 110: product-sum operation unit, 1
20: edge extraction unit, 121: luminance difference calculation unit, 122
... Edge determination unit, 130... Edge pattern extraction unit, 13
DESCRIPTION OF SYMBOLS 1 ... Edge pattern generation part, 132 ... Edge pattern database, 140 ... Filter coefficient change part, 141 ... Line determination part, 142 ... Shift determination part, 143 ... Filter processing change database, 144 ... Line determination database.

Claims (7)

[Claims] 1. A digital filter processing system having a filter processing means for an input digital signal,
Edge extracting means for determining and extracting an edge portion of the input digital signal; edge pattern extracting means for extracting an edge pattern based on a combination of the plurality of extracted edge portions; and And a filter coefficient changing unit for changing a filter coefficient of the filter processing unit. 2. The digital filter processing system according to claim 1, wherein the filter coefficient changing unit determines a position on a time axis of an interpolation signal used for filter processing based on the plurality of extracted edge patterns. A digital filter processing system comprising shift determination means for performing a shift determination. 3. The digital filter processing system according to claim 1, wherein the filter coefficient changing unit changes a line portion of the input digital signal in accordance with a type of the extracted edge pattern and the number of times of extraction of the type. Line determining means for determining, and shift determining means for performing a shift determination of a position on the time axis of the interpolation signal used for the filtering process based on the edge pattern and the determination result of the line portion. Digital filter processing system. 4. The digital filter processing system according to claim 1, wherein, when the digital filter processing system has the extracted edge pattern and the judgment result of the line part, the judgment is performed based on the judgment result of the line part. The filter coefficient converting means shifts the position on the time axis of the interpolation signal located immediately adjacent to the edge to the position on the time axis of the input digital signal having the extracted information on the edge. A digital filter processing system characterized by the above-mentioned. 5. The digital filter processing system according to claim 1, wherein said edge pattern extracting means has an edge pattern database for storing a combination of a plurality of edge portions. Processing system. 6. The digital filter processing system according to claim 3, wherein said line determination means has a line determination database for storing a type of said extracted edge pattern and said number of times of extraction. A digital filter processing system, characterized in that: 7. The digital filter processing system according to claim 1, wherein the filter coefficient changing unit has a type of filter processing, the edge pattern, and a determination result of the line portion. A digital filter processing system having a filter processing change database in which change information of a filter coefficient according to the determination result of the line portion is stored in advance.