JPS61230583A - Interpolation circuit for scanning line - Google Patents

Abstract

PURPOSE:To obtain a picture reproducing with smooth and strong sharpness by interpolating selectively from a signal low in resolution in the horizontal and vertical directions by utilizing the correlation of the signal, and synthesizing a signal with double information on appearance. CONSTITUTION:Luminance signals A(N) and A(N+1) of continuing two scanning lines are discriminated whether they are larger or smaller than a reference level at comparators 1 and 2 respectively. When both of the A(N) and the A(N+1) are smaller than the reference level, or when both of them are larger than that, an EXOR gate 3 takes out a low level signal. The luminance average of the two scanning lines is obtained by an average arithmetic unit 6. An edge detecting circuit 10 counts up a counter 9 at every detection of an edge. At an interpolation arithmetic circuit 14, an interpolation calculation is executed when the signals of the two scanning lines are existed separately at different areas. At multiplexers 18, 19 and 20, addresses given to line memories 15, 16 and 17 are switches. At an address counter 21, the address in the horizontal direction is counted and it is reset at every one horizontal scanning.

Description

[Detailed Description of the Invention] [Field of Application of the Invention] The present invention relates to a signal transmission system for displaying a high-definition display using fewer signals in a conventional information room.
3J1, which is useful for displaying images that are related to architecture, relates to the circuit between Takeo and Oke.

[Background of the invention]

As a method to make the term f, * several i times multiple, by interpolation.

As shown in Japanese Unexamined Patent Publication No. 58-79578, a method for calculating the average brightness between two running lines is described.

According to this method, it is possible to reduce the jagy that appears in the diagonal part and double the number of strokes with a relatively simple circuit, but it is difficult to recognize the relationship between the occurrence of intermediate brightness and the sharpness of the image. It had not been done.

[Purpose of the invention]

An object of the present invention is to provide a smooth and highly sharp 1m1Il signal conversion device for displaying high-definition television using a low-speed signal with a small amount of information. The object of the present invention is to provide a scanning line interpolation circuit that can be reproduced.

[Summary of the invention]

The feature of the present invention is that the edges of one image are marked.

Insect the 5-sized Bima-sashi *m* with an edge located midway between the edges on two adjacent running rocks.

If the two points at the same horizontal position on the upper and lower scanning lines are in the same area, the brightness is the average brightness of the two points, and if they are in different areas, the brightness is the average brightness of the two points at the same horizontal position on the upper and lower scanning lines. The maximum slope [[
8, and by setting the average brightness at each intersection with the vertical movement fM, it is possible to reproduce a smooth and highly sharp image.

[Embodiments of the invention]

Figure #&1 is a 1012 figure showing an example of one-English M according to the present invention.

The raw salt of the present invention will be explained below with reference to FIG.

A (IV) and A (N◆ are two consecutive scanning-like luminance signals (digital signals). 1.2 are the luminance signals of A (#) and A (N◆1) by comparators, respectively. Determine whether the value is greater than or less than the reference level.

Normally, the reference level is set to KO.

3 is Exclusi us-OR gate (ACN)
, A (#+1) When both signals are smaller than the reference level or both are larger than the reference level, an a-level signal is taken out. 4.5 is a 5-ztat- buffer.

6 is an average calculation Xa, that is, an adder 0. This calculates the average brightness of the two set meal maroons. 7 is a multiplexer, 8
is a buffer of 3-ztat-, 9 is a counter, and 10 is an edge detection circuit. Every time an edge is detected in the edge detection cycle wr10, the counter 9 increments the count by 1. 1
1 is a 5-5 tate buffer, and 12 is a line memory. The line memory 12 stores the edge type and edge position (horizontal address) at the address output from the counter 9. 15 is a latch % 14 is an interpolation calculation circuit. Scanning-sama's faith-41
Interpolation calculation is performed when jA CN) and A (#◆1) are in different areas. 15 and 16 are line memories, and two consecutive scanning signals A (N).

Store A (N◆1). 17 is a line memory that stores the interpolated scanning f-signal A CN''/2);18;
19゜20 is a multiplexer, line memory 15,
Switch the address given to 16°17. 21 is an address counter and the human input signal A (N), A C#+Sl
The horizontal address force is 9 times synchronous with IC and is reset every horizontal shot.

ACN) and Jtyφ1) are two consecutive scanning-like signals, which are formed from the original signal by the circuit shown in '#&2'.

In Fig. 2, the Ai path is an analog input signal, Din is a digital input signal, 22 is an Anami-Digital Kahong device, 2
5 is a delay circuit with a width of one dot period, 24° and 25 are comparators,
26 is an average calculation circuit (adder).

27 is Excltbs*ua -ORgu), 28.2
9 is a multiplexer, and 30 is a delay circuit with a width of one horizontal period.O First, if the original No. 11 is an analog signal, it is converted into a high-resolution digital signal by the analog-to-digital converter 22, and then passed through the multiplexer 29. The signal is output as is, and is also output through the delay circuit 50t for one horizontal period. -1 if 10,000 digital signals are input
An average yt calculating circuit 26 calculates the luminance average through a delay path 25 having a dot period width. And 43 of 1 dot edition by multiplexer 28! ! , the output of the circuit 23 and the output of the average m-nose path (Kanai 赫) 26 are switched and input to the multi-7 lexer 29.
1 dot of A signal) Divide the interval between Al1 into 2, and both lives are delayed by 1f! The output of the i path 25 is 0 or 1/2, which is the ratio I.
The R devices 24, 25 and the Exclusive-Q R gate 27 set the level of the 'fN waves across the bending circuit w & 23 to a certain base! 1!1111, when both are in the same area and when they are in the opposite area,
The output of the average arithmetic circuit (adder) 26 is selected in the Ha group, and the output of the Suitchi 12J path 25 is selected in the Ling group. The operations after multiplexer 29 are the same as when inputting analog *V.

These are interpolations in the horizontal direction, and in order to maintain the sharpness of the edge, in the edge city, 91, which is separated by one dot, is used as the Kikuma signal.

Otherwise, the average luminance of the dots of both waves is used as the interpolation signal.
There is.

Returning to m 1m, this tvA (n), (Is it possible to find the average of Ka and A(N+1)?
The signal is input to the data terminal of the line memory 17 via the buffer 8 of 7.5-ztat-.

At this time, comparators 1, 2, Exclusive-O
The circuit consisting of R gate 3 is J (N), A CM
If both φ1) are in the same area with respect to a certain reference value,
ct - outputs the level, and its output is line memory 1
It is input to the write enable terminal of No.7.

According to the above, the line memory 17 interpolates the average luminance at the same horizontal position above and below when the two luminances are in the same area, that is, when the luminances are close to each other with respect to the reference 1 Stored as a signal. In addition, in one line after these processes, the edge detection circuit 10 calculates A (N), A
(#φ1) Edges including K are detected including their orientation. Each time an edge is detected, the counter 9 counts up and its output is input to the address terminal of the line memory 12.

-1 The type of edge detected by the edge detection circuit 10 and the edge position (horizontal address space is 3 zta)
It is connected to the data terminal of the line memory 12 via the t- buffer 11, and is sequentially stored every time an edge is input.

All the above operations are completed within one horizontal period of the input signal. At this time, multiplexers 1s and 19°20 both select the output of address counter 21. Address number 21 is A CM), A (N+13KIW
It counts once and is reset at the beginning of the horizontal period. The output of the address counter 21 indicates the horizontal position on the screen.

After filling line memories 12, 15°IS and 17 for a long time, line memory 1
2, the address counter 9 sequentially generates addresses and reads edge information (type, horizontal address). In addition, Etsuji's I [gi4 is Mame-age.

! Another 1M downhill run and 2 runs f:? Indicates which side of #1 it belongs to.

The edge information is input to the interpolation calculation circuit 14 via the latch 13 as information on two adjacent edges. The interpolation calculation circuit 14 uses a combination of two types of edges to find the required address from the addresses input at the same time as the edge, supplies it to the line memories 15 and 1d through multiplexers 18 and 1, and reads out the luminance number m from each. Ding.

Further, the average luminance read from the line memories 15 and L6 is used as data, and the average of the addresses given to the line memories 15 and 16 is used as an address to perform convolution in the line memory 17. The brightness of a point on the scanning plane is determined by a straight line that passes through the point and intersects with the upper and lower scanning patterns, and the brightness of the two intersection points is in the same area with respect to the base *[.

If we take the average of the g degrees between the two intersecting points and use it as the brightness of the point on the interpolation scanning grid, we are performing nothing but a loop operation. This operation achieves both the sharpness of edges and the smoothness of diagonal lines.

The above operation is completed within the 2'lF horizontal period. Then, in the 311th horizontal period, the interpolation running signal is output from the line memory 17 in the order of addresses.

FIG. 3 is an explanatory diagram showing the generation of interpolated traces according to the present invention. ◎The circles represent dots, and the color inside represents the brightness region. Also, the number 1 section above the dot indicates the horizontal address. First, if the original signal is an analog signal, it is converted into a linear signal in units of 1 dot according to the required resolution.
In addition, in the case of digital signal input, in order to double the resolution, new dots (odd dots) are inserted in the middle of the dots (even dots) of the original signal in the horizontal direction. The brightness of this odd-numbered dot is set to be the average of the brightness before and after m when the dot after m is within the direction -w degree range, and when it is within the 1 opposite pancreatic summer vX range, it is set to be the same as the j4L degree of the previous dot. (in Fig. 3, the run f, f1MLcN) is the brightness A of the address M+1.
(N9M◆1) Also, in the running style L (N+1), address M+
This corresponds to the case where the brightness Et A (#+1, Af+5) of 5 is in the opposite brightness region. It is assumed that the # degree signal of the horizontal address y of the Z-targeted movement is represented by ACx, y). After completing these horizontal interpolations, perform vertical interpolation.

Similarly to the horizontal direction, if the luminances of two points at the same horizontal position on the two hill runs are within the same luminance region, then
The m-line averages are combined to form an interpolated signal for points at the same horizontal position.

However, if the brightness levels of the upper and lower foot sources are in opposite regions, tracing back to the point at the same horizontal position on the supplementary dark foot trace, and the intersection of the upper and lower legs; Considering an adult straight line with a slope of 5, the brightness of each intersection point is averaged and used as the Fukuma running ffi signal. The LI indicated by the broken line
[Group fN that divides the luminance into two for yen! 11 [0
, by resetting the memory 17 in advance, A (N+/2.N+4) A (Nφ'/2
It is possible to omit the processing for the part m''5).

Next, a specific example of Tanibe will be given and explained.

Figure m4 shows one specific example of the edge detection circuit 10.

31 and 32 are diagonals, 33 and 54 are one-dot delay circuits, and 35 and 56 are fja:albziug-
At the OR gate.

As a result, ashing with respect to the reference value of the luminance level of successive dots is detected for each of the two strokes f:. 57
, 5B, 59, and 40 are AND gates that contain A (N) K, respectively.

Mameagari Tsuji included in A (#), A (N+1)
Included in Bunkudari Edge, included in A (7v41)! Detect upstream edge. Further, the OR gate 41 detects all edges included in the two runs f-. OR gate 4
2, the two legs at the edge detection position;
A case where at least one of the A lines is on the high brightness side is detected.

Next, the fifth example shows a specific example of the arithmetic circuit 14.

In the figure, 43 is a latch, and 46 and 49 are multiplexers.

55, 54 are Crisella tuple counters, 56, 59
57 and 58 are multiplexers, and 60 is a 3-ztat- buffer. Again 1
5, 16.17 is the line memory indicated by 1-. To the latch 45, the erase @t (edge type and horizontal address) of the edges included in the two runs:f are sequentially input two by two. RU is included in A (fit), and FU is included in A (#)! The falling edge, R1) is a signal that detects the rising edge included in A (A'+1), FD is the signal that detects the falling edge included in A (A'+1), and AE and AE-1 are the horizontal edges of the edge. Represents one address less than the address and horizontal address.

Note that the number after the alphabet is used to distinguish between two edges that touch 1111, and 1 is closer to the left edge of the screen. HH, is 2 at address AE
Book run f: # This is a signal indicating that at least one of the signals A (#) and A (N+1) is on the high luminance side with respect to the reference value. Among the input signals, zyjif)*li4'1e
bYQ@R, U, FU, RD.

FD passes through latch 45 and then OR go) 44, 45
.

47 and 48, and the first edge's ! Signals R□, F, , which detect the rising edge, the falling edge of the first edge, the rising edge of the second edge, and the falling edge of the second edge.
R□F, are each synthesized. So KRs m's l
R1 *FaHAANDG) 50 and 51.52 respectively, detects when two adjacent edges are both rising edges, when they are both falling edges, or when the first edge is a falling edge and the next edge is a rising edge. The signals RR, FF, and FR are combined. On the other hand, the addresses AE1, 'AE@-1, AE!' are input to the latch 43 at the same time as the edge. , A E, -1
are respectively translated by multiplexers 46 and 49.

Here, I will explain by quoting Figure 3.・AEiAE1 in the figure
-1 and AEhA4 j are each N+2°N+1
．． Corresponds to N+6 and N+5. First, to obtain 4m of the interpolated signal of the left block of the edge boundary and where the two running f:feed signals are in different luminance areas, set AE, -1, AE, -1 as the edge address. Proper translation, 0≦, <AE
, -Δ The total number I of the whiskers is calculated. Here, the luminance signal at the horizontal address y of the X★-th running eye is A Cx, y), and the Z-eye and x + l
The luminance signal at the horizontal address y of the l-th match is represented by A (a+/2.y). To find the brightness of the interpolated signal on the right side of the edge boundary and where the two running 3f:M signals are in different brightness ranges, edge addresses Az, , AE are selected and calculated. In Figure 3, there is a sharp boundary and there is a strong correlation between the two scan outputs, so the interpolation scan f? It is possible to determine the Is signal. However, in actual shapes, the correlation is weak, and the boundaries are I! There are times when Ir is not enough.

An example is shown in FIG. The part indicated by Nikatsumaru in the figure is
Depending on whether emphasis is placed on the correlation of the white circles or the correlation of the 1M circles, it is possible to perform two types of interpolation according to the straight lines indicated by *m and 1, respectively. Therefore, when a pattern that is considered to have a relatively soft correlation, that is, a single double edge in the opposite direction is input, it is assumed that there is a correlation only on the high brightness side or the low #direction side, and interpolation is performed. Return to Figure 5. In this example, we emphasize the correlation on the high brightness side, and let's assume that the standard value of brightness is set to the black level. First, the output RR of the AND gate 50 is When the level is high, presettable counters 55 and 54K
AE, , A
E, is set to 0, and the addresses set to each are input as they are to the addition@56, and the average value is determined. On the other hand, the addresses set in input terminals 55 and 54K are distributed and input to line memories 15 and 16 by multiplexers 57 and 58, and are placed at the edge positions. The brightness of the two scanning circles is read.

(TektA (7v, m+2), A (#
+1. jf+, 6). ]These 2 runs fill
The brightness information on l is 27+1 true! ) 59, and the average thereof is written to the line memory 17 via a 35-state buffer 60. Next, when -ml- is finished,
The counter 53 counts up 1 by 2, and each time,
Convolution is performed in the line memory 17 along the same path as before. The addresses of the counter 53 are A, E,
fIfc is kicked until it reaches .

111, when the output FF of the AND gate 52 becomes high level, the 1 resetter 1 counters 53 and 54 receive AE, -
1, AE, -1 are set, and convolution is performed in the line memory f7 when RR goes high.

After that, the counter 54 counts down by two, and convolution is performed in the line memory 17 as before. and.

This process continues until the output of the power inputter 54 becomes smaller than AE, -1. In Figure 3, A(A'"'/2 J" 2
) *A (7y+!/1+3 to 2.1) is the interpolation gI number accumulated by these operations.

Next, when the output PR of the AND gate 51 becomes high level, at least one of the two scanning signals at the address AE is on the high brightness side.

In other words, when the Hjl and tlc signals are high, the counters 55, 5411CAE, 5411CAE, A
When E is input and RR becomes high level, convolution is performed in the line memory 17. When this operation is completed, the multiplexers 46 and 49 are switched to output A4 1 to the input terminals 55 and 54.

Input A E, -1 and perform convolution with the colleague when FF becomes high level. Figure 7 shows the convolver of interpolation for these three cases. 11%.

The brightness of the two dots at the edge is averaged in darkness, and the n of the center dot is
Indicates that the degree of

Through this series of moat works, the construction of signal areas and non-signal areas,
To make the background clear while maintaining the sharpness of the image η)
oJ-becomes the rudder. Regarding the change of 14 skin within the Chigo habitual area, it is ignored here because it is less noticeable compared to the Mushinku habitual area. If it is not possible to connect consecutively, instead of detecting an edge by changing the luminance chain acid with respect to the quasi-electrode, detect the case where the straight run exceeds a straight run or F1 as an edge.
6Jt! ! 1! It is necessary to do some work.

Therefore, the eighth figure shows the time relationship between the source 41g and the output signal at each position in the entire system of the present invention. 61 is a time pressure device.

t-represented, the original signal 5 inches is time-compressed (more than twice) and its output SP is input to the circuit wr62 of the present invention.

The output of the circuit 62 of the present invention is temporarily stored in a 7-frame memory 63 if necessary, and input to a 7-frame memory 64. Next, the time relationship between each signal will be described.

@Figure 8 (in the case of 2J, the original signal of 5 inches is compressed twice by the 1-hour compressor 61, and C3P is output in 2 runs f: sparse division once). This output SP is the (2) path 6 shown in the present invention.
Entered in 2, 5 places! 1 process ■■■ (convolution■, W
The interpolation signal is output in the process (2) after the R output process (2) is completed. As you can see from Yuu, the time required for processing is 3 horizontal cycles, and it takes several runs! It is impossible to slow down the process by delaying the process.It is possible to achieve this by providing two circuits of the present invention and processing each other at t5. Provide two 17
When one is being read (process ①), if the other is enabled for input, the time relationship as shown in FIG. 8(3) will be established, and interpolation processing can be performed continuously.

Furthermore, as shown in FIG. 8 (4), it is also possible to perform continuous interpolation processing by compressing the time of the original signal, 5LW, by a factor of 4 using the 1-hour compression @61. The circuit of the present invention must be operated with a 2i bias.

In the embodiments mentioned above, only the interpolation calculation using software has been dealt with.

Use frame memory if real-time processing is not required.

It is possible to process Iryl images using the software 9Air using a central processing unit.

[Efficacy of invention]

According to the present invention, from a high-resolution signal, interpolation is selectively performed in the horizontal and vertical directions using the correlation, and the signal is divided into two signals in both the horizontal and W directions. It is possible to synthesize, and by taking the vertical average along the edge, there is a young fruit that can make the smoothness of the diagonal line the same without tracing the cape device of the edge.

[Brief explanation of drawings]

Figure 1 is a diagram 12 showing one embodiment of the present invention, and Figure 2 is a diagram 7' showing a circuit that generates the input shown in Figure 1 from an original signal.
3 is an explanatory diagram showing the principle of interpolation, FIG. 4 is a circuit diagram showing a specific example of the edge detection circuit 10, and FIG. 5 is a circuit diagram showing a specific example of the interpolation calculation circuit 14. Figure 6 shows the interpolation situation for patterns with weak correlation! l!
7 is an explanatory diagram showing the state of interpolation by a specific example of the interpolation calculation circuit 14 (FIG. 5), and FIG. 8 is an explanatory diagram showing the entire interpolation system including the present invention and input/output time F This is an explanatory diagram showing: 1.2...Comparators 4, 5, 8.11...5 ztat-buffer 76
...Adder (average calculation circuit) 7.18.19.20...Multiplexer 9.21.
...Counter 1G-...Edge detection circuit Figure 1 Figure 2 Figure 3 Figure 4 M M+I M+Z M+3M++ Between 5
MtG M+'7 Dol+8 Mshi'7A
E+ Figure 8 (α)

Claims (1)

[Claims] In a scanning line interpolation circuit that inserts an intermediate scanning line between two scanning lines of a television signal, etc., the luminance of two adjacent scanning lines at the same horizontal position is detected and both are within the same luminance area. is the interpolated signal at the same horizontal position by averaging the brightness, and if they are in different areas, the brightness of the interpolated signal at the same horizontal position is calculated as the brightness at the intersection of the upper and lower scanning lines passing through that point. 1. A scanning line interpolation circuit characterized in that a straight line with the maximum slope that lies within the same brightness area is determined, and the average of the brightness of two intersection points along the straight line is used as an interpolation signal.