JPS63311812A - Signal processing method - Google Patents

Abstract

PURPOSE:To attain high speed processing with a simple device by executing simultaneously an interpolation processing and a filtering processing efficiently so as to save remarkably the quantity of arithmetic operation and the arithmetic time. CONSTITUTION:Equation 1 relating to the interpolation processing with respect to the input sampling signal is substituted into equation II relating to the filtering processing, and the equation of the result of substitution is used to obtain an interpolation signal and the filtering processing signal with respect to the input sampling signal simultaneously, where f1-f3 are input signal values, g1-g3 are interpolation signals and a1-a3 are weight coefficients of the input signal 1 in equation I, and b1-b3 are weight coefficients, g11, g21, g31 are signals as the result of the filtering processing and f21 is the input signal 1 in equation II. Thus, number of times of arithmetic operation is saved remarkably or the arithmetic operation speed is quickened.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a signal processing method, and particularly to a signal processing method that efficiently performs both interpolation processing and filtering processing of human signals.

(Prior Art) Generally, as a method for processing a signal, such as an image signal, a signal that is not obtained from a sampling signal obtained as a human signal, that is, a signal between two sampling signals, is estimated, and the estimated signal is used as the Interpolation processing inserted between two sampling signals is particularly important in image processing technology.

For example, when used for image signals, the number of pixels can be increased, which has the advantage of improving image quality.

Additionally, filtering processing is often performed on the obtained signal. According to the filtering process, input signal smoothing, edge emphasis, edge extraction, etc. can be easily performed.

However, conventionally, both the interpolation processing and filtering processing require complicated and time-consuming calculations, and performing interpolation processing on the original signal and then filtering processing on that signal requires a particularly large amount of calculations. It took a long time to process.

For example, a simple example of the conventional interpolation processing and filtering processing operations will be described below.

FIG. 6 is an explanatory diagram showing conventional signal interpolation processing. In the figure, 1 is an input signal, which indicates a sampled signal. The input signal 1 is obtained at equal intervals, and its values are fl, f2 . f3...etc. 2 is an interpolation signal, which is created based on the input signal 1, and its values are gl, g2 . g
3...etc. The interpolated signals 2 are generally set at equal intervals between the human input signals 1. It is assumed that the interpolated signal 2 is determined by the weighted average of the input signals 1 on both sides thereof. Therefore, the values of interpolation signal 2 gI, g2 . g'3 can be expressed by the following equation.

However, al, a2. a3. a4. a5. a5 is a weighting coefficient of input signal 1.

and the value g4 of interpolation signal 2. g5. Similarly, for g6, the value f2. f3 and Fuuki gl, g2.
It is calculated in the same way using the same coefficients used in g3.

Filtering processing is performed on the signal obtained as described above. For simplicity, assume that it is a three-point convolution, let its weighting coefficients be l, b2゜b3, and let the values of the signal after filtering be gll and g, respectively.
21. If g31 and the value of input signal 1 are f21, they are expressed by the following equation.

gll in formula (2), g21. To find g31 and f21, gl, g2. g3. Adding the calculation of g4 gives 2
This requires 0 multiplications and 12 additions. , Therefore, it can be seen that there is a problem that the amount of calculation is large as shown in section 11η.

(Problems to be Solved by the Invention) An object of the present invention is to provide a simple signal processing method that can perform signal interpolation processing and filtering processing efficiently and at high speed.

(Means for solving the problem) Substitute the equation related to interpolation processing for the human sampling signal into the equation related to filtering processing, and use the equation obtained from this to simultaneously obtain the interpolation signal and filtering processing signal for the human sampling signal. This is what I did.

(Embodiment) FIG. 1 is a block diagram showing a signal processing method of the present invention. The present invention substitutes a formula representing an interpolation signal for a human sampling signal, for example, formula (1) shown in the prior art, into a filtered formula, such as formula (2), and directly obtains the final result from the formula. This is to obtain a signal through filtering processing.

In the figure, step (a) is the interpolation (8
Step (b)
In step (e), the value gl of the interpolated signal of the filter link processing equation (2) obtained in step (b) is calculated using the value of the interpolation processing equation (1) to obtain equation (2). , g2. g3. g4 is the interpolation process formula (
1) is substituted. From this, the following formula can be obtained.

gll = blfl+b2(alfl+a2f2
)+b3(a3fl+a4f2)g21-bl(
alfl+a2f2)+b3(a3fl+a4f2)+
b3 (a5f 1+a6 f2) 331-bl(a:lfl+a4f2)+b2(a5
fl+a6f2)+b3f2f21=bl(a
5fl+a6f2)+b2f2+b3(alf2+a2
f3) If we organize this, we get the following.

Then, in step (d), equation (3) is solved to obtain the final result for the sampling signals fl, f2. The coefficients related to f3 only need to be calculated once at the beginning, and the same coefficients can be used repeatedly thereafter.

Therefore, according to this embodiment, if the effort of calculating the coefficient is ignored, the final result can be obtained by 9 multiplications and 5 additions, so the number of calculations can be significantly reduced and the calculation speed can be increased. .

Next, an example using actual numerical examples will be shown. The cubic convolution method is used as the interpolation method. In the Cupic convolution method, the following interpolation function 1 (x)
is used.

. . . -(3a) FIG. 2 is a graph using the interpolation function 1(x). In the figure, 11 is an interpolation function I (x).

FIG. 3 is an explanatory diagram showing interpolation processing using the cubic convolution method in this embodiment. In the figure, reference numeral 21 indicates a human input signal sampled at equal intervals. The value of the input signal 21 is old, h2. -, h5, etc. 22 is an interpolation function whose basic form is shown in FIG. 2, and whose input signal 21 is the maximum value. 23 is an interpolation signal;
It is applied to each position divided at equal intervals between two human input signals, and its value is kl, 2. -, it is 6th magnitude. however,
Only a portion of the interpolated signal 23 is shown for simplicity.

2 to the value kl of the interpolation signal 23. 3 is expressed by the following formula.

kl -1 (5/4) old +1 (+/4) h2+1 (
-3/4) h3◆1 (-774) h4--(9/
64) Old + (57/64) h2+ (19764) h
3- (3/64) h4 to 2- T (3/2) old ÷ I
(1/2) h2÷1 (-1/2) h3+T (-
3/2) h4--(+/8) old + (578) h2+
(5/8) h3- (178) 3-1 (7/
4) hl+1 (3/4) h2+1 (-174) h3+1
(-574) h4--(3/64) old + (19/64)
h2+ (57/64) h3- (9/64) h4
・・・・・・・・・・・・(4) Also, among the interpolation signals 23, the remaining 4. 5. 6, the human power signal 21 h2. h3. h4. 2 to kl of the interpolated signal 23 using h5. : can be determined using the same coefficients used to calculate l. For example, K4
can be expressed by the following formula.

K4--(9/64) h2÷(57764) h3+
(19/64) h4- (3/64) h5 Other interpolation signals may be obtained in the same manner.

Next, as an example of filtering processing, a case will be described in which, for example, second-order differentiation is executed to extract edges. It is performed using a 3-point convolution, and its weighting coefficients are -1, +2. −
It is 1. That is, interpolated signal 2 after filtering processing is performed.
The value of 3 and the human input signal 21 are respectively kll and 21. to 31
' and h31, they become as follows.

Here, as mentioned above, the value kl of the interpolation signal 23 in equation (5) is 2. By substituting equation (4) into 3, the following equation can be obtained.

K11--(5/32)hl+(5/32)h2-(1
/32)h3+(1/32)h4 has 21−-(171B
) Old + (1/16) h2+ (1/16) h3- (
1/16) h4 to 31- (1/32) old-(1/
32) h2◆(5/32) h3- (5/:32)
h4h31 = (3/64)hl-(5/32)
h2+(7/32)h3-(5/32)h4+(3/6
4) h5 By performing calculations using equation (6), it is possible to obtain both interpolated and filtered signals from the input signal, but the amount of calculations exceeds that of ordinary interpolation processing other than this cubic convolution method. For example, (3)
Compared to the formula, there is almost no difference except that one term has been added.

Although the above embodiments were related to one-dimensional signals,
Expansion to two dimensions can be done relatively easily.

For example, in the cubic convolution method, interpolation functions are used as I (x) and I (y), and in filtering processing, a two-dimensional weighting function is used, so the same procedure as described above can be performed in two dimensions.

Also, regarding the amount of calculation in the two-dimensional signal processing, the cubic convolution method uses 16 input signal values, whereas in the one-dimensional case, interpolation is performed using four input signal values. The weighting function most often used in filtering processing has a size of 3×3 9 pixels. Therefore, in the conventional method that does not use the calculation method of the present invention, the two
The amount of calculation required for dimensional signal processing is enormous, and the use of the calculation method of the present invention, including the cubic convolution method, is effective and can significantly reduce the amount of calculation.

The interpolation processing described above can be used not only for signal densification but also for coordinate transformation.

FIG. 4 is a schematic diagram of one embodiment showing coordinate transformation. In the figure, 31 represents the old coordinate system, and 32 represents the old coordinate system 31.
This is the signal at 33 is a new coordinate system, and 34 is a signal in the new coordinate system 33.

The old coordinate system 31 has been converted into a new coordinate system 33 according to certain rules. At the intersection of the new coordinate system 33, a signal 34 in the new coordinate system is required, but this is not normally obtained and must be obtained by interpolation processing from the signal 32 in the old coordinate system 31. When performing filtering processing on the signal 34, if the results of the interpolation and filtering processing are directly calculated using the signal 32 in the old coordinate system according to the calculation method of the present invention, the amount of calculation can be reduced and the calculation time can be shortened. Can be done.

However, the above method is effective for signal 3 in the new coordinate system.
4, when multiple items have the same positional relationship or a relatively simple relationship with the signal 32 in the old coordinate system, and the relationship between the two can be classified into several types, the new coordinate system Processing is required, such as dividing the signal 34 into groups and applying a separate weighting function to each group.

The signal processing or arithmetic methods of the embodiments described above are usually digitally processed using an electric circuit or computer, or may be processed using an optical device.

Image processing using light has already been proposed by the present applicant, for example, in Japanese Patent Laid-Open No. 191165/1983, and the signal processing method of the present invention can be carried out using the image processing device proposed therein. Can be done.

FIG. 5 is a schematic perspective view of an image processing apparatus using light disclosed in the publication. In the figure, numeral 38 represents an input image, which may be an original, film, photographic paper, or the like. 39 is an imaging lens, 41 is a charge transfer type solid-state image sensor (image sensor), and pixels are arranged in two-dimensional directions of x and y.
Charges can be transferred in the x and y directions in accordance with the clock from the solid-state image sensor driver 42. 44 is an electric signal, 45 is an amplifier, and 43 is a controller, and charge transfer and information sweeping of the solid-state image sensor driver 42 are controlled by the controller 43.

The human image 38 is captured by a solid-state image sensor 41 using an imaging lens 39.
imaged on top. The image information read by the solid-state image sensor 41 is extracted as an electrical signal 44, amplified by an amplifier 45, and then sent to an appropriate processing circuit at the next stage.

In the image input device as described above, the input image 38 is exposed to the solid-state image sensor 41, the induced charge is transferred for one pixel, and further exposure is performed. When this operation is repeated, the input image 38 that has undergone the convolution process is displayed on the solid-state image sensor 41.
A charge image of .

If this is extracted as an electrical signal 44, a convolution output can be obtained.

Furthermore, the weight of the convolution weighting function can be controlled by the timing of charge transfer, intensity modulation of illumination light, etc.

If further negative weights are required, for example, as disclosed in Japanese Patent Application No. 60-220291 filed by the present applicant, two solid-state image sensors are installed, and polarized light perpendicular to the positive and negative weights is transmitted. It may be used as imaging light.

When performing interpolation processing and filtering processing with the image processing apparatus shown in FIG. 5, for example, in the case of interpolation processing using cubic convolution and filtering processing using quadratic differentiation, the processing is performed as follows.

With reference to FIG. 3, the values of the interpolated signal 23 and the human input signal 21 that have been filtered using equation (5) are determined by 11. 21. 31. h31 is the value hl of the input signal 21,
h2. h3. It is the average of the river mitts of h4 and h5 added to it.

First, when we realize the weighting coefficients used to calculate kll and perform convolution on the human image, kll
is obtained. At this time, at the same time, a series of filtering processed signals corresponding to the interpolation position of 1 and the value of the interpolation signal 23 such as 41 are simultaneously obtained in parallel over the entire screen. These signals are read out as first electrical signals. Next, by changing the weighting coefficients, 21 series of interpolated and filtered signals are obtained, and the signals read from the solid-state image sensor 41 are used as the second electrical signals. Similarly, 31 series, h3
The signals of one series (see equation (5)) are converted to the third. It is obtained as a fourth electrical signal. Finally, by arranging the first to fourth electrical signals in the signal order shown in FIG. 3, a desired output signal can be obtained.

(Effects of the Invention) According to the present invention, by efficiently performing interpolation processing and filtering processing, which conventionally required a large amount of calculation amount and time, at the same time, the amount of calculation and calculation time can be significantly reduced. A signal processing method that can perform high-speed processing with a simple signal processing device can be achieved.

[Brief explanation of drawings]

FIG. 1 is a block diagram of an embodiment of the method of the present invention, FIG. 2 is a graph showing the interpolation function of the cubic convolution method, FIG. 3 is a graph showing the interpolation function of the cubic convolution method, and FIG. An explanatory diagram showing coordinate transformation, and FIG. 5 is a schematic diagram showing an image processing device using light. FIG. 6 is an explanatory diagram showing conventional signal interpolation processing. In the figure, 21 is a human input signal, 22 is an interpolation function, 23 is an interpolation signal, 32 is a signal in the old coordinate system, 34 is a signal in the new coordinate system, 3
8 is an input image, 41 is a solid-state image sensor, 42 is a solid-state image sensor driver, 43 is a controller, and 45 is an amplifier. Patent applicant: Canon Co., Ltd. Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 [¥] Figure 6

Claims (1)

[Claims] Signal processing characterized in that an equation related to interpolation processing for an input sampling signal is substituted into an equation related to filtering processing, and the resulting equation is used to simultaneously obtain an interpolated signal and a filtered signal for the input sampling signal. Method.