JP2004187203A - Signal processing circuit and signal processing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve a problem wherein a subtracted value resulting from directly subtracting a signal level of a certain pixel and a signal level of a pixel just after the above pixel becomes a total value of a crosstalk component and a random noise differential when random noise is superimposed on both signal levels, to thereby increase the noise. <P>SOLUTION: In the signal processing circuit for processing a signal having cyclicity resulting from sampling/holding and then digitizing an output signal of a color solid-state imaging device, a subtractor 13 subtracts an output signal B of a low-pass filter 11 with a position of a signal inputted after a signal A of a concerned pixel for one pixel as centroid from a signal level of the signal A, the result of multiplying a positive correction coefficient by a subtracted output (A-B) in a multiplier 14 is defined as a correction quantity (correction value), the correction quantity is added to the signal A by an adder 15 and outputted, such that without increasing the noise, crosstalk caused by sampling/holding is corrected. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a processing circuit and a processing method for a signal having a periodicity that has passed through a sample-and-hold circuit, and in particular, samples and holds an output signal of a color solid-state imaging device and then processes a signal having a periodicity obtained by digitization. The present invention relates to a signal processing circuit and a signal processing method.
[0002]
[Prior art]
In a system using a solid-state imaging device such as a color CCD (Charge Coupled Device) type imaging device, as shown in FIG. 6, an output signal of the solid-state imaging device 101 is usually converted into a CDS (Correlated Double Sampling). The reset noise is reduced by passing through the circuit 102, and the waveform is shaped. Thereafter, the signal is amplified at an appropriate amplification factor by the amplifier circuit 103 so that the value becomes an appropriate value.
[0003]
The waveform of the amplified signal is destroyed by passing through the amplifier circuit 103, so the waveform is shaped again by the sample / hold (S / H) circuit 104, and then digitized by the A / D converter 105 and the DSP ( Digital Signal Processor (Digital Signal Processing) circuit 106. In this DSP 106 circuit, various signal processes are performed digitally.
[0004]
In general, when the output signal of the color CCD solid-state imaging device 101 passes through the sample-and-hold circuit 104, the signal of two consecutive pixels in the sample-and-hold circuit 104 interferes with each other, and the signal level differs from the original signal level. There is. FIG. 7 shows an example of a circuit configuration of the sample-and-hold circuit 104 that causes the above.
[0005]
In FIG. 7, an input signal is applied to the gate of MOS transistor Q2 via MOS transistor Q1. MOS transistor Q1 is turned on when clock Clk is applied to its gate. The drain of the MOS transistor Q2 is connected to the power supply Vdd, and the source is connected to the ground Gnd via the resistor R1. The capacitor C1 is connected between the gate of the MOS transistor Q2 and the ground.
[0006]
In the sample / hold circuit 104 having the above configuration, if the signals of two pixels input to the circuit are signal 1 / signal 2 in the order of time, the signal 1 is input, and the signal 1 is input via the MOS transistor Q1 in the ON state. It is applied to the gate of MOS transistor Q2. Then, after the gate voltage of the MOS transistor Q2 reaches the signal level of the signal 1, the clock Clk goes low, so that the MOS transistor Q1 is turned off.
[0007]
The signal 2 is input after the signal 1. At this time, the signal levels before and after the MOS transistor Q1 are different. That is, although not shown in FIG. 7, since the parasitic capacitance exists between the source and the drain of the MOS transistor Q1, the gate voltage of the MOS transistor Q2 is slightly changed by the input of the signal 2, and the actual signal is changed. The value is different from the level. This change is substantially proportional to the level difference between signal 1 and signal 2. This phenomenon is called crosstalk (or color mixing).
[0008]
FIG. 8 is a waveform diagram showing the crosstalk phenomenon. In the figure, a solid line is an input signal, and a signal changed by crosstalk is shown by a dotted line. Δ1 and Δ2 are the level differences of the input signals for each signal, and the product of Δ1 and Δ2 multiplied by the color mixture ratio is the change between the input signal and the output signal. Here, the color mixing ratio is a ratio at which the signal 1 actually changes with respect to the level difference between the signal 1 and the signal 2, and is a value that changes for each sample and hold circuit. In other words, the color mixing ratio is determined by the characteristics of the sample and hold circuit.
[0009]
By the way, considering a case where a color solid-state imaging device having a color filter of a primary color R (red), G (green), B (blue) Bayer array is used, a G signal / B signal is output for each line. A line (hereinafter, referred to as an R line) and a line (hereinafter, referred to as a B line) from which an R signal / B signal is output are switched. When a red subject is imaged, the R signal is considerably large even though the G signal / B signal is almost zero.
[0010]
Therefore, the signal level of the G signal on the R line increases due to the influence of the R signal, but since the G signal and the B signal are substantially equal, the G signal on the B line hardly changes. As a result, the signal levels of the G signal of the R line and the G signal of the B line are different. As a result, even if signal processing is performed based on these signals, a good image cannot be reproduced.
[0011]
Conventionally, in order to solve such a problem, in a signal processing circuit that samples and holds an output signal of a color solid-state imaging device and processes a digitized signal, a signal level of a signal input one bit ahead is used. The result of subtracting the signal level of the next input signal one bit later and multiplying the subtracted output by a positive correction coefficient is used as a correction amount (correction value), and this correction amount is input one bit ahead. The crosstalk (color mixture) generated in the sample-and-hold is corrected by adding the signal having the periodicity and outputting the resultant signal (for example, see Patent Document 1).
[0012]
[Patent Document 1]
JP-A-11-177998 [0013]
[Problems to be solved by the invention]
In the signal processing circuit and the signal processing method according to the conventional example described above, a good processing result can be obtained when no noise component is present in the image signal. However, usually, an image signal contains a noise component whose level changes in units of one pixel. This noise component is largely classified into fixed pattern noise and random noise. Of these two types of noise components, random noise is dominant as a noise component mixed in the image signal.
[0014]
The fixed pattern noise is a noise component mainly generated due to, for example, dark current of a photodiode constituting a pixel, variation in aperture of a pixel, or uneven aperture. On the other hand, random noise is caused by fluctuation of the number of photons of incident light, fluctuation of dark current of a photodiode or a vertical transfer unit, kTC noise generated when resetting FD (floating diffusion), and an amplifier connected to FD. Noise. That is, random noise has a level difference in pixel units, and it is difficult to recognize correlation between adjacent pixels.
[0015]
Therefore, when the signal level of a certain pixel and the signal level of the immediately succeeding pixel are directly subtracted, if random noise is superimposed on both, the subtracted value is the sum of the difference between the crosstalk and the random noise. I will. Since it is difficult to recognize the correlation between adjacent pixels in the difference of the random noise, the influence of the difference on the subtraction value cannot be ignored. Therefore, if the correction process is performed by directly calculating the correction value of the difference of the random noise, the noise may be increased more than the removal of the crosstalk component.
[0016]
The present invention has been made in view of the above problems, and an object of the present invention is to improve crosstalk (color mixture) generated in a sample hold without increasing a random noise component, thereby improving crosstalk. It is an object of the present invention to provide a signal processing circuit and a signal processing method capable of obtaining an original signal without noise.
[0017]
[Means for Solving the Problems]
In order to achieve the above object, according to the present invention, in a signal processing circuit for processing a signal having periodicity that has passed through a sample-and-hold circuit, first, a random noise component included in the signal having periodicity is reduced by a noise reduction unit. . Then, the signal level of the signal after noise reduction is subtracted from the signal level of the periodic signal input one bit ahead, and the result of multiplying the subtracted output by a positive correction coefficient is one bit ahead. The signal is added to the input signal having periodicity and output.
[0018]
In the signal processing circuit having the above configuration, a signal having periodicity to be a circuit input includes a crosstalk component generated in the sample and hold circuit. This signal is first passed through noise reduction means to reduce random noise components contained in the signal. Next, by subtracting the signal level of the signal from which the random noise component has been reduced from the signal level of the signal having the periodicity input one bit ahead, the effect of the difference of the random noise on the subtraction output is substantially reduced. It is reduced to a negligible extent. Then, a value obtained by multiplying the subtraction output by a positive correction coefficient is a correction amount. Therefore, by adding this correction amount to the signal one bit earlier, the crosstalk component is canceled. As a result, crosstalk (color mixture) generated in the sample hold can be favorably corrected without increasing the random noise component.
[0019]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram illustrating a configuration example of a signal processing circuit according to an embodiment of the present invention.
[0020]
The signal processing circuit according to the present embodiment samples and holds, for example, an output signal of a color CCD solid-state imaging device, and then processes a digitized signal. Further, the color CCD solid-state imaging device has a color filter having a primary color RGB Bayer array as shown in FIG.
[0021]
Note that the color arrangement is not limited to the primary color Bayer arrangement. Also, the color filters are not limited to the color arrangement of the RGB primary colors, but may be the color arrangement of other primary colors or a color arrangement using complementary colors (for example, Ye (yellow) / Cy (cyan) / Mg (magenta) / G (green)) is also applicable.
[0022]
As is clear from FIG. 1, the signal processing circuit according to the present embodiment includes a noise reduction unit that reduces a random noise component included in an input signal, for example, a low-pass filter 11 and the latency (delay time) of the low-pass filter 11 A) a register 12 for delaying by a delay time of +1 bit, a subtractor 13 for subtracting the output signal B of the low-pass filter 11 from the output signal A of the register 12, and a positive correction coefficient set externally. Is multiplied by the subtraction output (A−B) of the subtractor 13, and an adder 15 that adds the multiplication result of the multiplier 14 to the output signal A of the register 12.
[0023]
As described above, the signal processing circuit configured as described above samples and holds an output signal of, for example, a color CCD solid-state imaging device, and then processes a signal obtained by digitization. At this time, the signal output from the color CCD solid-state imaging device is a dot-sequential signal corresponding to each pixel. Therefore, a signal output from the color CCD solid-state imaging device, that is, a signal input to the signal processing circuit according to the present embodiment is a signal having periodicity.
[0024]
Then, the signal processing circuit according to the present embodiment performs the correction when receiving the influence of the signal of the adjacent pixel, that is, in order to correct the crosstalk (color mixture) generated in the sample hold, FIG. In the signal processing system for the output signal of the CCD solid-state imaging device shown in FIG. In this case, a signal including a crosstalk component generated by the sample and hold circuit 104 in the signal processing system is input to the signal processing circuit according to the present embodiment.
[0025]
The input signal is supplied to the subtractor 13 as a signal A passing through the register 12 and a signal B passing through the low-pass filter 11. The low-pass filter 11 has a structure capable of selectively processing only pixels of the same color, that is, a coefficient distribution for extracting and processing pixels of the same color. As the low-pass filter 11, for example, as shown in FIG. 3, a Laplacian filter having a 5 × 5 tap structure and processing only the same color pixel portion can be considered.
[0026]
FIG. 4 shows an example of a specific configuration of the low-pass filter 11. Here, D represents a unit delay amount (time). In the case of the low-pass filter 11 according to this example, the coefficients are set to 1,0,2,0,1. However, the configuration of the low-pass filter 11 is not limited to this, and the low-pass filter 11 may be configured to process pixels of the same color at a different ratio with another number of taps.
[0027]
The register 12 has a delay function of delaying the input signal by a delay time of one bit (one clock) corresponding to the latency of the low-pass filter 11 plus one pixel. Thereby, as shown in FIG. 5, the signal A is synchronized with the signal B by being delayed by the delay of the low-pass filter 11 plus 1 bit in the register 12. As a result, the signal B corresponds to the signal of the pixel following the signal A. As is clear from the RGB color Bayer arrangement diagram of FIG. 2, for example, if the signal A is an R signal, the signal B becomes a G signal (R line). Become.
[0028]
Next, the signal B is subtracted from the signal A in the subtractor 13. The subtraction result (A−B) of the subtractor 13 is multiplied by a positive correction coefficient set externally using a multiplier 14. The multiplication result of the multiplier 14, that is, a signal obtained by multiplying the subtraction output (A-B) by a positive correction coefficient becomes a correction amount (correction value) when correcting a crosstalk component generated in the sample hold. This correction amount is added to the output signal A of the register 12 by the adder 15.
[0029]
Here, the principle of the above-described crosstalk correction will be described using equations below. It is assumed that the signal amount when no crosstalk is generated is signal 1 / signal 2, and signals corresponding to signal 1 / signal 2 are signal 1 '/ signal 2' due to crosstalk.
[0030]
Due to the occurrence of crosstalk, the signal 1 'in which crosstalk has occurred is
Signal 1 ′ = Signal 1− (Signal 1−Signal 2) × Color Mixing Ratio Here, the color mixing ratio is a value determined by the characteristics of the sample-and-hold circuit 104 in the output signal processing system (see FIG. 6) of the CCD image sensor as described above.
[0031]
By correcting the signal 1 'in which the crosstalk has occurred in the signal processing circuit according to the present embodiment, the output signal becomes
Output signal = signal 1 '+ (signal 1'-signal 2') × correction coefficient.
[0032]
Here, assuming that both the correction coefficient and the color mixing ratio are positive values sufficiently smaller than 1,
Output signal ≒ signal 1− (signal 1−signal 2) × color mixture rate + (signal 1−signal 2) × correction coefficient.
[0033]
Furthermore, if the correction coefficient is selected so that the color mixing ratio ≒ the correction coefficient,
Output signal ≒ Signal 1
Thus, a change in signal due to crosstalk can be corrected. As is apparent from this, the correction coefficient set externally in accordance with the color mixing ratio is also a value determined by the characteristics of the sample and hold circuit 104 in the output signal processing system of the CCD solid-state imaging device.
[0034]
By the way, as described above, the image signal usually contains a noise component. There are various types of noise, but random noise is dominant. Since this random noise causes a level difference in pixel units due to its cause, it is difficult to recognize the correlation between adjacent pixels, and in severe cases, the image quality is significantly degraded. However, due to its nature, the random noise exists as a band in a higher band than the image signal. Therefore, random noise components can be suppressed by performing high-frequency reduction processing such as averaging in terms of area or time.
[0035]
Focusing on this point, in the signal processing circuit according to the present embodiment, the signal B is subjected to high-frequency reduction processing such as area or time averaging through the low-pass filter 11 and the random noise component included in the signal B is processed. Is reduced, so that the difference between random noises is reduced. Therefore, the coefficient of the low-pass filter 11 is determined based on the band of the random noise component.
[0036]
Specifically, the signal output from the CCD solid-state imaging device is first passed through the low-pass filter 11 to reduce a random noise component included in the signal. The signal B input with one bit (one pixel) after the signal A of the signal B in which the random noise component is reduced has a center of gravity as a signal position. Then, the signal level of the signal B is subtracted by the subtractor 13 from the signal level of the signal A input one bit ahead. At this time, the difference of the random noise is reduced to such an extent that the influence on the subtraction output (A−B) of the subtractor 13 can be almost ignored.
[0037]
A value obtained by multiplying the subtraction output (A−B) of the subtractor 13 by a positive correction coefficient by the multiplier 14 is a correction amount (correction value). Accordingly, the adder 15 adds this correction amount to the signal A one bit before through the register 12 to cancel the crosstalk component of the signal A of the pixel of interest without increasing the random noise component. can do.
[0038]
Here, it is also conceivable to employ a method of reducing the random noise component of the signal A through a low-pass filter as in the case of the signal B. However, if this method is adopted, the high band of the image signal will be dropped. Therefore, it is preferable to adopt a method of passing only the signal B through the low-pass filter.
[0039]
As described above, in a signal processing circuit that samples and holds an output signal of a color solid-state imaging device represented by a CCD solid-state imaging device and then processes a signal having a periodicity obtained by digitization, From the signal level of the signal A, subtracts the output signal B of the low-pass filter 11 whose center of gravity is the signal position input one pixel after the signal A, and multiplies the subtracted output (AB) by a positive correction coefficient. The correction result is used as the correction amount, and the correction amount is added to the signal A and output, whereby the crosstalk (color mixture) generated in the sample and hold can be satisfactorily corrected without increasing the random noise component.
[0040]
As a result, an original signal free of crosstalk can be obtained, and a good image can be obtained by using an image pickup signal that has been subjected to various kinds of signal processing in the subsequent signal processing system (the DSP circuit 106 in FIG. 6). Can be reproduced. Moreover, as is apparent from FIG. 1, the signal processing circuit for correction is an extremely simple circuit including the low-pass filter 11, the register 12, the subtractor 13, the multiplier 14, and the adder 15, so that a complicated circuit is required. The desired objective can be achieved without adding
[0041]
In the above-described embodiment, a case has been described in which the output signal of the color solid-state imaging device is sampled and held, and then a signal having a periodicity obtained by digitization is processed by the signal processing circuit according to the present embodiment. However, the present invention is not limited to this, and generally applies to a signal having a periodicity that has passed through a sample-and-hold circuit, in particular, a signal having a large amount of change in signal level per bit. can do.
[0042]
In the above embodiment, the low-pass filter is used as the noise reduction means. However, the present invention is not limited to the low-pass filter, and any other configuration can be used as long as it can reduce the random noise component included in the periodic signal. good.
[0043]
【The invention's effect】
As described above, according to the present invention, in a signal processing circuit that processes a signal having periodicity that has passed through a sample-and-hold circuit, the signal level of a signal input one bit ahead and one bit after the next input The signal level of the signal in which the random noise component has been reduced by performing high-frequency reduction processing on the signal of No. 1 is subtracted, and the result obtained by multiplying the subtracted output by a positive correction coefficient is used as a correction amount. By adding the signal to the input signal and outputting the signal, it is possible to satisfactorily correct the crosstalk generated in the sample hold without increasing the noise.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating a configuration example of a signal processing circuit according to an embodiment of the present invention.
FIG. 2 is a diagram showing a primary color Bayer array of a color filter.
FIG. 3 is a conceptual diagram of generation of a signal B.
FIG. 4 is a circuit diagram showing an example of a specific configuration of a low-pass filter.
FIG. 5 is a conceptual diagram of synchronizing a signal A and a signal B;
FIG. 6 is a block diagram illustrating an example of a configuration of an output signal processing system of the CCD solid-state imaging device.
FIG. 7 is a circuit diagram illustrating an example of a circuit configuration of a sample and hold circuit.
FIG. 8 is a waveform diagram showing a crosstalk phenomenon.
[Explanation of symbols]
11: Low-pass filter, 12: Register, 13: Subtractor, 14: Multiplier, 15: Adder, 101: CCD solid-state imaging device, 102: CDS (correlated double sampling) circuit, 104: Sample hold circuit, 106: DSP (digital signal processing) circuit

Claims (6)

A signal processing circuit having a periodicity through a sample and hold circuit,
Noise reduction means for reducing a random noise component included in the signal having the periodicity,
Delay means for delaying the signal having the periodicity by a delay time of the noise reduction means + 1 bit,
Subtraction means for subtracting the signal level of the output signal of the noise reduction means from the signal level of the output signal of the delay means,
Multiplication means for multiplying the subtraction result of the subtraction means by a positive correction coefficient,
A signal processing circuit comprising: an adding unit that adds a multiplication result of the multiplying unit to an output signal of the delay unit. 2. The signal processing circuit according to claim 1, wherein the noise reduction unit is a low-pass filter whose coefficient is determined based on a band of the random noise component. 2. The signal processing circuit according to claim 1, wherein the signal input to the sample and hold circuit is an output signal of a solid-state imaging device having a color filter having a predetermined color arrangement on a light receiving surface. 4. The signal processing circuit according to claim 3, wherein the noise reduction unit is a low-pass filter in which coefficients of the same color are extracted and processed. A method for processing a signal having periodicity that has passed through a sample and hold circuit,
Reducing a random noise component included in the signal having the periodicity, and subtracting the signal level of the signal after the noise reduction from the signal level of the signal having the periodicity input one bit ahead;
A signal processing method, wherein a result obtained by multiplying the subtraction output by a positive correction coefficient is added to the signal having the periodicity input one bit ahead and output. 6. The signal processing method according to claim 5, wherein the signal input to the sample and hold circuit is an output signal of a solid-state imaging device having a color filter having a predetermined color arrangement on a light receiving surface.

Images