JP3894177B2 - Video signal processing apparatus and video signal processing method - Google Patents

Description

  The present invention relates to a video signal processing apparatus and a video signal processing method for performing digital non-linear processing on a digital video signal obtained by digitizing an analog imaging signal, for example.

  In general, video signals handled by video cameras and video tape recorders are subjected to various signal processing such as gamma correction, knee correction, white black clip, contour correction, white balance adjustment, hue adjustment, detail crispening, and level depend. The

  For example, in the case of a normal color video camera, in an imaging signal processing circuit that forms and outputs a luminance signal or a chroma signal from an imaging signal obtained by an imaging unit using a CCD image sensor or the like built in the camera, Various signal processing such as gamma correction, contour correction, white balance adjustment, and hue adjustment are performed.

  In a digital signal processing camera that digitizes an imaging signal, performs digital processing on the digital imaging signal, and outputs the digital imaging signal, it is written in a nonvolatile memory such as a so-called EEPROM (electrically erasable and programmable read only memory) in advance. Based on the control parameters, the gamma correction, the contour correction, the white balance adjustment, the hue adjustment, and the like are performed, and a signal processing unit for performing these various signal processes is provided.

  Here, the configuration of a general digital signal processing camera will be described with reference to FIG.

  In FIG. 29, light from a subject enters through the optical system 100 and is imaged by the CCD image sensor 110. The imaging signal from the CCD image sensor 110 is composed of, for example, three imaging signals of three primary colors R (red), G (green), and B (blue), and this imaging signal is sent to the preamplifier 111. The imaging signal amplified by the preamplifier 111 is sent to the video amplifier circuit 112. The video amplifier circuit 112 performs processing such as black / white balance adjustment, black / white shading distortion correction, and flare correction on the R, G, and B image signals, and also performs signal amplification. The output signal of the video amplifier circuit 112 is converted into digital video data by an analog / digital (A / D) converter 113 and sent to the defect correction circuit 114. The defect correction circuit 114 performs correction corresponding to the defective portion of the CCD image sensor 110.

  The digital video data after the defect correction by the defect correction circuit 114 is sent to a contour enhancement signal generation circuit that generates a contour enhancement signal for performing horizontal and vertical contour enhancement processing. The contour emphasis signal generation circuit is composed of components from the 1H delay circuits 115, 116, 117 to the limiter 129.

  In the edge emphasis signal generation circuit, first, in the 1H delay circuits 115, 116, 117 connected in series, digital video data supplied via the defect correction circuit 114 is sequentially supplied by 1H (H is a horizontal period). In addition to delaying, each delayed video data is output. As a result, digital video data shifted by three lines in the vertical direction is output from these 1H delay circuits 115, 116, and 117. The 1H delay circuits 115, 116, and 117 are provided for generating a vertical edge enhancement signal.

  Next, the digital video data shifted by the three lines from the 1H delay circuits 115, 116, and 117 passes through the vertical digital high-pass filter (HPF) 121 and further in the horizontal digital low-pass filter (LPF) 122. Pass through. As a result, a vertical contour component is extracted from the digital video data. At the same time, the digital video data shifted by the three lines from the 1H delay circuits 115, 116, and 117 passes through the vertical digital low-pass filter (LPF) 124, and further, the horizontal digital high-pass filter (HPF) 125 also passes through. pass. As a result, a horizontal contour component is extracted from the digital video data.

  The vertical contour components taken out by the HPF 121 and the LPF 122 are sent to a multiplier 123, where they are multiplied by a gain adjustment coefficient for adjusting the degree of vertical contour emphasis supplied via a terminal 144. Is done. At the same time, the horizontal contour components taken out by the LPF 124 and the HPF 125 are sent to a multiplier 127 where a gain adjustment coefficient for adjusting the degree of horizontal contour enhancement supplied via a terminal 145 is obtained. Is multiplied.

  The output data of these multipliers 123 and 127 are added by an adder 128, limited to a predetermined level by a limiter 129, and then added as a contour enhancement signal for performing horizontal and vertical contour enhancement processing. 130, and the adder 130 adds the digital video data to the main line. The limiter 129 is provided to limit the input level to the adder 130.

  Here, in the configuration example of FIG. 29, the main digital video data to which the contour emphasis signal is added is digital video data after the output from the 1H delay circuit 116 is corrected by the linear matrix circuit 132. Yes. The linear matrix circuit 132 is provided to correct a color reproduction error caused by the imaging characteristics of the CCD image sensor 110 being different from the ideal imaging characteristics.

  The digital video data subjected to edge enhancement in the horizontal and vertical directions output from the adder 130 is subjected to knee correction by a knee correction circuit 133 with a predetermined knee characteristic, and then subjected to a predetermined gamma correction circuit 134. Gamma correction is performed, and black / white clip processing is further performed by the B / W clip circuit 135.

  Next, the digital video data subjected to nonlinear processing by the knee correction circuit 133, gamma correction circuit 134, and B / W clip circuit 135 is sent to the matrix circuit 138. In the matrix circuit 138, digital video signals of luminance (Y) and color differences (RY) and (BY) are formed from the R, G, B digital video data.

  The digital video signal from the matrix circuit 138 is converted into serial digital video data by a parallel / serial (P / S) converter 141 and then output from a terminal 143. The digital video signal from the matrix circuit 138 is converted into digital composite video data by the encoder 139 and further converted into an analog composite video signal by the digital / analog (D / A) converter 140. Is output.

  By the way, when nonlinear processing such as gamma correction, knee correction, black / white clipping, etc. is performed on the video signal, the waveform is distorted. Therefore, the signal after the nonlinear processing has an integer of frequency components included in the video signal. Double harmonic components are generated.

  In the conventional digital signal processing camera, since the high frequency signal such as the contour emphasis signal is added before the non-linear processing unit for the main line signal, particularly before the gamma correction circuit, the high frequency signal is Harmonics are generated by a non-linear process uniformly performed by the gamma correction circuit, and all signals exceeding the Nyquist frequency (fs / 2) that is 1/2 of the sampling frequency fs are somewhere in the band of 0 to fs / 2. It will be folded back as a false signal (aliasing). As a result, the image quality is remarkably impaired, such as a low-frequency beat that is not present in the original signal.

  Here, how a false signal (aliasing) is generated by adding a high frequency signal such as a contour enhancement signal before gamma correction to the main line signal will be described below.

  FIG. 30 shows an essential part of the configuration shown in FIG. 29 for simplifying the extraction of the aliasing component caused by the gamma correction.

  In FIG. 30, an analog signal is supplied to a terminal 160, and this analog signal is sent to an A / D converter 161 corresponding to the A / D converter 113 of FIG. It is sampled at fs = 18 MHz and converted into a digital signal. This digital signal a is sent to the HPF 162 showing the outline emphasis signal generation circuit in FIG. 29 in a simplified manner and the LPF 164 showing various components inserted in the main signal path in FIG. 29 in a simplified manner. The output signal c of the LPF 164 corresponds to the main line signal of FIG. 29, and the output signal b of the HPF 162 corresponds to the contour emphasis signal, that is, the high frequency signal of FIG. The high-frequency signal b from the HPF 162 is added to the output signal c of the LPF 164 by an adder 168 corresponding to the adder 130 of FIG. Thereafter, the output signal d of the adder 168 is sent to a gamma correction circuit 167 corresponding to the gamma correction circuit 134 of FIG. 29 (omitted from the knee correction circuit 133), and the gamma correction circuit 167 performs gamma correction. Non-linear processing is performed and output as a signal e from the output terminal 169, and this signal e is sent to the B / W clip circuit 135 of FIG.

Here, the signals of the respective parts in FIG. 30 will be described in terms of analog waveforms for convenience, and the digital signal a from the A / D converter 161 with the sampling frequency fs = 18 MHz is, for example, 0 to 9 MHz as shown in FIG. when a is a sweep signal _{a S,} the high frequency signal is the signal _{b S} becomes as shown in FIG. 32 which is output from the HPF162 Further, the signal _{c S} as the output signal are shown in Figure 33 from the LPF164 It becomes. Therefore, by adding the signals b _S and c _S by the adder 168, a signal d _S as shown in FIG. 34 is obtained. Thereafter, a gamma correction process is performed on the signal d _S by the gamma correction circuit 167, whereby the signal e _S on which a pseudo signal component (aliasing distortion) as shown in FIG. Will be output.

In the configuration of FIG. 30, when the digital signal a from the A / D converter 161 is, for example, a burst signal a _B as shown in FIG. 36, the high frequency signal output from the HPF 162 is as shown in FIG. signal _{b B} becomes as shown in 37, the output signal from the LPF164 is a signal _{c B} as shown in FIG. 38. Therefore, by adding the signals b _B and c _B by the adder 168, a signal d _B as shown in FIG. 39 is obtained. Thereafter, the signals by performing gamma correction processing by the gamma correction circuit 167 with respect to d _B, the gamma pseudo signal component (aliasing distortion) as shown in FIG. 40 from the correction circuit 167 aboard signal e _B Will be output.

  In order to avoid the generation of the false signal as described above, for example, if a contour emphasis signal (high frequency signal) is added after the gamma correction circuit, the generation of the false signal is surely suppressed. Since the contour emphasis signal is generated from the signal before the signal, and this contour emphasis signal is added after the gamma correction circuit, it is difficult to perform the contour emphasis on the low level main signal, and conversely the high level main signal. Tends to be over-enhanced. More specifically, if the contour emphasis signal is added only after the gamma correction, there is a problem that the contour emphasis is hardly added near the black of the main line signal.

  Therefore, an object of the present invention is made in view of the above-described circumstances, and can reduce generation of false signals due to nonlinear processing, particularly gamma correction processing, and can be performed regardless of the level of the main line signal. An object of the present invention is to provide a video signal processing apparatus and a video signal processing method capable of performing edge enhancement.

The present invention relates to a video signal processing apparatus that changes the signal level of a digital video signal by digital nonlinear processing using a nonlinear processing curve, and filter means for performing low-pass filter processing to reduce the level of a high-frequency component on an input digital video signal; Multiplication coefficient data and addition coefficient for linearly approximating each non-linear processing curve divided into a plurality of continuous sections corresponding to the level of the digital video signal subjected to the low-pass filter processing by the filter means coefficient data generation means for generating data, and multiplying means for multiplying the multiplication coefficient data to the input digital video signal not subjected to the low-pass filter processing by the filter means, the multiplication coefficient data is multiplied by said multiplying means Coefficient data generating means for digital video signal And an adding means for adding said addition coefficient data generated by, low-pass filtered by the filter means into a digital video signal as a source for generating the coefficients of the nonlinear characteristics approximated by the coefficient data generation means, the high-frequency It is characterized by using a narrow area on the nonlinear characteristic curve.

The present invention is also a video signal processing method for changing the signal level of a digital video signal by digital non-linear processing using a non-linear processing curve, wherein the input digital video signal is subjected to low-pass filter processing for reducing the level of a high frequency component. Corresponding to the level of the digital video signal subjected to the low-pass filter processing, the multiplication coefficient data and the addition coefficient data are generated for linear approximation for each section obtained by dividing the nonlinear processing curve into a plurality of continuous sections. A step of multiplying the input digital video signal not subjected to the low-pass filter processing by the multiplication coefficient data, and a step of adding the addition coefficient data to the digital video signal multiplied by the multiplication coefficient data. de that is, the source for generating the coefficients of the nonlinear approximator By the digital video signal subjected to the low-pass filtering, characterized in that it has to use the narrow region on the non-linear characteristic curve the higher the frequency.
A video signal processing method for changing a signal level of a digital video signal by digital non-linear processing using a non-linear processing curve, the step of performing low-pass filter processing for reducing the level of a high frequency component on an input digital video signal, and the low-pass filter processing Generating the multiplication coefficient data and the addition coefficient data for linear approximation for each section obtained by dividing the nonlinear processing curve into a plurality of continuous sections corresponding to the level of the digital video signal subjected to the processing, and the low pass Multiplying the input digital video signal not filtered by the multiplication coefficient data; and adding the addition coefficient data to the digital video signal multiplied by the multiplication coefficient data; Digital video signal that generates the coefficients of By applying the low-pass filtering, characterized in that it has to use the narrow region on the high-frequency as a nonlinear characteristic curve.

In the present invention, the input digital video signal is subjected to low-pass filter processing for reducing the level of high-frequency components, and the nonlinear processing curve is divided into a plurality of continuous sections corresponding to the level of the digital video signal subjected to the low-pass filter processing. Multiplication coefficient data and addition coefficient data for linear approximation are generated for each divided section, the input digital video signal not subjected to the low-pass filter processing is multiplied by the multiplication coefficient data, and the multiplication coefficient data is multiplied. For example, non-linear processing according to a gamma curve is performed on low frequency signals that are difficult to generate false signals (aliasing) even if they are distorted by adding the above addition coefficient data to the digital video signal. In addition, high-frequency signals that tend to generate false signals when distorted are processed more linearly. It is possible to suppress the generation of a false signal Te. In addition, if the level is made closer to linear processing based on the level of the main line signal in this way, the luminance level dependency of the high frequency component can be eliminated, thereby eliminating the problem that black contour enhancement cannot be achieved. . That is, in the present invention, generation of false signals due to nonlinear processing, particularly gamma correction processing, can be reduced, and contour enhancement can be performed regardless of the level of the main line signal.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  FIG. 1 shows a configuration example of a digital signal processing camera as an example of the imaging apparatus of the present invention. First, the basic operation of the configuration of FIG. 1 will be described.

  In FIG. 1, light from a subject enters through an optical system 1 and is imaged by a CCD image sensor 10. The CCD image sensor 10 includes three CCD image sensors corresponding to three channels of R (red), G (green), and B (blue) of the three primary colors of light. One pixel constituting the primary color pixel (in this case, for example, G pixel) is spatially (optically) shifted by 1/2 pixel pitch with respect to the remaining two pixels (R pixel and B pixel). Has been placed. These three-channel CCD image sensors each have 500,000 pixels and are operated at a horizontal drive frequency of 18 MHz.

  The three-channel R, G, B imaging signals from the CCD image sensor 10 are sent to the preamplifier 11. The imaging signal amplified by the preamplifier 11 is sent to the video amplifier circuit 12. The video amplifier circuit 12 performs processing such as black / white balance adjustment, black / white shading distortion correction, flare correction, and signal amplification on the R, G, B image signals. The output signal of the video amplifier circuit 12 is converted into digital video data by an analog / digital (A / D) converter 13 and sent to a defect correction circuit 14. The defect correction circuit 14 performs correction corresponding to the defective pixel of the CCD image sensor 10.

  The digital video data after the defect correction is performed by the defect correction circuit 14 is a contour that is a high-frequency signal for increasing the resolution by correcting the contour of the image in the horizontal and vertical directions, that is, correcting the contour of the image on the video signal. The signal is sent to a contour enhancement signal generation circuit that generates an enhancement signal. The edge emphasis signal generation circuit includes a 1H delay circuit 15, 16, 17, a vertical digital high-pass filter 21 (hereinafter referred to as HPF 21), a horizontal digital low-pass filter 22 (hereinafter referred to as LPF 22), and a vertical direction. Digital low-pass filter 24 (hereinafter referred to as LPF 24), horizontal digital high-pass filter 25 (hereinafter referred to as HPF 25), multipliers 23 and 27, adder 28, and limiter 29.

  In the edge emphasis signal generation circuit, first, in the 1H delay circuits 15, 16, and 17 connected in series, digital video data supplied via the defect correction circuit 14 is sequentially supplied by 1H (H is a horizontal period). In addition to delaying, each of the delayed digital video data is output. As a result, digital video data shifted by three lines in the vertical direction is output from these 1H delay circuits 15, 16, and 17. The 1H delay circuits 15, 16, and 17 are provided for generating a vertical edge enhancement signal.

  Next, the digital video data shifted by 3 lines from the 1H delay circuits 15, 16, and 17 passes through the HPF 21 and further passes through the LPF 22 through zero insertion circuits 18, 19, and 20 described later. . By passing through the HPF 21 and the LPF 22, a vertical contour component is extracted from the digital video data. At the same time, the digital video data shifted by the three lines from the 1H delay circuits 15, 16, and 17 passes through the LPF 24 and further passes through the HPF 25 through zero insertion circuits 18, 19, and 20 described later. By passing through the LPF 24 and the HPF 25, a horizontal contour component is extracted from the digital video data.

  The vertical contour components taken out by the HPF 21 and the LPF 22 are sent to the multiplier 23, where they are multiplied by a gain adjustment coefficient for adjusting the degree of vertical contour enhancement supplied via the terminal 44. Is done. At the same time, the horizontal contour components taken out by the LPF 24 and the HPF 25 are sent to a multiplier 27 through a digital low-pass filter 26 (hereinafter referred to as LPF 26), which will be described later, and supplied through a terminal 45 here. A gain adjustment coefficient for adjusting the degree of edge enhancement in the horizontal direction is multiplied.

  The output data of the multipliers 23 and 27 are added by an adder 28, limited to a predetermined level by a limiter 29, and then described as an edge enhancement signal for performing edge enhancement processing in the horizontal and vertical directions. To the adder 52 via the multiplier 30, and the adder 52 adds the digital video data to the main line. The limiter 29 is provided to limit the input level to the multiplier 30.

  Here, in the configuration example of FIG. 1, the main line digital video data to which the contour enhancement signal via the multiplier 30 is added by the adder 52 is the output data from the 1H delay circuit 16, which will be described later. Through the zero insertion circuit 19 and the digital low-pass filter 31 (hereinafter referred to as LPF 31), correction is further performed by the linear matrix circuit 32, and the digital video data is subjected to nonlinear processing by the knee correction circuit 33 and gamma correction circuit 34. ing.

  The linear matrix circuit 32 is provided to correct a color reproduction error caused by the imaging characteristics of the CCD image sensor 10 being different from the ideal imaging characteristics. The knee correction circuit 33 performs knee correction based on a predetermined knee characteristic, and the gamma correction circuit 34 performs gamma correction using a gamma curve as shown in FIG. Here, the knee correction circuit 33 and the gamma correction circuit 34 function as level compression / decompression processing means, for example. That is, the knee correction circuit 33 performs knee correction processing on the main line digital video data using a coefficient indicating a predetermined knee slope and knee point as knee characteristics. As a specific example of the knee correction process, for example, a plurality of types (for example, two types) of coefficients indicating knee slope and knee point are prepared, and the plurality of types of coefficients are appropriately selected and used. be able to. The gamma correction circuit 34 also applies gamma correction processing using the correction coefficient based on the gamma curve shown in FIG. 2 to the main line digital video data. In this gamma correction processing, as a specific example, for example, a plurality of types (for example, two types) of gamma correction coefficients are prepared, and the plurality of types of coefficients can be appropriately selected and used.

  Next, the digital video data obtained by adding the horizontal and vertical contour emphasis signals by the adder 52 to the output data via the knee correction circuit 33 and the subsequent gamma correction circuit 34 is further obtained. The B / W clip circuit 35 performs black / white clip processing which is nonlinear processing. The digital video data subjected to nonlinear processing by the B / W clip circuit 35 is sent to a matrix circuit 38 via a digital low-pass filter 36 (hereinafter referred to as LPF 36) and a decimation circuit 37 which will be described later. The matrix circuit 38 forms digital video signals of luminance (Y) and color differences (RY) and (BY) from the R, G, B digital video data.

  The digital video signal from the matrix circuit 38 is converted into serial digital video data by a parallel / serial (P / S) converter 41 and then output from a terminal 43. The digital video signal from the matrix circuit 38 is converted into digital composite video data by the encoder 39, and further converted into an analog composite video signal by the digital / analog (D / A) converter 40, from the terminal 42. Is output. In the configuration example of the present invention, the signal band is set to DC (direct current) to 6 MHz corresponding to the Rec 601 standard of the so-called CCIR (International Radio Communications Advisory Committee).

  By the way, in the digital signal processing camera shown in FIG. 1 of the present invention, in order to suppress the image quality deterioration caused by the aliasing component caused by the gamma correction process among the nonlinear processes as described above, The digital low-pass filter 50 (hereinafter referred to as LPF 50) to which main line digital video data is supplied via the zero insertion circuit 19 and the output data of the LPF 50 and those used in the gamma correction circuit 34 in FIG. A gamma slope coefficient generation circuit 51 for generating coefficient data corresponding to a differential value of a point on the gamma curve, that is, a slope of a tangent line based on the gamma curve shown, and an edge emphasis signal via the limiter circuit 29 for the coefficient data. And a multiplier 30 for multiplying the signal, and further, a contour emphasizing signal via the multiplier 30 is An adder 52 for adding the digital video data, and to arrange immediately after the gamma correction circuit 34.

  Hereinafter, prevention of the aliasing component due to the gamma correction will be described.

  FIG. 3 shows a simplified configuration of a main part of the configuration shown in FIG. 1 for explaining prevention of aliasing components caused by gamma correction.

  In FIG. 3, an analog signal is supplied to a terminal 60, and this analog signal is sent to an A / D converter 61 corresponding to the A / D converter 13 of FIG. It is sampled at fs = 18 MHz and converted to a digital signal A. The digital signal A from the A / D converter 61 shows a simplified configuration of the HPF 62 showing the outline enhancement signal generation circuit in a simplified manner, and the LPF 31 inserted in the main signal path of FIG. 1 to the knee correction circuit 33. It is sent to the LPF 63 and the LPF 63 corresponding to the LPF 50 of FIG.

  The output signal D of the HPF 62 is sent to a multiplier 66 corresponding to the multiplier 30 of FIG. The output signal B of the LPF 64 is sent to a gamma correction circuit 67 corresponding to the gamma correction circuit 34 of FIG. 1, and the output signal E of the LPF 63 is a coefficient generation circuit 65 corresponding to the gamma slope coefficient generation circuit 51 of FIG. Sent to. The output signal F of the gamma slope coefficient generation circuit 65 is sent to the multiplier 66 as a multiplication coefficient for the output signal D of the HPF 62.

  The output signal G of the multiplier 66 is sent to an adder 68, and is added by the adder 68 to the output signal C of the gamma correction circuit 67, which is a main line signal, as a contour emphasis signal (that is, a high frequency signal). The The output signal H of the adder 68 is sent to the B / W clip circuit 35 of FIG.

Here, the signals of the respective parts in FIG. 3 are described in terms of analog waveforms for convenience, and the digital signal A from the A / D converter 61 with the sampling frequency fs = 18 MHz is, for example, 0 to 9 MHz as shown in FIG. when as the sweep signal a _S, the main line signal B outputted from the LPF64 is a signal B _S as shown in FIG. Further, the signal B _S is subjected to the gamma correction processing by the gamma correction circuit 67 to become a signal C _S as shown in FIG.

Further, from the HPF62 is output signal _{D S} as shown in FIG. 7, from the LPF63 signal _{E S} as shown in FIG. 8 is output. That is, the signal E _S are the signal low-pass has been made by the LPF 63, the amplitude decreases as the frequency of the signal. In the gamma inclination coefficient generating circuit 65 the signal E _S is transmitted, it generates a signal F _S, as shown in FIG. 9 in accordance with the gradient of tangent points on the gamma curve corresponding to the level (amplitude) of the signal E _S To do. In other words, in the gamma inclination coefficient generating circuit 65, in FIG. 2, when the level of the input signal E _S on the horizontal axis (amplitude), the length of the horizontal axis corresponding to the level of the input signal E _S for each point in the range on the gamma curve in response to, determine the tangent slope respectively, it generates and outputs a signal F _S corresponding to the tangent slope of each point. The signal _{F S} is sent to the multiplier 66 as the multiplication factor, it is multiplied by the multiplier 66 to the signal _{D S} from the HPF 62.

Here, for example, the level of the low frequency side of the signal E _S is large (the amplitude is large), even becomes wide as the range of the gamma curve corresponding thereto, in this case corresponding to the level of the signal E _S The value of the tangent slope at each point on the gamma curve also changes greatly. That means that the value of the tangent slope changes significantly, (coefficient value corresponding to the signal F _S) number is multiplied to the low frequency side of the signal D _S at the multiplier 66 it means that changes significantly, This indicates that the nonlinearity becomes strong (large distortion). On the other hand, for example, the level of the high frequency side of the signal E _S is small (the amplitude is narrow) for which the corresponding range of the gamma curve also becomes narrow the gamma curve on which corresponds to the level of the case where the signal E _S is The value of the slope of the tangent at each point will not change much. That means that the value of the tangent slope is not so different is that change number (coefficients corresponding to the signal F _S) is multiplied to the high frequency side of the signal D _S is small, this linearity is strengthened Represents that.

  As described above, in the configuration of the present invention, even if the signal is distorted, a low-frequency signal that hardly generates a false signal (aliasing) is subjected to nonlinear processing according to the gamma curve, and conversely, if the signal is distorted, the signal is false. For high-frequency signals that are likely to generate signals, processing that is closer to linear is performed to suppress generation of false signals. In addition, when the level is made closer to the linear processing based on the level of the main line signal in this way, there is no problem that black outline enhancement cannot be performed.

Incidentally, if over strong high-pass limited by LPF 63, the signal also E _S of the high-frequency side amplitude (level) as was 0, corresponding to the level (i.e., amplitude 0) of the signal E _S at that time gamma Since the signal F _S corresponding to the slope of the tangent of the point on the curve is multiplied by the signal D _S (that is, the contour emphasis signal), the compression characteristic of the gamma correction process is maintained. That is, the gradient of tangent points on the gamma curve when the amplitude of the signal E _S is 0 is a complete linear processing so that the value does not vary.

The signal output from the multiplier 66 in the manner described above the signal G _S becomes as shown in FIG. 10, the signal G _S is sent to the adder 68, the gamma correction processing facilities of the main line at the adder 68 It is added to the signal C _S. The output signal of the adder 68 is a signal H _S as shown in FIG. 11.

In the above example, although depicted as no deterioration of the frequency characteristic at all signal A _S, since the actual high-frequency amplitude as signal of the signal E _S is falling, the LPF63 in FIG It is also possible not to provide it. However, in the present invention, with respect to the signal A _S, by actively performing high-pass restrictions LPF 63, so that to suppress higher-frequency aliasing occurs.

In the configuration of FIG. 3, when the digital signal A from the A / D converter 61, which is for example a burst signal A _B as shown in FIG. 12, the main line signal B outputted from the LPF64 Figure a signal _{B B} as shown in 13. The signal B _B than being gamma correction process by the gamma correction circuit 67, the signal C _B as shown in FIG. 14.

On the other hand, from the HPF62 output signal _{D B} as shown in FIG. 15, from the LPF63 is output signal _{E B} as shown in FIG. 16. That is, in the example of the burst signal A _B, the signal E _B are the signal low-pass has been made by the LPF 63, the amplitude decreases as the frequency of the signal. In the gamma inclination coefficient generating circuit 65 the signal E _B is sent, similarly to the above the signal E level (amplitude) signal as shown in FIG. 17 corresponding to the gradient of the tangent points on the gamma curve corresponding to F and _{B B} is generated. The signal _{F B} is sent to the multiplier 66 as the multiplication factor, it is multiplied by the multiplier 66 to the signal _{D B} from the HPF 62.

In the example of the burst signal A _B, for example, the level of the low frequency side of the signal E _B is large (the amplitude is large), even becomes wide as the range of the gamma curve corresponding thereto, in this case the signal the value of the tangent slope at each point on the gamma curve corresponding to the level of E _B also becomes vary large, resulting in the number to be multiplied to the low frequency side of the signal D _B at the multiplier 66 (signal F ₍ Coefficient value corresponding to _B ) greatly changes, which indicates that nonlinearity becomes stronger (distorts greatly). Further, for example, the level of the high frequency side of the signal E _B is small (the amplitude is narrow), so the gamma curve on which corresponds to the level of the signal E _B in this case also becomes narrow range on the gamma curve corresponding to this the value of the tangent slope at each point will be not much, as a result it means that change the number (coefficients corresponding to the signal F _B) multiplied to the high frequency side of the signal D _B is small, which is This shows that the linearity becomes stronger.

Furthermore, if LPF63 over strong high-pass restrictions, even as a high-frequency side of the amplitude (level) and 0 of the signal E _B, gamma corresponding to the level (i.e., amplitude 0) of the signal E _B at that time Since the signal F _B corresponding to the slope of the tangent of the point on the curve is multiplied by the signal D _B (that is, the contour enhancement signal), the compression characteristic of the gamma correction process is maintained. That is, the gradient of tangent points on the gamma curve when the amplitude of the signal E _B is 0 is a complete linear processing so that the value does not vary.

The signal output from the multiplier 66 in the manner described above the signal G _B becomes as shown in FIG. 18, the signal G _B is sent to the adder 68, the gamma correction processing facilities of the main line at the adder 68 It is added to the signal C _B. The output signal of the adder 68 is a signal H _B as shown in FIG. 19.

Thus, in the example of the burst signal A _B, because false signals even likened distorting for low frequency signals (aliasing) hardly occurs performs nonlinear processing in accordance with the gamma curve, conversely, If a high frequency signal is distorted, a false signal is likely to be generated. Therefore, it is possible to suppress the generation of a false signal by performing processing that is more linear. Also in this example, the problem that black contour enhancement cannot be achieved is eliminated by making it closer to linear processing as the level of the main line signal is smaller.

  Note that the technique for preventing the generation of false signals as described above can be applied not only to the contour emphasis signal generation circuit part but also to the gamma correction circuit itself. That is, in FIG. 2, when the input is the x axis and the output is the y axis, the gamma curve can be approximated by a linear expression of y = ax + b when viewed in a narrow section. In this equation, a represents the slope of the tangent line on the gamma curve, and b represents the y-axis intercept. Therefore, when the amplitude of the high frequency is reduced by the low-pass filter in the same manner as described above, the level change becomes smaller as the frequency becomes higher, and the fluctuation range of the slope a and the intercept b becomes smaller. As the distortion becomes smaller, aliasing is suppressed.

  A specific configuration of the gamma correction circuit in this case is as shown in FIG. In FIG. 20, output data from the knee correction circuit 33 in FIG. 1 is supplied to a terminal 70, and the data is restricted to high-pass by a low-pass filter (LPF) 71. The data input through the terminal 70 is sent to the output terminal 75 through the multiplier 73 and further through the adder 74. The output of the LPF 71 is sent to a coefficient generation circuit 72. The coefficient generation circuit 72 calculates and outputs the slope a and the intercept b according to the input level. The data of the slope a is sent as the multiplication coefficient of the multiplier 73, and the data of the intercept b is added to the output data of the multiplier 73 by the adder 74.

  Further, the coefficient generation circuit 72 is specifically configured as shown in FIG. In FIG. 21, the output of the LPF 71 is supplied to the terminal 80, and the output is sent to the level comparator 81, where the output level of the LPF 71 is measured. The output of the level comparator 81 is sent to a coefficient a table 82 and an intercept b table 83. The coefficient a data corresponding to the measurement level is read from the coefficient a table 82, and the intercept b table 83 reads the above-described data. Data of the intercept b corresponding to the measurement level is read out.

  Next, in the digital signal processing camera shown in FIG. 1 of the present invention, in order to suppress deterioration in image quality caused by aliasing components caused by nonlinear processing such as knee correction other than the above gamma correction and black / white clip, The sampling frequency in the nonlinear processing portion is increased so that the aliasing component does not occur in the signal band. That is, in the configuration of FIG. 1, the zero converters 18, 19, 20 and the LPFs 22, 26, 31 in the preceding stage of the nonlinear processing portion constitute an up converter, and the sampling in the nonlinear processing portion is achieved by the configuration of these up converters. I try to raise the frequency.

  Hereinafter, the generation prevention of the aliasing component due to the non-linear processing by providing the up-converter will be described.

  Here, when the original sampling frequency is fs and the sampling frequency after up-conversion is fs ′, the frequency fs ′ is usually twice or four times the frequency fs in order to simplify the circuit configuration. Often an integer multiple of. However, in a digital circuit constituted by a so-called CMOS (Complementary Metal Oxide Semiconductor), the power consumption is proportional to the operating frequency, so it is not preferable to raise it to a very high frequency.

  For this reason, in this configuration example, the frequency fs ′ = 2fs is set in the nonlinear processing portion from the knee correction circuit 33 to the B / W clip circuit 35 for the main line signal. It should be noted that the frequency fs ′ = 4 fs can be set for the limiter 29 portion. As described above, when the frequency fs ′ in the non-linear processing portion is set to fs ′ = 2fs, it is possible to prevent the third-order harmonic from returning to the original signal band as shown in FIG. .

  Returning to FIG. 1, the configuration of the up-converter will be described.

  The up-converter inserts zero data into the output data (that is, the sampled data string) from the 1H delay circuits 15, 16, and 17 as shown in FIG. 23A, and inserts zero to raise the sampling frequency. The circuits 18, 19, and 20 and LPFs 22, 26, and 31 for obtaining a data string as shown in B of FIG. 23 by performing band limitation on the data string into which the zero data is inserted. That is, in the up-converter, when zero is inserted as shown in A of FIG. 23 into output data (that is, a sampled data string) as shown in A of FIG. 24 from the 1H delay circuits 15, 16, and 17, FIG. As shown in B, the aliasing component from the frequency fs is generated in the band of the frequency fs / 2 or higher. At this time, if the band is limited to the frequency fs / 2 or lower by the LPF as shown in FIG. Output data as shown in FIG. 23B from which the aliasing component has been removed is obtained. In FIGS. 23 and 24, the case where the frequency is fs ′ = 2fs is taken as an example.

  Here, if the band limitation by the LPF after the up-conversion is low, for example, if the aliasing component remains in the band of frequency fs / 2 or more as shown by the dotted line in FIG. Such folding distortion occurs. In the analog signal processing camera, the signal band or more is attenuated by, for example, a low-pass filter, and there is no problem because the circuit does not have much gain.

  That is, in FIG. 25, if the LPF characteristic of the up-converter is poor and the aliasing component leaks at the band limiting unit, the signal component of the original frequency f and the aliasing component of the frequency fs-f as shown in FIG. Is the result input to the nonlinear processor. As a result, the non-linear processing unit has a form of mutual modulation, and as shown in FIG. 25B, the frequency f and an integer multiple of the frequency fs-f and the difference between the frequency fs-f and the frequency f are obtained. That is, a term that is an integral multiple of the frequency fs-2f is generated. For this reason, for example, even if the sampling frequency is increased to the frequency fs ′ so as to prevent the occurrence of aliasing, a low frequency component such as the frequency fs-2f is generated, which is not very effective. In FIG. 25B, the harmonic components of f are 2f and 3f, the aliasing components of f harmonic components from fs' are fs'-2f and fs'-3f, and the harmonic components of fs-f. The wave components are 2 (fs-f) and 3 (fs-f), and the folded components from fs' of the harmonic components of fs-f are fs'-2 (fs-f) and 3 (fs-f). -Fs', 2fs'-3 (fs-f), the difference between f and (fs-f) and its harmonic components are fs-2f, 2 (fs-2f), and f and (fs-f ) And the return components of the harmonic components from fs ′ are fs ′ − (fs−2f) and fs′−2 (fs−2f).

  As described above, when the aliasing component as shown in FIG. 25 is generated by the non-linear processing, the third-order of the aliasing component (fs−f) from fs even if the band is limited up to fs / 2 thereafter. Two of the harmonic folding component 3 (fs−f) −fs ′ and the difference fs−2f of the folding component (fs−f) from f and fs remain as folding in the band.

  Therefore, the band limitation after the zero insertion must sufficiently prevent fs / 2 or more (usually -40 dB or less, particularly conspicuous frequency is -60 dB or less).

  Here, when fs ′ = 2fs, the following frequencies are particularly noticeable.

  For example, if f = 2fs / 2 + α (α << fs / 2), the difference between the aliasing component (fs−f) from f and fs = fs−2f = −2α, and if α is very small, Near-frequency aliasing distortion is noticeable.

  For example, when f = fs / 3 + α (α << fs / 2), the third harmonic return component of the return component (fs−f) from fs = 3 (fs−f) −fs ′ = 3α Thus, low-frequency aliasing distortion is conspicuous. Further, the difference between the aliasing components from f and fs (fs = f) = fs−2f = fs / 3−2α, and the frequency 3α between the frequency f (= fs / 3 + α) of the original signal This produces a low frequency beat.

  For example, if f = fs / 4 + α (α << fs / 2), the difference between the aliasing components (fs-f) from f and fs = fs−2f = fs / 2−α A low frequency beat of a frequency 2α is generated between the harmonic component 2f (= fs / 2 + 2α).

  For this reason, the attenuation factor in the vicinity of the frequencies fs / 2, 2fs / 3, and 3fs / 4 returned from fs described above must be as large as possible. However, since the pass band is up to fs / 2, the vicinity of the first frequency fs / 2 becomes a transition area of a normal low-pass filter, and sufficient attenuation cannot be obtained. In this case, sampling at a frequency fs higher than a necessary band in advance and taking a large attenuation rate in the vicinity of fs / 2 or giving up the aliasing distortion in the vicinity of fs / 2 may cause the aliasing component to leak. It is done.

  Here, when the aliasing distortion that appears near the frequency fs / 2 is given up as in the latter case, the aliasing component is leaked by extending the passband and leaking the aliasing component as shown in FIG. Since the distortion component fs-f is canceled if the high-frequency component is replaced with R + G by shifting the spatial pixel of the CCD image sensor 10, it is considered that the characteristics of the LPF for band limitation may be sweet. . However, in reality, there is no effect of shifting the spatial pixels of the CCD image sensor 10 due to the influence of lateral chromatic aberration at the periphery of the screen, and therefore aliasing distortion at the periphery of the screen is very conspicuous (however, The result is about an order of magnitude less than without up-conversion.

  On the other hand, in the configuration example of the present invention, the signal band is reduced to the minimum necessary, and instead, the attenuation factor in the high frequency generated as the aliasing component is increased to reduce the aliasing distortion. Yes. In the configuration example of the present invention, since the CCD image sensors 10 for three channels each have the same 500,000 pixels, the frequencies fs and fs ′ are as shown in FIG. Although the frequency is the same, the signal band is limited to DC (direct current) to 6 MHz in the CCIR Rec 601 and, instead, the attenuation factor of 9 MHz or more is increased to reduce the aliasing distortion.

  As described above, if the attenuation at the frequency fs / 2 or more is sufficiently taken, even if the signal as shown in FIG. 27A is inputted to the nonlinear processing unit, the signal band as shown in B of FIG. Only the harmonic components of, and their aliasing components from fs ′ appear.

  Next, after the nonlinear processing, the processing may be continued with the up-conversion as described above. However, since the power consumption of the circuit is proportional to the sampling frequency, in the configuration example of the present invention, the sampling frequency is used. Is reduced to a lower sampling frequency, that is, down-converted (or decimated).

  However, if the decimation is performed as it is, the harmonic component will be folded back, so a digital low pass filter (LPF) 36 is provided at the subsequent stage of the B / W clip circuit 35 of the nonlinear processing unit, and the LPF 36 is the same as B in FIG. As shown in FIG. 28B, band limitation is applied to the signal A shown in FIG.

  The LPF 36 performs band limitation of 0 to fs / 2 when the frequency fs ″ after down-conversion is fs ″> fs, and 0 to fs ″ when fs ″ is fs ″ <fs. Band limiting of fs ″ / 2 is performed.

  After the LPF 36, the decimation circuit 37 performs down-conversion, as shown in FIG.

  In the case of down-conversion, it is not always necessary to return to the original sampling frequency fs. For example, sampling is performed at fs = 18 MHz in accordance with the horizontal drive frequency of the CCD image sensor 10 of 500,000 pixels, nonlinear processing is processed at fs ′ = 36 MHz, and then down-converted to fs ″ = 13.5 MHz for serial digital communication. For example, it can be output in accordance with the standard.

  As described above, in the configuration example of the present invention, an up-converter is inserted before the non-linear processing, further band limiting is performed to remove the aliasing component, and the harmonic aliasing does not interfere with the fundamental wave (signal frequency). Image quality caused by the aliasing component due to nonlinear processing by performing nonlinear processing at such a high sampling frequency, and then band-limiting to remove harmonic components and lowering the sampling frequency with a down converter Deterioration can be suppressed. In addition, according to the configuration example of the present invention, compared to processing the whole at a high sampling frequency, power consumption is small, and by increasing the sampling frequency only in a necessary portion, signal processing with less aliasing distortion can be performed. It is possible.

  In the example of FIG. 1, the digital low-pass filter for performing band limitation includes an LPF 22 in a path for generating a horizontal outline enhancement signal, and an LPF 26 provided in a path for generating a horizontal outline enhancement signal. The LPF 31 provided in the main route is divided into three because the required bandwidth is different in each route. Therefore, instead of these three LPFs 22, 26, and 31, LPFs that perform the same operations can be inserted immediately after the zero insertion circuits 18, 19, and 20.

1 is a block circuit diagram illustrating a schematic configuration example of a digital signal processing camera as an imaging apparatus of the present invention. It is a figure which shows a gamma curve. It is a block circuit diagram which simplifies and shows the structure of the principal part of the digital signal processing camera concerning this invention. It is a figure which shows the sweep waveform signal as an example of the signal input into the structure of FIG. It is a figure which shows the sweep waveform signal after carrying out the high pass restriction | limiting of the sweep waveform signal input into the structure of FIG. 3 with a low-pass filter (LPF64). It is a figure which shows the sweep waveform signal after the waveform signal of FIG. 5 was processed by the gamma correction circuit. It is a figure which shows the waveform signal after carrying out the low-pass restriction | limiting of the sweep waveform signal input into the structure of FIG. 3 with a high pass filter. It is a figure which shows the waveform signal after carrying out the high-pass restriction | limiting of the sweep waveform signal input into the structure of FIG. 3 with a low-pass filter (LPF63). It is a figure which shows the waveform signal after the waveform signal of FIG. 8 was processed by the gamma inclination coefficient generation circuit. FIG. 5 is a diagram showing a waveform signal (contour emphasis signal for the sweep waveform signal in FIG. 4) after multiplying the sweep waveform signal after passing through the high-pass filter by a coefficient from the gamma slope coefficient generation circuit. It is a figure which shows the sweep waveform signal after adding an outline emphasis signal to the sweep waveform signal after passing a gamma correction circuit. It is a figure which shows the burst waveform signal as an example of the signal input into the structure of FIG. It is a figure which shows the burst waveform signal after carrying out high-pass restriction | limiting of the burst waveform signal input into the structure of FIG. 3 with a low-pass filter (LPF64). It is a figure which shows the burst waveform signal after the waveform signal of FIG. 13 was processed by the gamma correction circuit. It is a figure which shows the burst waveform signal after carrying out the low-pass restriction | limiting of the burst waveform signal input into the structure of FIG. 3 with a high pass filter. It is a figure which shows the waveform signal after carrying out high-pass restriction | limiting of the burst waveform signal input into the structure of FIG. 3 with a low-pass filter (LPF63). It is a figure which shows the waveform signal after the waveform signal of FIG. 16 was processed by the gamma inclination coefficient generation circuit. FIG. 14 is a diagram showing a waveform signal (contour emphasis signal with respect to the burst waveform signal in FIG. 13) after multiplying the burst waveform signal after passing through the high-pass filter by a coefficient from the gamma slope coefficient generation circuit. It is a figure which shows the burst waveform signal after adding an outline emphasis signal to the burst waveform signal after passing a gamma correction circuit. It is a block circuit diagram which shows the structural example in the case of suppressing aliasing by the gamma correction circuit itself. FIG. 21 is a block circuit diagram showing a specific configuration of a coefficient generation circuit of the gamma correction circuit of FIG. 20. It is a figure for demonstrating that a folding | turning component does not generate | occur | produce by non-linear processing by up-conversion. It is a figure for demonstrating the zone | band limitation by zero insertion and LPF. It is a figure for demonstrating the example in case a folding | turning component remains by band limitation. It is a figure for demonstrating a harmonic component and a folding component in case a folding component leaks also by zone | band limitation. It is a figure for demonstrating the case where attenuation | damping in fs / 2 or more is taken, and the case where it does not take. It is a figure for demonstrating the harmonic and the aliasing component at the time of fully taking attenuation in fs / 2 or more. It is a figure for demonstrating downconversion and the band limitation before it. It is a block circuit diagram which shows the example of schematic structure of the conventional digital signal processing camera. It is a block circuit diagram which simplifies and shows the structure of the principal part of the conventional digital signal processing camera. It is a figure which shows the sweep waveform signal as an example of the signal input into the structure of a prior art example. It is a figure which shows the waveform signal after carrying out the low-pass restriction | limiting of the sweep waveform signal input into the structure of the prior art example with a high pass filter. It is a figure which shows the sweep waveform signal after carrying out the high-pass restriction | limiting of the sweep waveform signal input into the structure of the prior art example with a low-pass filter. It is a figure which shows the sweep waveform signal after adding a high region signal and a main line signal by the structure of a prior art example. It is a figure which shows the sweep waveform signal after adding a high region signal and a main line signal by the structure of a prior art example, and performing a gamma correction. It is a figure which shows the burst waveform signal as an example of the signal input into the structure of a prior art example. It is a figure which shows the waveform signal after carrying out the low-pass restriction | limiting of the burst waveform signal input into the structure of the prior art example with a high pass filter. It is a figure which shows the waveform signal after carrying out the high-pass restriction | limiting of the burst waveform signal input into the structure of the prior art example with a low-pass filter. It is a figure which shows the burst waveform signal after adding a high region signal and a main line signal by the structure of a prior art example. It is a figure which shows the sweep waveform signal after adding a high region signal and a main line signal by the structure of a prior art example, and performing a gamma correction.

Explanation of symbols

  71 Low-pass filter, 72 coefficient generation circuit, 73 multiplier, 74 adder

Claims (8)

A video signal processing apparatus that changes a signal level of a digital video signal by digital nonlinear processing using a nonlinear processing curve,
Filter means for performing low-pass filter processing to reduce the level of high-frequency components in the input digital video signal;
Multiplication coefficient data and addition coefficient data for linear approximation for each section obtained by dividing the nonlinear processing curve into a plurality of continuous sections corresponding to the level of the digital video signal subjected to the low-pass filter processing by the filter means. Coefficient data generating means for generating
Multiplying means for multiplying the input digital video signal not subjected to the low-pass filter processing by the filter means with the multiplication coefficient data;
Adding means for adding the addition coefficient data generated by the coefficient data generation means to the digital video signal multiplied by the multiplication coefficient data by the multiplication means ;
A low-pass filter is applied to the digital video signal from which the coefficient data generating means generates a coefficient for approximating the nonlinear characteristic by the filter means, and a narrower region on the nonlinear characteristic curve is used for higher frequencies. Video signal processing device. The coefficient data generating means is
Level detecting means for detecting the signal level of the digital video signal from which the low-pass filter processing has been performed,
Selection means for selecting a section for linearly approximating the nonlinear processing curve according to the signal level of the digital video signal detected by the level detection means;
2. The video signal processing apparatus according to claim 1, further comprising: coefficient data output means for outputting multiplication coefficient data and addition coefficient data for the section selected by the selection means.   2. The video signal processing apparatus according to claim 1, wherein the non-linear processing curve gives a gamma correction characteristic. A zero insertion means for up-converting the frequency of the input digital video signal is provided in the previous stage,
2. The video signal processing apparatus according to claim 1, wherein after the sampling frequency of the input digital video signal is increased, digital nonlinear processing using the nonlinear processing curve is performed. A video signal processing method for changing a signal level of a digital video signal by digital non-linear processing using a non-linear processing curve,
Applying low pass filter processing to reduce the level of high frequency components to the input digital video signal;
A step of generating multiplication coefficient data and addition coefficient data for linear approximation for each section obtained by dividing the nonlinear processing curve into a plurality of continuous sections corresponding to the level of the digital video signal subjected to the low-pass filter processing. When,
Multiplying the input digital video signal not subjected to the low-pass filter processing by the multiplication coefficient data;
Adding the addition coefficient data to the digital video signal multiplied by the multiplication coefficient data,
A video signal processing method characterized in that a narrower region on a nonlinear characteristic curve is used in a higher frequency region by applying the low-pass filter process to a digital video signal that is a source for generating a coefficient of the nonlinear characteristic approximation. The step of generating the multiplication coefficient data and the addition coefficient data is as follows:
Detecting a signal level of the digital video signal subjected to the low-pass filter processing;
Selecting a section for linearly approximating the nonlinear processing curve according to the detected signal level of the digital video signal;
6. The video signal processing method according to claim 5, further comprising a step of outputting multiplication coefficient data and addition coefficient data for the selected section.   6. The video signal processing method according to claim 5, wherein the nonlinear processing curve gives a gamma correction characteristic. Performing zero insertion processing on the input digital video signal for up-converting the frequency,
6. The video signal processing method according to claim 5, wherein digital nonlinear processing using the nonlinear processing curve is performed after increasing the sampling frequency of the input digital video signal.

Images