JP2012028937A - Video signal correction apparatus and video signal correction program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a video signal correction apparatus capable of performing locally contrast correction to a low luminance portion without affecting a tone of entire video.SOLUTION: A video signal correction apparatus 1 comprises virtual radiance calculation means 20 calculating a radiance component by performing operation to a luminance component of a video signal due to reverse characteristic of camera characteristics, virtual radiance correction means 30 performing contrast correction of the radiance component by performing a convolution operation using Gaussian kernel having a plurality of sizes defined in advance for each pixel and correcting a luminance value of a pixel lower than a predetermined threshold with a coefficient defined in advance, and video pixel value conversion means 40 calculating a luminance value after video adjustment by performing an operation due to the camera characteristics to the luminance value of the pixel of the corrected luminance image and generating the video signal after correction.

Description

  The present invention relates to a video signal correction apparatus and a video signal correction program that perform contrast correction of a video signal.

Generally, in order to display a vast dynamic range in the real world on a device such as a display, a photographing camera adjusts an optical aperture and a filter to compress the dynamic range. However, in outdoor sports broadcasts and the like, there is a large difference in brightness between the sunshine and the shadow. Therefore, with only optical adjustment, the image quality (details) of both the high and low luminance areas is high. ) Is very difficult to display. Therefore, it has been necessary to limit the object to be photographed to either the high luminance side or the low luminance side, and optical adjustment has been performed so that one of them can be seen well. At this time, the other image quality is deteriorated.
Various proposals have been made as a technique (tone reproduction technique) for solving such a problem (for example, Non-Patent Document 1).

Here, a method proposed in Non-Patent Document 1 (hereinafter referred to as a conventional method) will be briefly described.
In the conventional method, when L _w (x, y) is a real-world radiance corresponding to the pixel (x, y) in the image, the logarithmic average L _w ′ of the radiance shown in the following equation (1) Is calculated. The logarithmic average L _w ′ is a reference brightness in the image, and serves as an index for determining an intermediate gray zone (intermediate luminance region) of a zone system known in photographic optics.

Here, N is the number of pixels of the image, and δ is a minute value for preventing a calculation underflow that occurs when the luminance value is “0”.
Further, the conventional method defines a scaled luminance L (x, y) expressed by the following equation (2).

Here, a is called a key value, and is a value that determines whether an image is subjectively bright (high key) or dark (low key). Usually, a = 0.18 is widely used as a key value.
Then, the conventional method specifies the provisional luminance value L _d (x, y) on the video signal (display) by scaling the display range of the entire image by the following equation (3).

  As a result, the luminance is scaled to a dynamic range of 0.0 to 1.0 times that can be displayed on the display.

Next, the conventional method emphasizes the contrast of pixels that are greatly different from the luminance values in the peripheral area.
That is, the conventional method convolves the luminance image specified by the scaled luminance value L (x, y) calculated by the above equation (2) with a circular Gaussian kernel defined by a plurality of spatial scales s _i. Using the resulting luminance value V (x, y, s _i ), the activity (scale of contrast determination) (Activity (x, y, s _i )) shown in the following equation (4) is calculated.

Here, φ is a sharpness parameter that works in a direction in which the edge is emphasized when the value is reduced. “a” is the same key value as in the formula (2). In the conventional method, the minimum s _{i is set} to “0.35”, and is sequentially increased by 1.6 times, thereby using _eight spatial scales s _{1 to} s ₈ .

In the conventional method, the minimum s _{1 to} s ₈ (here, s _m ) satisfying | Activity (x, y, s _i ) |> ε (ε is a threshold for determining the presence or absence of contrast). The luminance value L _d (x, y) after contrast correction as a video signal is calculated by the following equation (5) using the luminance value V (x, y, s _i ) specified by

  Thus, the conventional method can keep the real world radiance distribution within the dynamic range of the display while maintaining the contrast.

E. Reinhard, M. Stark, P. Shirley, and J. Ferwerda, "Photographic Tone Reproduction for Digital Images", ACM Transactions on Graphics, vol. 21, no. 3, pp. 267-276, 2002.

Since the above-described conventional method uses an input luminance distribution as a real-world radiance distribution, applying the method to a moving image and processing it in real time means that the optical system of the photographing camera or electrical signal processing, Furthermore, since the subsequent video processing and the like are complicated, it is impossible at present.
In addition, the conventional method has a problem that the effect of contrast correction in a low-luminance portion cannot be sufficiently obtained due to the nature of computation for performing contrast correction. That is, since the conventional method performs the contrast correction calculation in the equation (5), the denominator value “1” is greatly influenced at low luminance, and the change in the low luminance portion cannot be expressed appropriately. is there.

  Furthermore, since the conventional method is intended to correct the radiance distribution in the real world, it cannot be applied to the contrast correction of an existing video signal already captured by a camera. For example, in order to improve image quality and expressive power, for example, various signal processing such as compressing a high luminance region by knee correction processing is generally performed on video captured by a camera. For this reason, the luminance distribution of the video signal photographed by the camera is different from the radiance distribution in the real world, and if the conventional method is applied as it is, there is a problem that the tone of the entire video changes.

  The present invention has been made in view of the above problems, and locally corrects the contrast of a low-brightness portion of a video signal captured by a camera without affecting the tone of the entire video. It is an object of the present invention to provide a video signal correction apparatus and a video signal correction program that can be used.

  The present invention has been made in order to solve the above-mentioned problems. First, the video signal correction apparatus according to claim 1 is a video signal correction apparatus for correcting the contrast of a video signal, and includes virtual radiance. The calculation unit, the filter unit, the change amount calculation unit, the kernel determination unit, the luminance value selection unit, the low luminance value correction unit, and the video pixel value conversion unit are provided.

  In such a configuration, the video signal correction apparatus calculates the luminance component of the video signal based on the inverse characteristic of the camera characteristic based on the camera characteristic indicating the characteristic of the video signal processed by the virtual radiance calculation unit. To calculate a virtual radiance component before signal processing. Normally, a video signal photographed by a camera is a non-linear signal by signal processing such as gamma and knee. Therefore, the virtual radiance calculation means applies the approximate inverse characteristic assumed from the camera characteristics standard to the video signal, thereby converting it into a luminance signal that can be regarded as sufficiently linear for contrast calculation. be able to.

  Then, the video signal correction apparatus performs a convolution operation on the radiance component calculated by the virtual radiance calculation unit using the Gaussian kernel having a plurality of predetermined sizes for each pixel. As a result, a plurality of luminance values (filtered luminance values) to which Gaussian kernels having different sizes are applied are generated.

  Further, the video signal correction apparatus has two sets of filters calculated by adjacent Gaussian kernels in the order of the size of the Gaussian kernels in the plurality of filtered luminance values in the same pixel calculated by the filter unit by the change amount calculation unit. A change amount is calculated for each finished luminance value.

  Then, the video signal correction apparatus determines, as the Gaussian kernel to be applied to the pixel, the minimum Gaussian kernel in which the change amount calculated by the change amount calculation unit is larger than a predetermined threshold by the kernel determination unit. This determines the smallest Gaussian kernel where the contrast difference is greater than a predetermined amount.

  Then, the video signal correction device uses the luminance value selection unit to calculate the filtered luminance calculated by applying the Gaussian kernel determined by the kernel determination unit from the plurality of filtered luminance values in the same pixel calculated by the filtering unit. A luminance image is generated by selecting a value as a luminance value in the pixel. As a result, a luminance image in which the luminance value having the highest contrast is selected is generated.

  Further, the video signal correction device uses the low luminance value correction unit to determine whether the luminance value is generated in the luminance image generated by the luminance value selection unit based on a predetermined threshold value indicating whether the luminance value is low luminance. The luminance value of the pixel having low luminance is corrected by a predetermined coefficient. As a result, contrast correction can be performed partially on the low-luminance portion.

  Then, the video signal correction device performs the calculation based on the camera characteristics on the luminance value of each pixel of the luminance image corrected by the low luminance value correction unit by the video pixel value conversion unit, thereby performing signal processing by the camera. Is calculated, and a corrected video signal is generated. Thereby, it is possible to reproduce the camera characteristics by the video adjustment performed by the camera on the video signal after the contrast correction.

  According to a second aspect of the present invention, there is provided the video signal correction apparatus according to the first aspect, wherein the video signal in the video signal correction apparatus according to the first aspect is a video signal expressed in a YCbCr color space that is a broadcast video signal, The conversion unit and the color space inverse conversion unit are further provided.

In such a configuration, the video signal correction device converts the input video signal into a video signal in the xyY color space by the color space conversion means. As a result, a video signal in the YCbCr color space that is generally used as a broadcast video signal is converted into a video signal in the xyY color space, and the luminance signal in consideration of color is subjected to contrast correction in the subsequent processing. And
Then, the video signal correction device reversely converts the corrected video signal generated by the video pixel value conversion unit into a video signal in the YCbCr color space by the color space reverse conversion unit. As a result, the contrast-corrected video signal can be used as a broadcast video signal.

According to a third aspect of the present invention, in the video signal correction apparatus according to the first or second aspect, the virtual radiance calculation means has a camera characteristic and a radiance component L ^* of qL ^* / The camera characteristic parameters p and q approximated as the characteristic that becomes the luminance component of (1 + pL ^* ) are input from the outside, and the signal L is calculated by calculating L _d / (q−pL _d ) for the luminance component L _d of the video signal. A virtual radiance component L ^* before processing is calculated.

  In such a configuration, the video signal correction apparatus uses the virtual radiance calculation means to calculate the camera characteristics as known characteristics and perform the calculation using the camera characteristic parameters indicating the characteristics, thereby obtaining the brightness before video adjustment by the camera. A component (radiance component) can be virtually generated.

The video signal correction apparatus according to claim 4 is the video signal correction apparatus according to claim 3, wherein the luminance value of each pixel of the luminance image corrected by the video pixel value conversion means by the low luminance value correction means. Is calculated by calculating qL ^* / (1 + pV _s ^* ) using camera characteristic parameters p and q, where V _s ^{* is} a virtual radiance component and L ^* is a virtual radiance component. It is characterized by that.

  In such a configuration, the video signal correction device performs the inverse conversion of the conversion performed by the virtual radiance calculation unit using the video pixel value conversion unit with the camera characteristic as a known characteristic, so that the luminance component after contrast correction is obtained. A luminance component having characteristics after image adjustment by the camera can be generated.

Furthermore, the video signal correction device according to claim 5 is the video signal correction device according to any one of claims 1 to 4, wherein the low luminance value correction means sets the luminance value before correction to V ^*. When the threshold value α and the coefficient β are input from the outside and the luminance value V ^* before correction is small with reference to the threshold value α, V ^* −β (1−V ^* / α) The luminance value after correction is calculated by the above calculation.

  In such a configuration, in the video signal correction apparatus, contrast adjustment corresponding to the coefficient is performed for the luminance component smaller than the threshold value by the low luminance value correction means. This makes it possible to perform contrast adjustment locally on the low luminance component.

  According to a sixth aspect of the present invention, there is provided a video signal correction program comprising: a computer for virtual radiance calculation means, filter means, change amount calculation means, kernel determination means, luminance value selection means; It is configured to function as a low luminance value correction unit and a video pixel value conversion unit.

In such a configuration, the video signal correction program calculates the luminance component of the video signal based on the inverse characteristic of the camera characteristic based on the camera characteristic indicating the characteristic of the video signal processed by the virtual radiance calculation means. To calculate a virtual radiance component before signal processing.
Then, the video signal correction program performs a convolution operation on the radiance component calculated by the virtual radiance calculation unit by the filter unit using a Gaussian kernel having a plurality of predetermined sizes for each pixel.

The video signal correction program also includes two sets of filters calculated by adjacent Gaussian kernels in order of magnitude of Gaussian kernels in a plurality of filtered luminance values calculated by the filter unit by the change amount calculation unit. A change amount is calculated for each finished luminance value.
Then, the video signal correction program determines, as the Gaussian kernel to be applied to the pixel, the minimum Gaussian kernel in which the change amount calculated by the change amount calculation unit is larger than a predetermined threshold value by the kernel determination unit.

  Then, the video signal correction program uses the luminance value selecting unit to calculate the filtered luminance calculated by applying the Gaussian kernel determined by the kernel determining unit from the plurality of filtered luminance values in the same pixel calculated by the filtering unit. A luminance image is generated by selecting a value as a luminance value in the pixel.

Further, the video signal correction program has a luminance value in the luminance image generated by the luminance value selection unit based on a predetermined threshold value indicating whether or not the luminance value is low luminance by the low luminance value correction unit. The luminance value of the pixel having low luminance is corrected by a predetermined coefficient.
Then, the video signal correction program performs the calculation based on the camera characteristics on the luminance value of each pixel of the luminance image corrected by the low luminance value correcting unit by the video pixel value converting unit, thereby performing the signal processing by the camera. Is calculated, and a corrected video signal is generated.

The present invention has the following excellent effects.
According to the first and sixth aspects of the invention, contrast correction is performed on the entire image with respect to the virtual luminance component before image adjustment by the camera, and contrast correction is further partially performed on the low luminance component, thereby The video signal corresponding to the adjustment can be regenerated. For this reason, the present invention can locally correct the contrast of a low-brightness portion without affecting the tone of the entire video by video adjustment for a video signal photographed by a camera, thereby reducing a high-luminance region and a low-brightness region. It is possible to improve the image quality of both details in the luminance range.

  According to the second aspect of the present invention, the low-luminance part is locally applied to the broadcast video signal adjusted in advance by the broadcast camera without affecting the tone of the already adjusted video. Contrast correction can be performed.

According to the third and fourth aspects of the present invention, the radiance component before adjustment of the video signal can be accurately generated according to the known camera characteristics. Contrast correction can be performed, and contrast correction can be performed without affecting the tone of the entire image due to image adjustment.
According to the fifth aspect of the present invention, it is possible to partially perform the contrast correction of the low-luminance portion, and it is possible to improve the image quality of the low-luminance portion as compared with the conventional tone reproduction method.

It is a block block diagram which shows the structure of the video signal correction apparatus which concerns on embodiment of this invention. It is a block block diagram which shows the detailed structure of the virtual radiance correction | amendment means of the video signal correction apparatus which concerns on embodiment of this invention. It is explanatory drawing for demonstrating a camera characteristic parameter, Comprising: (a) is a characteristic curve (camera characteristic curve) of the camera in a certain setting example, (b) is a characteristic curve (virtual) approximated to the characteristic curve of a camera. A typical camera characteristic curve). It is a flowchart which shows the whole operation | movement of the video signal correction apparatus which concerns on embodiment of this invention. It is a flowchart which shows the detailed operation | movement of the contrast correction | amendment of the video signal correction apparatus which concerns on embodiment of this invention.

Embodiments of the present invention will be described below with reference to the drawings.
[Configuration of video signal correction device]
First, the configuration of a video signal correction apparatus according to an embodiment of the present invention will be described with reference to FIG.

  The video signal correction apparatus 1 performs contrast correction of the entire video signal while adjusting the brightness of the low-brightness portion of the video signal. The video signal correction apparatus 1 uses a video signal photographed by a camera as a correction target, and here, as a video signal, a general broadcast video signal expressed in a YCbCr color space is a correction target. For example, this video signal is converted to ITU-R Rec. BT. When the luminance is 8 bits as defined in 709, it is assumed to be a floating-point YCbCr digital signal normalized by setting the value of “16” to “0” and the value of “235” to “1” in the full-range signal.

  The video signal correction apparatus 1 includes a color space conversion unit 10, a virtual radiance calculation unit 20, a virtual radiance correction unit 30, a video pixel value conversion unit 40, a color space inverse conversion unit 50, and a parameter setting unit. 60.

  The color space conversion means 10 converts the color space of the input video signal. Here, the color space conversion means 10 converts the input luminance Y and a video signal in the YCbCr color space composed of two color differences (Cb, Cr) into chromaticity coordinates (x, y) and luminance Y. It is converted into a video signal in the xyY color space.

  Specifically, the color space conversion means 10 converts a video signal in the YCbCr color space into a video signal in the RGB color space by the following formula (6), and a video signal in the RGB color space by the formula (7). Are converted into video signals in the XYZ color space. Furthermore, the color space conversion means 10 converts the video signal in the XYZ color space into the video signal in the xyY color space according to the following equation (8).

As will be described below, the video signal correction apparatus 1 performs contrast correction by adjusting the luminance component. Since the luminance Y of the video signal in the YCbCr color space is a signal having no color, when only the luminance is adjusted, only white with no color is emphasized at high luminance, and color is low at low luminance. Since only black with no taste is an emphasized color, proper contrast correction cannot be performed. Therefore, the video signal correction apparatus 1 uses the color space conversion unit 10 to convert the video signal in the YCbCr color space into the xyY including the luminance with the color added, as can be seen from the equations (6) to (8). The video signal is converted to a color space video signal.
In addition, although said Formula (6)-(8) was expressed by each conversion formula in order to show conversion of each color space intelligibly, said Formula (6)-(8) should be converted as one conversion formula. It is good.

  In this way, the video signal converted by the color space conversion unit 10 is output to the virtual radiance calculation unit 20 and the video pixel value conversion unit 40. Here, the color space conversion unit 10 outputs only the luminance component Y of the video signal converted by the color space conversion unit 10 to the virtual radiance calculation unit 20 and only the color component to the video pixel value conversion unit 40. Output.

The virtual radiance calculation unit 20 calculates a virtual radiance component before signal processing (video adjustment) by the camera is performed from the luminance component of the video signal converted by the color space conversion unit 10.
Here, the virtual radiance calculation means 20 is input via the camera characteristic parameter setting means 61 of the parameter setting means 60, and the camera parameters (cameras) obtained by adjusting the video signal in advance (signal processing such as gamma and knee). The luminance component (virtual radiance component) of the video signal before the signal processing is obtained by calculating the inverse characteristic from the characteristic parameters p and q).

Here, it is assumed that the luminance (Y) component on the video signal converted by the color space conversion means 10 is L _d (x, y), and the virtual radiance component is L ^* (x, y). Note that (x, y) is a two-dimensional pixel space coordinate, L _d (x, y) is a luminance value in the (x, y) coordinate of the luminance component, and L ^* (x, y) is a virtual The luminance values at the (x, y) coordinates of the radiance component are shown.
Then, with the prior correction of the camera, the luminance component L _d (x, y) can be assumed to be the virtual radiance component L ^* (x, y) scaled by the following equation (9).

Here, the camera characteristic parameters p and q are camera characteristic parameters obtained by adjusting the video signal in advance.
Therefore, the virtual radiance calculation means 20 calculates the virtual radiance component L ^* (x, y) by the following equation (10), which is an inverse conversion equation of the equation (9).

The virtual radiance component L ^* (x, y) calculated by the virtual radiance calculation means 20 is output to the virtual radiance correction means 30 and the video pixel value conversion means 40.

  Here, the camera characteristic parameters p and q used by the virtual radiance calculation means 20 will be described with reference to FIG. 3 (refer to FIG. 1 as appropriate). 3A shows a characteristic curve (camera characteristic curve) of a camera in a certain setting example, and FIG. 3B shows a characteristic curve approximated to the characteristic curve of the camera (virtual camera characteristic curve). In FIGS. 3A and 3B, the horizontal axis represents the relative input luminance value (u) and the vertical axis represents the display (output) luminance value (v). In FIG. 3B, in order to compare the two characteristic curves, the camera characteristic curve is indicated by a broken line and the virtual camera characteristic curve is indicated by a solid line.

  FIG. 3A shows a specific setting example in which the setting of the camera that has captured the video signal input to the video signal correction apparatus 1 starts to knee at approximately 95% of the display (output) luminance value (knee). Point) is a characteristic curve assuming that the relative input luminance value reaches 110%, which is the maximum value of the display (output) luminance value at 400%. That is, the characteristic curve of FIG. 3A can be expressed by the function of the following formula (11).

  Here, when u = 0.09025, v = √ (0.9025) = 0.95, which corresponds to the characteristic that knee starts to be applied at about 95% of the display (output) luminance value. Further, when u = 4, v = √ (1.21) = 1.1, which corresponds to the characteristic that the relative input luminance value reaches 400% when the relative luminance value reaches 400%, which is the maximum value of the display (output) luminance value.

  A characteristic curve that approximates the camera characteristic curve of FIG. 3A is a virtual camera characteristic curve shown in FIG. That is, the virtual camera characteristic curve of FIG. 3B is defined as the following expression (12), and p and q are determined in advance as camera characteristic parameters so as to approximate the camera characteristic curve of FIG.

  In this case, v = 1.1 when u = 4, and p and q are determined so as to be a curve that approximates the camera characteristic curve of FIG. For example, in this case, values of p = 3 and q = 3.575 can be set.

As described above, the adjustment performed in the camera is set in advance as the camera characteristic parameter, so that the virtual radiance calculation unit 20 can perform the adjustment before the adjustment from the luminance component L _d (x, y) of the video signal. The virtual radiance component L ^* (x, y) can be calculated (see the above formula (10)).
Returning to FIG. 1, the description of the configuration of the video signal correction apparatus 1 will be continued.

  The virtual radiance correction unit 30 performs contrast correction on the virtual radiance component calculated by the virtual radiance calculation unit 20. The virtual radiance correction means 30 performs contrast correction for the entire image and also emphasizes contrast correction for the low-luminance portion. The luminance value (corrected luminance value) corrected by the virtual radiance correction unit 30 is output to the video pixel value conversion unit 40.

Here, the detailed configuration of the virtual radiance correction means 30 will be described with reference to FIG. As shown in FIG. 2, the virtual radiance correction means 30 includes a filter means 31 (31 ₁ , 31 ₂ ,..., 31 _n ) and an activity calculation means 32 (31 ₁ , 31 ₂ ,..., 31 _n-1 ). A kernel determining unit 33, a contrast correction pixel selecting unit 34, and a low luminance value correcting unit 35.

The filter unit 31 applies a Gaussian kernel (Gaussian filter) to the input virtual radiance component L ^* (x, y), and uses the Gaussian kernel at the luminance of the surrounding pixels with respect to the luminance at the pixel coordinates (x, y). Is used to perform a convolution operation. In this way, by applying the Gaussian kernel, it is possible to smooth the luminance value by removing the noise by increasing the weight as the pixel is closer to the target pixel.

Further, here, the filter means 31 is constituted by a plurality of filter means 31 ₁ , 31 ₂ ,..., 31 _{n to} which Gaussian kernels having different predetermined sizes (scales) are applied. The number of the filter means 31 is arbitrary, but is eight, for example.
The filter means 31 ₁ , 31 ₂ ,..., 31 _n respectively convolve the virtual radiance component L ^* (x, y) using a Gaussian kernel R as shown in the following equation (13). A smoothed luminance value (filtered luminance value) V ^* (x, y, s _i ) is calculated by calculation.

The (α _i s _i ) of the Gaussian kernel R is a coefficient for adjusting the scale of the Gaussian kernel, and α _i is a fixed value independent of i. Further, s _i is a coefficient that is different for each filter means 31 ₁ , 31 ₂ ,..., 31 _n .
Here, the Gaussian kernel filter unit 31 ₁ is Gaussian kernel _{R 1,} Gaussian kernel _{R 2} filter unit 31 ₂ using a coefficient _{s 2} using the coefficients _{s 1,} ..., the filter means 31 _n is using the coefficients _{s n} The respective (filtered) luminance values V ^* (x, y, s _i ) are calculated by R _n .

Also, here, a kernel in which scaling of each Gaussian kernel is sequentially scaled to 1.6 times larger than i = 1 is used, s ₁ = 0.35 in the smallest Gaussian kernel R ₁ , and s _i Let the kernel be 1.6 times larger.

Incidentally, to increase the size of the s _i by 1.6 times, generally, when approximating the human brightness perception model using DOG (Difference of Gaussian), the ratio of the scale of adjacent Gaussian kernel 1.6 (see, for example, MARR, D. 1982. Vision, a computational investigation into the human representation and processing of visual information. WH Freeman and Company, San Francisco.).

The smoothed luminance value V ^* (x, y, s _i ) is output to the activity calculation unit 32 and the contrast correction pixel selection unit 34. In addition, the brightness value V ^* calculated by two Gaussian kernels having close scales among the filter units 31 ₁ , 31 ₂ ,..., 31 _n is output to the same activity calculation unit 32. For example, the filter means 31 ₁ and 31 ₂ output the brightness value V ^* (x, y, s ₁ ) and the brightness value V ^* (x, y, s ₂ ) to the same activity calculation means 32 ₁ , respectively. filter means ₃₁ 2, 31 _3, and outputs the luminance value in the same activity calculation means ^{_{32 2 V * (x, y}} , s 2) and the luminance value _{^{V * (x, y, s}} 3) and, respectively. Similarly, the filter means 31 _n−1 and 31 _n are connected to the same activity calculation means 32 _n−1 by the luminance value V ^* (x, y, s _n−1 ) and the luminance value V ^* (x, y, s _n). ) And.

  The activity calculation means (change amount calculation means) 32 is used to change the luminance value associated with the scale based on the luminance value (filtered luminance value) calculated by the Gaussian kernel having a different scale (size) in the filter means 31. The degree (activity; change amount) is calculated.

Here, the activity calculation means 32 is composed of a plurality of activity calculation means 32 ₁ , 32 ₂ ,..., 32 _n−1 . The number of activity calculation means 32 is arbitrary, but is configured to be one less than the filter means 31.

The activity calculation means 32 ₁ , 32 ₂ ,..., 32 _n−1 each have two (filtered) luminance values V ^* (x, y, s _i ) in the same pixel calculated by a Gaussian kernel having a close scale. , V ^* (x, y, s _{i + 1} ), and the degree of change in brightness value (Activity (x, y, s _i )) is calculated by the following equation (14).

  Here, φ is a sharpness parameter that works in a direction in which the edge is emphasized when the value is reduced. For example, if the value of φ is reduced, the amount of change with respect to the luminance scale to be evaluated is adjusted, and the edge enhancement can be made steeper.

Further, a is a key value, and is a parameter that works in a direction in which an image becomes subjectively dark when the value is decreased.
Here, for example, φ = 8.0 and a = 0.18. The sharpness parameter φ and the key value a are predetermined values here, but may be configured to be set as parameters from the outside.
Further, since the equation (14) is the degree of change in the luminance value, the denominator “2 ^φ as _i ⁻² ” is not necessarily required.

The activity calculating means 32 ₁ calculates Activity (from the luminance value V ^* (x, y, s ₁ ) and the luminance value V ^* (x, y, s ₂ ) calculated by the filtering means 31 ₁ and 31 _2. x, y, _{s 1)} is calculated, the activity calculation unit 32 _2, the filter unit ₃₁ 2, 31 luminance values calculated in ^{_{3 V * (x, y,}} s 2) and the luminance value ^V * (x, y , S ₃ ), Activity (x, y, s ₂ ) is calculated. Similarly, the activity calculation means 32 _n-1 has the brightness value V ^* (x, y, s _n-1 ) and the brightness value V ^* (x, y, s) calculated by the filter means 31 _n-1 , 31 _{n. n} ), Activity (x, y, s _n-1 ) is calculated.
The degree of change in luminance value (Activity (x, y, s _i )) calculated by the activity calculating means 32 ₁ , 32 ₂ ,..., 32 _n−1 is output to the kernel determining means 33.

The kernel determination unit 33 is based on the degree of change in luminance value (change amount; activity) that varies depending on the scale of the Gaussian kernel calculated by the plurality of activity calculation units 32 (32 ₁ , 32 ₂ ,..., 32 _n-1 ). This determines which Gaussian kernel to apply.

The kernel determination unit 33 calculates a Gaussian kernel with the smallest scale in which the degree of change (Activity) calculated by the plurality of activity calculation units 32 is greater than a predetermined threshold. The Gaussian kernel to be applied to the pixel is determined.
That is, the kernel determination unit 33 determines s _m (m = 1, 2,..., N) that satisfies the following expression (15).

Here, ε is a threshold value serving as a criterion for determining that the luminance value has greatly changed. For example, ε is set to “0.05”. Here, the threshold value ε is set to a predetermined value, but may be configured to be set as a parameter from the outside.
As a result, the kernel determining unit 33 can specify a Gaussian kernel that further increases the change in contrast by changing the scale of the Gaussian kernel.

The kernel determination unit 33 outputs an identifier m (m = 1, 2,..., N) that identifies the determined Gaussian kernel to the contrast correction pixel selection unit 34. Here, an identifier m (m = 1,2, ..., n) , respectively, the filter means ₃₁ 1, ..., it shows a Gaussian kernel applied at 31 _n.

Contrast correction pixel selecting means (luminance value selecting means) 34, a plurality of filter means _{31 (31 1, 31 2,} ..., 31 n) after Gaussian kernel applied input from the luminance value _{^{V * (x, y, s}} m ) (M = 1, 2,..., N), the luminance value to which the Gaussian kernel determined by the kernel determining means 33 is applied is selected.

The contrast correction pixel selection unit 34 selects the luminance value determined by the kernel determination unit 33 to which the Gaussian kernel that increases the contrast change is applied for all the pixels, and thus performs contrast correction on the entire luminance image. It ’s gone.
Thus selected luminance values, i.e., the contrast in the entire luminance image (global) corrected luminance value ^{V * (x, y, s} m) is output to the low luminance value correcting means 35.

The low luminance value correction means 35 is selected by the contrast correction pixel selection means 34 and is partially (local) in the luminance image composed of the luminance values V ^* (x, y, s _m ) whose contrast is corrected globally. In addition, the contrast of the low luminance part is corrected.

Note that the luminance image composed of the luminance values V ^* (x, y, s _m ) whose contrast has been corrected by the contrast correction pixel selection unit 34 is obtained by the virtual radiance calculation unit 20 (see FIG. 1) by the above formula ( This is generated from the virtual radiance component L ^* (x, y) calculated by 10).
However, in the formula (9), the contrast-corrected luminance value V ^* (x, y, s _m ) is similar to the formula (5) of the conventional method described in “Problems to be solved by the invention”. In some cases, the contrast correction effect in the low-luminance portion is not sufficient.

Therefore, the low luminance value correction means 35, via the correction amount adjustment parameter setting means 62 of the parameter setting means 60 shown in FIG. 1, the threshold value (correction amount adjustment parameter α) for specifying the low luminance component and the correction amount (correction). By setting the amount adjustment parameter β), the set correction is performed on the low luminance component based on the set threshold value, and the contrast correction of the low luminance component is performed.
Specifically, the low luminance value correcting means 35 calculates a corrected luminance value V _s ^* (x, y, α, β) obtained by performing contrast correction on the low luminance portion by the following equation (16).

Here, the correction amount adjustment parameters α and β are parameters that can be appropriately adjusted by an operator operating the video signal correction apparatus 1.
This correction amount adjustment parameter α is a threshold value that serves as a reference for low luminance, and is a value that indicates which low luminance component or less the luminance value the operator wants to adjust. Here, the brightness value ^{V * (x, y, s} m) is, if set correction amount adjustment value than the parameter α is small, so that the correction is performed.

The correction amount adjustment parameter β is a coefficient that is a correction amount for correcting the low luminance component, and is also a value indicating how much contrast the operator wants to apply to the low luminance portion. The correction amount adjustment parameter β works in a direction in which the contrast is increased if the value is large, and works in a direction in which the contrast is weak when the value is small.
The correction amount adjustment parameters α and β can be set in advance, but the operator can also perform adjustment while visually confirming the corrected image in real time. is there.

As described above, the virtual radiance correction unit 30 converts the luminance component of the video signal into the virtual radiance component before the camera adjustment is performed once. Then, the virtual radiance correction unit 30 performs contrast correction on the entire luminance image (global) with respect to the converted virtual radiance component, and further performs contrast correction on the low luminance part (local).
Returning to FIG. 1, the description of the configuration of the video signal correction apparatus 1 will be continued.

The image pixel value conversion means 40 includes the virtual radiance component L ^* (x, y) calculated by the virtual radiance calculation means 20 and the corrected brightness value V _s ^* (x, y) corrected by the virtual radiance correction means 30. Based on y, α, β), a corrected video signal converted into a luminance component of the video signal adjusted by the camera is generated.

Here, the video pixel value converting means 40 is configured to set the virtual radiance component L ^* (x, y), the corrected brightness value V _s ^* (x, y, α, β), and the camera characteristic parameter setting of the parameter setting means 60. Based on the camera parameters (camera characteristic parameters p, q) input in advance via the means 61 and adjusted in advance, the corresponding luminance component is generated after the camera adjustment, and the contrast-corrected video is generated. Generate a signal.
Specifically, the video pixel value conversion means 40 generates the luminance component L _d ^* (x, y, α, β) of the video signal whose contrast has been corrected by the following equation (17).

The luminance component generated by the video pixel value conversion means 40 is output to the color space inverse conversion means 50 as a corrected video signal together with the color components.
Here, the corrected luminance value V _s ^* (x, y, α, β) after contrast correction is used for the denominator of the equation (17), and the virtual radiance component L ^* (x, y) before contrast correction is used for the numerator. The reason for using is described briefly.

It is assumed that the pixel (x, y) that is currently focused on is a pixel whose luminance is slightly reduced in a bright surrounding area. In this case, the virtual radiance correction means 30 outputs V _s ^* to which a Gaussian kernel having a small scale is applied in accordance with the luminance change. This V _s ^* is larger than the virtual radiance component L ^{* due} to the Gaussian weighted average effect with surrounding bright pixels. Thus, by using this V _s ^* for the denominator of the equation (17), the luminance value L _d ^* of the target pixel (x, y) becomes a value smaller than the original luminance value.

On the other hand, since the surrounding pixels of the target pixel are bright, a Gaussian kernel having a large scale is applied, and the value is not different from the original virtual radiance component L ^* even by the Gaussian weighted average.

Thus, the image pixel value conversion means 40 uses the corrected luminance value V _s ^* after contrast correction as the denominator of the equation (17) and the virtual radiance component L ^* before contrast correction as the numerator, so that the pixel of interest Only the luminance value of is reduced more than the original luminance value compared to the surroundings, and the contrast can be enhanced.

  The color space inverse conversion means 50 converts the color space of the corrected video signal whose luminance component has been converted by the video pixel value conversion means 40. Here, the color space inverse conversion unit 50 converts the corrected video signal in the xyY color space output from the video pixel value conversion unit 40 into a video signal in the same YCbCr color space as the input video signal.

  Specifically, the color space inverse conversion means 50 converts the video signal in the xyY color space into the video signal in the XYZ color space by the following formula (18), and the video in the XYZ color space by the formula (19). The signal is converted into a video signal in the RGB color space. Further, the color space inverse conversion means 50 converts the video signal in the RGB color space into the video signal in the YCbCr color space by the following equation (20).

In addition, although said Formula (18)-(20) was expressed with each conversion formula in order to show conversion of each color space intelligibly, converting said Formula (18)-(20) as one conversion formula. It is good.
The color space reverse conversion unit 50 performs the reverse conversion of the color space conversion unit 10. Therefore, if the color space conversion means 10 performs different color space conversion, it is needless to say that the color space reverse conversion means 50 is configured to perform reverse conversion corresponding to the conversion.

  The parameter setting means 60 inputs various parameters from the outside. Here, the parameter setting unit 60 includes a camera characteristic parameter setting unit 61 and a correction amount adjustment parameter setting unit 62.

The camera characteristic parameter setting means 61 inputs camera characteristic parameters p and q indicating the characteristics of the camera in which the input video signal is adjusted (signal processing) in advance by signal processing by the camera. The input camera characteristic parameters p and q are output to the virtual radiance calculation means 20 and the video pixel value conversion means 40.
Since the camera characteristic parameters p and q have been described in the virtual radiance calculation means 20, the description thereof is omitted here.

The correction amount adjustment parameter setting unit 62 includes a threshold value (correction amount adjustment parameter α) for specifying a low luminance component for performing contrast correction of a low luminance portion, and a coefficient (a correction amount for correcting the low luminance component). The correction amount adjustment parameter β) is input. The correction amount adjustment parameters α and β input here are output to the virtual radiance correction means 30.
The correction amount adjustment parameters α and β have been described in the low luminance value correction unit 35 (see FIG. 2) of the virtual radiance correction unit 30 and thus will not be described here.

  With this configuration, the video signal correction apparatus 1 globally performs contrast correction on the entire image from signal processing such as knee correction processing or color correction that has already been performed by the camera, and further, The contrast correction can be partially performed on the low luminance part.

  Since the video signal correction device 1 temporarily converts the luminance component of the video signal into a virtual radiance component and performs contrast correction, the video signal correction device 1 performs contrast correction on the video signal before camera adjustment. Thus, the contrast can be improved without affecting the adjustment contents by the camera, that is, without affecting the tone of the entire image.

  The video signal correction apparatus 1 can be configured by a general CPU, RAM, ROM, and the like, and can be realized by a program (video signal correction program) that causes a computer to function as each of the above-described means.

Although one embodiment of the present invention has been described above, the present invention is not limited to this embodiment.
Here, the video signal to be corrected is a video signal in the YCbCr color space, but even a video signal in other color space can be applied to the present invention as long as the luminance component can be extracted. For example, general video signals such as video signals in YUV color space and video signals in RGB color space can be used. In this case, the color space conversion unit 10 and the color space reverse conversion unit 50 may perform conversion / inverse conversion so as to correspond to each color space.

  Here, the video signal to be corrected is a color video signal. However, if the video signal to be corrected is a black and white video signal, it is composed only of a luminance signal. The spatial inverse transform unit 50 may be omitted from the configuration.

[Operation of video signal correction device]
Next, the operation of the video signal correction apparatus according to the embodiment of the present invention will be described with reference to FIG.

  First, the video signal correction apparatus 1 converts the color space of the input video signal by the color space conversion means 10 (step S1). Here, the color space conversion means 10 converts a video signal in the YCbCr color space into a video signal in the xyY color space. In other words, the color space conversion means 10 converts the video signal in the YCbCr color space into the video signal in the xyY color space according to the equations (6) to (8).

Then, the video signal correction apparatus 1 uses the virtual radiance calculation unit 20 to convert the luminance component L _d (x, y) in the video signal in the xyY color space, which has been color space converted in step S1, to the camera characteristic parameter setting unit. The virtual radiance component L ^* (x, y) is calculated by scaling (reverse conversion) with the camera characteristic parameters p and q of the camera input in advance through 61 and adjusted in advance. Step S2). That is, the virtual radiance calculation means 20 calculates the virtual radiance component L ^* (x, y) by the above equation (10).

As described above, the virtual radiance component L ^* (x, y) calculated by the virtual radiance calculation means 20 is scaled (reversely converted) with the camera characteristic parameters p and q of the camera obtained by adjusting the video signal in advance. Therefore, this corresponds to the luminance component of the video signal before camera adjustment.

Then, the video signal correction apparatus 1 contrast-corrects the scaled virtual radiance component L ^* (x, y) calculated in step S2 by the virtual radiance correction means 30, and sets a correction amount adjustment parameter. Based on the correction amount adjustment parameters α and β input by the means 62, the contrast is partially adjusted for the low luminance portion (step S3). The contrast correction operation in step S3 will be described later in detail with reference to FIG.

Thereafter, the video signal correction device 1 uses the video pixel value conversion means 40 to set the camera characteristic parameter setting means from the corrected luminance value V _s ^* (x, y, α, β) of the virtual radiation component whose contrast has been corrected in step S3. 61. A corrected video signal that is input via 61 and corrects the luminance component of the video signal after being scaled by the camera by reproducing the adjustment with the camera characteristic parameters p and q of the camera that has been adjusted in advance. Is generated (step S4). That is, the video pixel value conversion means 40 generates a corrected video signal according to the above equation (17).

  Then, the video signal correction apparatus 1 converts the color space of the corrected video signal generated in step S4 by the color space inverse conversion means 50 (step S5). Here, the color space inverse conversion means 50 converts the corrected video signal in the xyY color space into a corrected video signal in the YCbCr color space. That is, the color space inverse conversion means 50 converts the corrected video signal in the xyY color space into the corrected video signal in the YCbCr color space according to the equations (18) to (20).

  Through the above operation, the video signal correction apparatus 1 performs contrast correction on the virtual video signal before adjustment even if the video signal is adjusted by the camera. Contrast correction can be performed without affecting the adjustment of the characteristics.

(Operation of virtual radiance correction means)
Next, the operation of the virtual radiance correction means 30 will be described with reference to FIG. This operation corresponds to the operation in step S3 in FIG.

First, the virtual radiance correction means 30 applies a plurality of filter means 31 ₁ , 31 ₂ , to the virtual radiance component L ^* (x, y) calculated by the virtual radiance calculation means 20 (see FIG. 1). .., 31 _n are used to perform convolution operations by applying Gaussian kernels having different sizes (scales) (step S31). That is, the filter means 31 ₁ , 31 ₂ ,..., 31 _n calculate the smoothed luminance value V ^* (x, y, s _i ) by the convolution operation of the equation (13). As a result, a plurality of luminance values calculated with Gaussian kernels having different scales are generated for the same two-dimensional pixel position (x, y).

The virtual radiance correction means 30, a plurality of activity calculation means _{32 1,} 32 2, ..., ₃₂ by _n-1, of the two the size of the scale is calculated in nearly Gaussian kernel luminance value ^V * (x, The degree of change in luminance value (activity: Activity (x, y, s _i )) is calculated from y, s _i ), V ^* (x, y, s _{i + 1} ). That is, the activity calculation means 32 ₁ , 32 ₂ ,..., 32 _n−1 each calculate an activity according to the equation (14). This activity makes it possible to determine which scale of Gaussian kernel is applied for each pixel to obtain a contrast effect.

Therefore, the virtual radiance correction unit 30 uses the kernel determination unit 33 to first obtain a luminance value to which the Gaussian kernel having the smallest scale s ₁ is applied and a luminance value to which the Gaussian kernel having the _second smallest scale s ₂ is applied. the obtained Activity (x, y, _{s 1)} it may determine whether a value greater than the threshold value ε a predetermined (step _S33 1), is greater (Yes in step S33 _1), Gaussian scale _{s 1} It is determined to apply the kernel (step S34 ₁ ).
Similarly, from Activity (x, y, s ₂ ) to Activity (x, y, s _n−1 ), it is sequentially determined whether or not the value is larger than the threshold value ε. Decide to apply the Gaussian kernel.

And all Activity is, if the result is not a value larger than the threshold epsilon (No at step _{S33 n-1),} decides to apply a Gaussian kernel of the largest scale _{s n} (step _S34 n).
Then, the virtual radiance correction unit 30 uses the luminance value V ^* calculated in step S31 as the luminance value to which the Gaussian kernel determined in any of steps S34 _{1 to} S34 _n is applied by the contrast correction pixel selection unit 34 ^. A selection is made from among (x, y, s _i ) (step S35).

Thereafter, the virtual radiance correction unit 30 compares the luminance value V ^* (x, y, s _m ) selected in step S35 with the value of the correction amount adjustment parameter α by the low luminance value correction unit 35 (step S35). S36). When the luminance value V ^* (x, y, s _m ) is smaller than the value of the correction amount adjustment parameter α (corresponding to the case where the luminance value V ^* is low luminance) (Yes in step S35), the luminance value Contrast correction (local contrast correction) is further performed on V ^* (step S37). That is, the low luminance value correction means 35 performs local contrast correction according to the equation (16).

With the above operation, the virtual radiance correction unit 30 performs contrast correction on the entire image with respect to the virtual radiance component L ^* (x, y), and further locally performs contrast correction on the low luminance part. Can do.
As described above, the video signal correction apparatus 1 can perform contrast correction on the video signal captured by the camera without affecting the tone of the entire video, and further, locally applies the low-luminance portion. The contrast can be corrected.

DESCRIPTION OF SYMBOLS 1 Video signal correction apparatus 10 Color space conversion means 20 Virtual radiance calculation means 30 Virtual radiance correction means 31 Filter means 32 Activity calculation means (change amount calculation means)
33 Kernel determination means 34 Contrast correction pixel selection means (luminance value selection means)
35 Low luminance value correcting means 40 Video pixel value converting means 50 Color space inverse converting means 60 Parameter setting means 61 Camera characteristic parameter setting means 62 Correction amount adjustment parameter setting means

Claims (6)

A video signal correction device for correcting the contrast of a video signal,
Based on the camera characteristics indicating the characteristics of the video signal that has undergone signal processing in the camera, a virtual radiation before the signal processing is performed by performing an operation based on the inverse characteristics of the camera characteristics on the luminance component of the video signal. Virtual radiance calculating means for calculating a luminance component;
Filter means for calculating a filtered luminance value by performing a convolution operation with a plurality of predetermined Gaussian kernels for each pixel for the radiance component calculated by the virtual radiance calculation means;
A change amount for calculating a change amount for each of two sets of filtered luminance values calculated by adjacent Gaussian kernels in a plurality of filtered luminance values in the same pixel calculated by the filter means. A calculation means;
Kernel determination means for determining a minimum Gaussian kernel in which the change amount calculated by the change amount calculation means is larger than a predetermined threshold as a Gaussian kernel to be applied to the pixel;
By selecting a filtered luminance value calculated by applying the Gaussian kernel determined by the kernel determining unit from a plurality of filtered luminance values at the same pixel calculated by the filtering unit as a luminance value at the pixel. A luminance value selection means for generating a luminance image;
Based on a predetermined threshold value indicating whether or not the luminance value is low luminance, in the luminance image generated by the luminance value selecting means, the luminance value of a pixel having a low luminance value is determined by a predetermined coefficient. Low luminance value correcting means for correcting;
The luminance value of each pixel of the luminance image corrected by the low luminance value correcting means is calculated according to the camera characteristics, thereby calculating the luminance value after the signal processing and generating a corrected video signal. Video pixel value conversion means for
A video signal correction apparatus comprising: The video signal is a video signal expressed in a YCbCr color space that is a broadcast video signal,
Color space conversion means for converting the input video signal into an xyY color space video signal;
Color space reverse conversion means for reversely converting the corrected video signal generated by the video pixel value conversion means into a video signal in the YCbCr color space;
The video signal correction apparatus according to claim 1, further comprising: The virtual radiance calculating means includes
Camera characteristic parameters p and q approximating the camera characteristic as a characteristic in which the radiance component L ^* is a luminance component of qL ^* / (1 + pL ^* ) are input from outside, and the luminance component L _d of the video signal 3. The video signal correction apparatus according to claim 1, wherein a virtual radiance component L ^* before the signal processing is calculated by calculating L _d / (q−pL _d ). . The video pixel value converting means includes
When the luminance value of each pixel of the luminance image corrected by the low luminance value correcting means is V _s ^* and the virtual radiance component is L ^* , qL ^* is used using the camera characteristic parameters p and q ^. The video signal correction apparatus according to claim 3, wherein a luminance value after correction is calculated by calculating / (1 + pV _s ^* ). The low luminance value correcting means is
When the luminance value before correction is V ^* , α that is a predetermined threshold indicating whether or not the luminance value is low luminance and β that is the coefficient are input from the outside, and before the correction, 5. The corrected luminance value is calculated by calculating V ^* −β (1−V ^* / α) when the luminance value V ^* is small with respect to the threshold value α. The video signal correction apparatus according to any one of the above. In order to correct the contrast of the video signal,
Based on the camera characteristics indicating the characteristics of the video signal that has undergone signal processing in the camera, a virtual radiation before the signal processing is performed by performing an operation based on the inverse characteristics of the camera characteristics on the luminance component of the video signal. Virtual radiance calculating means for calculating a luminance component;
Filter means for calculating a filtered luminance value by performing a convolution operation on a radiance component calculated by the virtual radiance calculation means for each pixel with a plurality of predetermined Gaussian kernels,
A change amount for calculating a change amount for each of two sets of filtered luminance values calculated by adjacent Gaussian kernels in a plurality of filtered luminance values in the same pixel calculated by the filter means. Calculation means,
Kernel determination means for determining a minimum Gaussian kernel in which the amount of change calculated by the change amount calculation means is greater than a predetermined threshold as a Gaussian kernel to be applied to the pixel;
By selecting a filtered luminance value calculated by applying the Gaussian kernel determined by the kernel determining unit from a plurality of filtered luminance values at the same pixel calculated by the filtering unit as a luminance value at the pixel. A luminance value selection means for generating a luminance image;
Based on a predetermined threshold value indicating whether or not the luminance value is low luminance, in the luminance image generated by the luminance value selecting means, the luminance value of a pixel having a low luminance value is determined by a predetermined coefficient. Low luminance value correction means for correcting,
The luminance value of each pixel of the luminance image corrected by the low luminance value correcting means is calculated according to the camera characteristics, thereby calculating the luminance value after the signal processing and generating a corrected video signal. Video pixel value conversion means
A video signal correction program characterized by functioning as

Images