JP2020003998A - Tone mapping circuit and tone mapping method - Google Patents

Abstract

To provide a tone mapping circuit configured to reduce circuit scale, power consumption, cost, and computation time without impairing continuity of output images, and a tone mapping method.SOLUTION: A tone mapping circuit includes: a global feature quantity calculation circuit which calculates global feature quantity of an HDR image of a current frame; a local feature quantity calculation circuit which calculates local feature quantity of a block image, and calculates local feature quantity of pixels by interpolating local feature quantities of a block image of a previous frame; a gain calculation circuit which calculates feature quantity of gain and total on the basis of global feature quantity of a previous frame, local feature quantity of pixels, and target luminance; and a gain application circuit which applies the feature quantity of gain and total to data of pixels to be processed.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a tone mapping circuit and a tone mapping method for performing tone mapping for displaying an HDR (High Dynamic Range) image having a wide dynamic range on a display having a narrow dynamic range.

  Formats such as HDR10 and HLG (Hybrid Log Gamma) have been proposed and used to record HDR images having a wider dynamic range than in the era when CRTs (Cathode Ray Tubes) were used. In the HDR 10, the absolute luminance of an image to be displayed is recorded. In the HLG, a relative value with the maximum luminance that can be displayed on the display as the maximum value is recorded similarly to the SDR (Standard Dynamic Range) image, but the low luminance area is the same as the SDR image, and the high luminance area is recorded. Is an exponential function.

  On the other hand, a conventional image in which pixel data is converted using an inverse conversion curve of a square gamma curve of 2.2, which is a standard gamma value γ of a display, is called an SDR image.

  A method of performing conversion for displaying the HDR image having a wide dynamic range on a display having a narrow dynamic range is called tone mapping, and a circuit for performing the tone mapping is a tone mapping circuit. However, since the display has various performances such as maximum luminance and minimum luminance, it is impossible to reproduce all the gradations of the HDR image in the limited luminance area, and it is not possible to reproduce all the gradations of the bright part or the dark part. The key will be lost.

  A gradation region (gradation range) to which the average luminance or a large number of pixel luminances belong is calculated for each image or a block image obtained by dividing the image, and a region in which gradation is destroyed is reduced by compressing and displaying the region. be able to. When processing using such local information is performed, processing is performed using not the luminance value itself but its logarithmic value for reasons such as human visual characteristics. The PQ (Perceptual Quantization) curve used in the HDR 10 and the high luminance area of the HLG conversion function take in such characteristics, and the luminance signal is expressed using a function close to a logarithmic function.

  For example, as a simple conversion function when tone mapping is performed within the range of y = [0, 1] regardless of image information, a simple conversion function of y = x / (1 + x) and a piecewise linear function and so on.

  As a conversion function incorporating local information, as described in Non-patent Document 1, a low-pass filter Ri (x, y, s) (s is a coefficient representing the size of the filter) to calculate an image Vi (x, y, s) shown in the following equation (2). Functions and the like having the relationship shown in the following equation (3) are known.

  In Non-Patent Document 2, tone mapping is performed by calculating a gradient in a logarithmic space and compressing the gradient.

  Non-Patent Document 3 proposes a method of creating images of a plurality of scales, performing S-shaped nonlinear conversion on the images of each scale, and then combining the images.

  Non-Patent Document 4 also proposes a method of achieving high-speed processing by performing processing on images of respective scales in parallel.

  In Patent Document 1, tone mapping is performed by generating at least two different scale ratio images for a logarithmic input image and adding the generated scale ratio image to the logarithmic input image.

  In Patent Literature 2, first, a logarithmic global tone mapped luminance image is calculated from a histogram of logarithmic luminance pixel values in an HDR image, a local log ratio image is calculated therefrom, and a second logarithm that is down-scaled based thereon is calculated. The image is encoded by repeating the process of generating the tone mapping image and the second ratio image, and a JPEG-HDR image is created.

  In Patent Literature 3, a transmitting unit transmits transmission video data together with auxiliary information (level mapping curve information and / or electro-optical conversion characteristic information), a receiving unit receives the transmission video data, and a processing unit transmits the level of the transmission video data. Is converted based on the auxiliary information. In FIGS. 15, 18, and 19, after performing the HDR inverse conversion on the received image, the level mapping is performed using an electro-optical transfer function (EOTF).

JP-A-2005-212978 JP-T-2015-508589A WO 2015/111467

Erik Reinhard, 3 others, "Photographic tone reproduction for digital images", [online], 2002-07-01, ACM New York, NY, USA, "ACM Transactions on Graphics (TOG)-Proceedings of ACM SIGGRAPH 2002", Volume 21 Issue 3, July 2002, [Searched February 26, 2018], Internet <URL: https://dl.acm.org/citation.cfm?id=566575> Raanan Fattal, 2 others, "Gradient Domain High Dynamic Range Compression", [online], "School of Computer Science and Engineering", "The Hebrew University of Jerusalem", [Search February 26, 2018], Internet < URL: http://www.cs.huji.ac.il/~danix/hdr/> Sylvain Paris, 2 others, "Local Laplacian Filters: Edge-aware Image Processing with a Laplacian Pyramid", [online], March 2015, "ACM Transactions on Graphics (Proceedings of SIGGRAPH 2011)", "Communications of the ACM", Vol. 58, No. 3, Internet <URL: https://people.csail.mit.edu/sparis/publi/2011/siggraph/> Mathieu Aubry, 4 others, "Fast Local Laplacian Filters: Theory and Applications", [online], 2014, "ACM Transactions on Graphics", [Search February 26, 2018], Internet <URL: http: // www.di.ens.fr/~aubry/llf.html>

  When a uniform conversion function that does not depend on information on an image or a block image is used, white portions are lost in a bright portion, and black portions are lost in a dark portion. Therefore, mapping that allocates many bits to the gradation area to which many pixels belong according to the luminance distribution of the screen or block image has been proposed and implemented, but the continuity of the output image, circuit scale, calculation Compatibility with computational costs in terms of time and power consumption has not been easy. In particular, in applications such as embedding in a mobile display, a lighter calculation load is required rather than a strict calculation result.

  If the processing is not completed within one frame, it is necessary to store the image to be processed in the frame memory, which increases the circuit scale, power consumption, and cost.

  An object of the present invention is to provide a tone mapping circuit and a tone mapping method that can reduce the circuit scale, power consumption and cost, and reduce the calculation time without impairing the continuity of an output image.

In order to achieve the above object, the present invention calculates a plurality of global feature amounts, which are feature amounts of the HDR image, based on data of all pixels of the HDR image of the current frame, and holds the calculated global feature amount for one frame time, A global feature amount calculating circuit that outputs a global feature amount of the HDR image one frame before;
For each of a plurality of block images obtained by dividing the HDR image of the current frame in the horizontal direction and the vertical direction, a local feature amount which is a feature amount of the block image is calculated based on data of all pixels of the block image. By interpolating between the local feature values of the block image of the HDR image one frame before and the surrounding block image corresponding to the pixel position to be processed of the HDR image of the current frame, and A local feature calculating circuit that calculates a local feature of a pixel of the block image of the HDR image one frame before corresponding to the pixel position to be processed;
A plurality of global feature values of the HDR image of the previous frame, local feature values of pixels of the HDR image of the previous frame corresponding to the pixel position to be processed, and an output of a display corresponding to the plurality of global feature values A plurality of gains corresponding to the plurality of global feature amounts and the plurality of target brightnesses for each block image of the HDR image one frame before, based on a plurality of target brightnesses representing brightness; and the plurality of global features. A gain calculation circuit for calculating a plurality of total feature amounts corresponding to the amount and the local feature amount of the pixel of the HDR image one frame before corresponding to the pixel position to be processed;
For the data of the pixel to be processed, a gain corresponding to the data of the pixel to be processed is selected from among a plurality of gains for the block image of the HDR image one frame before corresponding to the pixel position of the pixel to be processed. And a gain application circuit that calculates pixel data after gain application by applying the feature amount of (1).

  Here, the global feature amount calculation circuit creates a histogram representing the number of pixels belonging to each region of the pixel data of the HDR image from the data of all the pixels of the HDR image, and uses the histogram to generate the histogram. It is preferable to calculate a plurality of global feature values.

  The global feature amount calculation circuit creates a cumulative histogram representing the cumulative number of pixels belonging to each area of the pixel data of the HDR image from the data of all the pixels of the HDR image, and uses the cumulative histogram. It is preferable to calculate the plurality of global feature amounts in this way.

  In addition, the global feature amount calculation circuit uses the cumulative histogram to determine, as the plurality of global feature amounts, a plurality of pixels specified by a plurality of thresholds with respect to the total number of pixels of the HDR image. It is preferable to output a representative value of data included in a plurality of data areas corresponding to the numbers.

  In addition, the global feature amount calculation circuit outputs, as one of the plurality of global feature amounts, a representative value of data included in a data region corresponding to a median pixel number of all pixels of the HDR image. Is preferred.

  In addition, the global feature amount calculation circuit is configured to determine, for each of a plurality of area images obtained by dividing the HDR image of the current frame in a horizontal direction and a vertical direction, based on data of all pixels of the area image, A plurality of area global feature amounts, which are feature amounts, are sequentially calculated and held for one frame time, and the pixel position of the processing target is selected from among the area global feature amounts of all the area images of the HDR image one frame before. It is preferable to sequentially output the area global feature amount of the area image of the HDR image one frame before corresponding to the above as the global feature amount.

  Further, it is preferable that the local feature value calculation circuit calculates, as a local feature value of the block image, an average value, a median value, a maximum value, a minimum value, or a representative pixel value of data of all pixels of the block image. .

  Further, the gain calculation circuit weights and adds each of the plurality of global feature amounts to a local feature amount of a pixel of the HDR image one frame before corresponding to the pixel position to be processed, thereby obtaining the plurality of gain values. It is preferable to provide a feature amount merging circuit for calculating a plurality of total feature amounts corresponding to the global feature amount.

  The gain calculation circuit may further include a gain output circuit that calculates the plurality of gains by dividing a difference between two adjacent target luminances by a difference between two adjacent total feature amounts. Is preferred.

Further, when the data of the pixel to be processed is smaller than the largest total feature amount, the gain application circuit subtracts the total feature amount from the data of the pixel to be processed, and the result of the subtraction is Multiplying by the gain corresponding to the total feature amount, calculating the pixel data after the gain application by adding the multiplication result and the target luminance corresponding to the total feature amount,
When the data of the pixel to be processed is equal to or greater than the largest total feature amount, the largest target luminance corresponding to the largest total feature amount is output as the pixel data after the gain application. Is preferred.

The pixel data of the HDR image is data of an RGB color space,
The image processing apparatus further includes an image space conversion circuit configured to convert data of the RGB color space of the pixel of the HDR image into data of a color space including a Y component.
The gain application circuit applies the gain to data of a Y component of a color space including a Y component of a pixel to be processed of the HDR image, and a color space including the Y component of the pixel after the gain application. Calculate the data of
Further, it is preferable that an image space inverse conversion circuit for converting data in a color space including the Y component of the pixel after the gain application into data in the RGB color space is provided.

Further, based on the data of the pixel of the HDR image of the current frame, a position information extraction circuit for extracting the horizontal and vertical addresses of the pixel to be processed as the position information of the pixel,
It is preferable that the global feature value calculating circuit and the local feature value calculating circuit calculate the plurality of global feature values and the local feature value of the block image based on an address of the pixel to be processed.

A linear space conversion circuit configured to convert pixel data of the HDR image of the current frame into linear space pixel data using an electro-optical conversion function;
The gain application circuit performs processing on the linear space from among a plurality of gains for the block image of the HDR image one frame before corresponding to the pixel position on the processing target for the data of the pixel to be processed in the linear space. Preferably, a gain corresponding to the data of the target pixel is applied to calculate the data of the pixel after the gain is applied.

A data aggregation circuit that outputs aggregated data in which a plurality of data input as pixel data of one pixel of the HDR image of the current frame is aggregated into one data;
The global feature amount calculation circuit calculates the plurality of global feature amounts based on aggregated data of all pixels of the HDR image of the current frame,
It is preferable that the local feature calculating circuit calculates a local feature of the block image based on aggregated data of all pixels of the block image.

  Further, it is preferable that the data aggregation circuit outputs, as the aggregated data, an average value, a maximum value, a minimum value, or a luminance value of a plurality of pieces of data input as the pixel data of the one pixel.

  Further, it is preferable that a gamma correction circuit for gamma-correcting the data of the pixel after the gain application be provided in accordance with the luminance characteristic of the display.

Further, according to the present invention, the global feature amount calculating circuit calculates a plurality of global feature amounts, which are feature amounts of the HDR image, based on data of all pixels of the HDR image of the current frame, and holds for one frame time. And outputting a global feature amount of the HDR image one frame before;
For each of the plurality of block images obtained by dividing the HDR image of the current frame in the horizontal direction and the vertical direction, a local feature amount calculation circuit calculates a feature amount of the block image based on data of all pixels of the block image. A local feature value is calculated and held for one frame time, and the local feature value of the block image of the HDR image one frame before and the surrounding block image corresponding to the pixel position to be processed of the HDR image of the current frame. Calculating the local feature amount of the pixel of the block image of the HDR image one frame before corresponding to the pixel position of the processing target by interpolating
A gain calculating circuit configured to calculate a plurality of global feature amounts of the HDR image of the previous frame, a local feature amount of a pixel of the HDR image of the previous frame corresponding to the pixel position to be processed, and the plurality of global feature amounts; A plurality of gains corresponding to the plurality of global features and the plurality of target luminances, for each block image of the HDR image one frame before, based on a plurality of target luminances representing output luminances of a corresponding display; Calculating a plurality of total feature values corresponding to the plurality of global feature values and local feature values of pixels of the HDR image one frame before corresponding to the pixel position to be processed;
The gain application circuit converts the data of the pixel to be processed from the plurality of gains of the block image of the HDR image one frame before corresponding to the pixel position of the pixel to be processed to the data of the pixel to be processed. Calculating the data of the pixel after applying the gain by applying the corresponding gain and the total feature amount to provide a tone mapping method.

  According to the present invention, even in a display having a narrow dynamic range, an HDR image having a wide dynamic range can be expressed by tone mapping without deteriorating a detailed portion thereof and continuity of an output image.

  Further, by calculating the global feature amount, the local feature amount, and the gain using the aggregated data, the circuit scale, the power consumption, and the cost can be significantly reduced, and the calculation time for that can be greatly reduced.

  Furthermore, since the global feature amount of the HDR image of the previous frame is retained and applied to the HDR image of the current frame, a memory for retaining the pixel data of the HDR image of the current frame for one frame time is not required. In addition, since the local feature amounts of a plurality of block images in the HDR image are stored, the memory capacity required for storing the local feature amounts can be significantly reduced. Therefore, the circuit scale can be significantly reduced, and power consumption and cost can be significantly reduced.

FIG. 2 is a block diagram of an embodiment showing a configuration of a tone mapping circuit according to the present invention. FIG. 2 is a block diagram of an embodiment illustrating a configuration of a global feature amount calculation circuit illustrated in FIG. 1. 6 is an example of a cumulative histogram showing the cumulative number of pixels belonging to each global region of the value of the aggregated data. FIG. 2 is a block diagram illustrating an example of a configuration of a local feature calculating circuit illustrated in FIG. 1. FIG. 2 is a block diagram of an embodiment illustrating a configuration of a gain calculation circuit illustrated in FIG. 1. 5 is an example graph showing gain applied to pixels of an HDR image. FIG. 2 is a block diagram of an embodiment illustrating a configuration of a gain application circuit illustrated in FIG. 1. FIG. 10 is a block diagram of another embodiment showing a configuration of a tone mapping circuit according to the present invention. FIG. 11 is a block diagram of still another embodiment showing the configuration of the tone mapping circuit according to the present invention.

  Hereinafter, a tone mapping circuit and a tone mapping method of the present invention will be described in detail based on preferred embodiments shown in the accompanying drawings.

  FIG. 1 is a block diagram of one embodiment showing a configuration of a tone mapping circuit according to the present invention. The tone mapping circuit 10 shown in FIG. 1 performs tone mapping for displaying an HDR image having a wide dynamic range on a display having a narrow dynamic range. The tone mapping circuit 10 includes a position information extraction circuit 30, a data aggregation circuit 12, a global feature calculation circuit 14, a local feature calculation circuit 16, a gain calculation circuit 18, a linear space conversion circuit 20, and a gain application circuit. A circuit 22 and a gamma correction circuit 24 are provided.

In the tone mapping circuit 10, the data RGB image of the pixel of the HDR image is input from the outside to the position information extracting circuit 30 every pixel time.
Based on the pixel data RGB image of the HDR image, the position information extracting circuit 30 determines the horizontal (horizontal) and vertical (vertical) addresses (coordinates) x, x, of the pixel to be processed as the pixel position information. Extract y.

Subsequently, the data aggregating circuit 12 receives, from the outside, pixel data RGB image of the HDR image every pixel time.
The data aggregating circuit 12 outputs aggregated data obtained by aggregating a plurality of pieces of data input as data RGB image of one pixel of the HDR image of the current frame into one data.

  The plurality of data is, for example, data RGB image of three channels of RGB (red, green, blue). Further, when data of a plurality of pixels is input as pixel data of one pixel as in a television or the like, the plurality of data is the data of the plurality of pixels.

The data aggregation circuit 12 outputs, as aggregated data, an average value, a maximum value, a minimum value, a luminance value, and the like of a plurality of data input as, for example, pixel data of one pixel.
When a plurality of data are RGB three-channel data RGB images, the average value is obtained by averaging R data, G data, and B data. The maximum value and the minimum value are the maximum and minimum values of R data, G data, and B data. The luminance value is Y component data obtained by converting data RGB image of a pixel in the RGB color space into data of a pixel in a color space including a Y component such as a YUV (luminance Y and color difference UV) color space.
The same applies to a case where a plurality of data is data of a plurality of pixels.

  In the case of the present embodiment, three-channel RGB data RGB image is input as pixel data of one pixel. The data aggregating circuit 12 aggregates RGB data RGB images of one pixel and outputs the average value.

Subsequently, the address x, y of the pixel of the HDR image is input from the position information extraction circuit 30 to the global feature amount calculation circuit 14 every pixel time, and the aggregated data of the pixel of the HDR image is input from the data aggregation circuit 12. Is entered.
The global feature amount calculating circuit 14 divides the HDR image of the current frame into M pieces in the horizontal direction and N pieces in the vertical direction (M × N pieces, where M and N are integers of 1 or more), and area images M, N Of each of the area images M and N based on the aggregated data of all the pixels of the area images M and N specified by the addresses x and y of the pixels to be processed. An area global feature amount global_ore [M, N] [i] (an integer of 1 or more, which represents the number of thresholds described later) is sequentially calculated and held for one frame time.
In addition, the global feature amount calculation circuit 14 selects, from among the area global feature amounts global_ore [M, N] [i] of all the area images M and N of the HDR image of the previous frame held one frame time earlier. The global global feature amounts global_ore [M, N] [i] of the area images M and N of the HDR image one frame before corresponding to the pixel position to be processed are sequentially output as global feature amounts global_ore '[i].

  As shown in FIG. 2, the global feature amount calculation circuit 14 includes an area global feature amount calculation circuit 32, a buffer 34, an area image selection circuit 36, and a global feature amount output circuit 38.

In the global feature amount calculation circuit 14, the area global feature amount calculation circuit 32 specifies the area images M and N including the pixel to be processed based on the addresses x and y of the pixel to be processed. Is calculated based on the aggregated data of all the pixels of the area images M and N including the global image feature global_ore [M, N] [i] of the area images M and N including the pixel to be processed. It is held for one frame time.
The area image selection circuit 36 selects area images M and N including the pixel to be processed based on the addresses x and y of the pixel to be processed.
Subsequently, the global feature output circuit 38 performs processing from the area global feature [global_ore [M, N] [i] of all the area images M and N of the HDR image one frame before stored in the buffer 34. The global feature amount global_ore [M, N] [i] of the area images M and N of the HDR image one frame before corresponding to the target pixel position is output as the global feature amount global_ore '[i].

  In the case of the present embodiment, M = N = 1. That is, the HDR image is not divided and becomes one area image. In this case, the global feature amount calculating circuit 32 calculates the global feature amount global_ore [i] of the HDR image based on the aggregated data of all the pixels of the HDR image of the current frame, and holds the global feature value global_ore [i] in the buffer 34 for one frame time. You. The area image selection circuit 36 and the global feature amount output circuit 38 are unnecessary, and the global feature amount global_ore [i] of the HDR image one frame before is output from the buffer 34 as the global feature amount global_ore '[i].

  The global feature value calculation circuit 14 determines a plurality of (i) threshold values input from the outside as a plurality of global feature values global_ore '[i] with respect to the total number of pixels of the HDR image. A central value or the like of aggregated data included in a plurality (i) of aggregated data regions corresponding to the number of pixels having a ratio of (i) is calculated.

  For example, when specifying the number of pixels of 50% (median value) with respect to the total number of pixels of the HDR image, if the total number of pixels of the HDR image is constant, the threshold is set to 50% of the total number of pixels of the HDR image. Is input. If the total number of pixels of the HDR image is variable, for example, the total number of pixels of the HDR image is set to 100, and 50 is input as a threshold value. Then, the actual total number of pixels of the HDR image, which is set in advance in the global feature amount calculation circuit 14, and 50%, which is the ratio of 50 of the threshold value to 100, which is the set size of the total number of pixels of the HDR image, Is used as the actual threshold.

  The area of the aggregated data represents a range of values obtained by dividing a range of values from the minimum value to the maximum value of the aggregated data by a fixed value (constant width). For example, in the case of 10-bit aggregate data, the value range is 0 to 1023. For example, when the fixed value is 1, the number of areas of the consolidated data is 1024. When the fixed value is 16, the number of areas of the aggregated data is 1024/16 = 64.

  The global feature amount calculation circuit 14 creates a cumulative histogram representing the cumulative number of pixels belonging to each area of the value of the aggregated data from the aggregated data of all the pixels of the HDR image, and uses the cumulative histogram to generate an HDR image. Corresponding to the number of pixels of, for example, 10% (dark portion), 50% (center portion) and 90% (bright portion) specified by three threshold values th1, th2 and th3 with respect to the total number of pixels Representative values such as a median value and a mode value of the aggregated data included in the one aggregated data area are output as global feature amounts global_ore '[i].

  FIG. 3 is an example of a cumulative histogram representing the cumulative number of pixels belonging to each global region of the aggregated data. The horizontal axis of the cumulative histogram shown in FIG. 3 represents the value of the aggregated data output as the global feature amount global_ore '[i], and the vertical axis represents the cumulative number of pixels belonging to each area of the aggregated data. On the horizontal axis of the cumulative histogram, the value of the aggregated data increases one by one toward the right side with the left end being 0, and becomes the maximum value at the right end.

  The cumulative number of pixels belonging to each region of the aggregated data is the total number of pixels belonging to each region from the leftmost region in a general histogram representing the number of pixels included in each region. That is, when the cumulative number of the left region is subtracted from the cumulative number of the right region of the two adjacent aggregated data regions, the number of pixels belonging to the right region in a general histogram is obtained. Further, the cumulative number of pixels belonging to the rightmost aggregated data area is the total number of pixels of the HDR image.

  For example, when the number of pixels of 10% is specified by the threshold value th1 (i = 1), the global feature amount calculation circuit 14 determines, in the cumulative histogram, the number of pixels on the left end that is the number closest to the number of pixels of 10%. The central value of the aggregated data included in the area of the aggregated data is output as the global feature amount global_ore '[1]. Similarly, when the number of pixels of 50% is specified by the threshold value th2 (i = 2), the central value of the aggregated data included in the area of the second aggregated data from the left is the global feature amount global_ore '[2]. When the number of pixels of 90% is designated by the threshold th3 (i = 3), the central value of the aggregated data included in the seventh aggregated data area from the left is the global feature amount global_ore '[3 ].

  In the case of 10-bit aggregated data, the number of areas of the aggregated data may be 1024, which is the maximum number, but is preferably a smaller number, for example, 12 to 80. The number of global feature values global_ore '[i] per area image is not limited to three, but may be two or more, and for example, is preferably about ten. Also, instead of calculating the median value of the aggregated data included in the area of the aggregated data corresponding to the median pixel number of the total number of pixels of the HDR image as one of the plurality of global feature amounts global_ore '[i], The median value of the aggregated data included in the aggregated data area corresponding to the average value of the data of all the pixels of the HDR image may be calculated. In addition to the cumulative histogram, a general histogram representing the number of pixels belonging to each region of the aggregated data of the pixels of the HDR image is created from the aggregated data of all the pixels of the HDR image. May be calculated as the global feature amount global_ore '[i].

Returning to FIG. 1, the address x, y of the pixel of the HDR image is input from the position information extraction circuit 30 to the local feature amount calculation circuit 16 every pixel time, and the pixel aggregation of the HDR image is input from the data aggregation circuit 12. Aggregated data is input.
The local feature calculating circuit 16 divides the HDR image of the current frame into m pieces in the horizontal direction and n pieces in the vertical direction (m × n pieces, where m and n are integers equal to or greater than 2). Of each of the block images m and n, based on the aggregate data of all the pixels of the block images m and n and the addresses x and y of the pixels to be processed, a local feature amount local_ore [m, n] which is a feature amount of the block images m and n Calculate and hold for one frame time.
In addition, the local feature calculating circuit 16 calculates the local feature between the block images m and n of the HDR image one frame before and held one frame before the frame position corresponding to the pixel position to be processed, and the local feature of the block images around it. To calculate the local feature value local_ore 'of the pixel of the block image m, n of the HDR image one frame before corresponding to the pixel position to be processed.

  As shown in FIG. 4, the local feature calculating circuit 16 includes a block local feature calculating circuit 40, a buffer 42, a block image selecting circuit 44, and a pixel local feature outputting circuit 46.

In the local feature calculating circuit 16, the block local feature calculating circuit 40 specifies block images m and n including the pixel to be processed based on the addresses x and y of the pixel to be processed, and the pixel to be processed. Is calculated based on the aggregate data of all the pixels of the block images m and n, and the local feature value local_ore [m, n] of the block images m and n is stored in the buffer 42 for one frame. Hold for time.
Further, the block image selection circuit 44 selects the block images m and n including the pixel to be processed based on the addresses x and y of the pixel to be processed.
Subsequently, the pixel local feature output circuit 46 outputs a pixel to be processed from the local feature local_ore [m, n] of all the block images m and n of the HDR image one frame before stored in the buffer 42. The block image m, n of the HDR image one frame before corresponding to the position and the local feature value local_ore [m, n] of the surrounding block image are read out, and the HDR image one frame before corresponding to the pixel to be processed Is interpolated between the block image m, n and the local feature value local_ore [m, n] of the surrounding block image, and the pixels of the block image m, n of the HDR image one frame before corresponding to the pixel to be processed Is output.

  The local feature value calculation circuit 16 calculates the average value, the median value, the maximum value, the minimum value, or the average value of the aggregated data of all the pixels of the block images m and n as the local feature value local_ore [m, n] of the block images m and n. A representative pixel value and the like are calculated. The representative pixel value is aggregate data of pixels at predetermined positions (addresses) in the block image, such as the upper left pixel and the center pixel among all the pixels of the block image.

  In the case of the present embodiment, the HDR image is divided into 100 (m = 100) × 100 (n = 100) = 10000 block images in the horizontal direction. The local feature calculating circuit 16 calculates an average value of aggregated data of all pixels of the block image for each block image as a local feature of the block image local_ore [m, n]. In addition, the local feature calculating circuit 16 calculates the average value of the aggregate value of the aggregate data of the block image of the HDR image one frame before and the surrounding nine block images corresponding to the pixel position to be processed. Interpolation is performed to calculate pixel data of the block image of the HDR image one frame before corresponding to the pixel position to be processed as the local feature amount Local_ore 'of the pixel.

  The number of divisions in the horizontal and vertical directions of the HDR image may be two or more, and the number of block images may be four or more. Further, the number of local feature values local_ore [m, n] of the block image used in the interpolation processing may be two or more. For example, using the local feature value local_ore [m, n] of a block image of the HDR image one frame before corresponding to the pixel position to be processed and five block images including four block images surrounding the block image in the cross direction. The interpolation process may be performed, or the local feature amount local_ore [m, n of the block image of the HDR image one frame before corresponding to the pixel to be processed and the 25 block images around the block image of the block image 24 ] May be used to perform the interpolation processing.

Subsequently, the gain calculation circuit 18 outputs a plurality of global feature amounts global_ore of the area images M and N of the HDR image one frame before corresponding to the pixel position to be processed from the global feature amount calculation circuit 14 every frame time. '[i] is entered. Further, the local feature value local_ore 'of the pixel of the HDR image one frame before corresponding to the pixel position of the HDR image of the current frame is input from the local feature value calculation circuit 16 to the gain calculation circuit 18 every pixel time. You. Further, a plurality of target luminances ore_y [i] and ore_y [0] representing the output luminance of the display corresponding to the plurality of global feature amounts global_ore ′ [i] are input from the outside to the gain calculation circuit 18.
The gain calculation circuit 18 calculates a plurality of global feature values global_ore '[i] of the area images M and N of the HDR image of the previous frame corresponding to the pixel position to be processed of the HDR image of the current frame, Based on the local feature local_ore 'of the pixel of the HDR image one frame before corresponding to the pixel position to be processed, and a plurality of target luminances ore_y [0] and ore_y [i], each of the HDR images one frame before For the block image of, a plurality of gains [i] corresponding to a plurality of global feature amounts global_ore '[i] and a plurality of target luminances ore_y [0] and ore_y [i] are calculated. When the local feature value local_ore 'of a pixel is calculated by interpolation as in the present embodiment, a plurality of gains gain [i] may be calculated for each pixel.

  As shown in FIG. 5, the gain calculating circuit 18 includes a feature amount merging circuit 48 and a gain output circuit 50.

In the gain calculating circuit 18, the feature amount merging circuit 48 sets a plurality of global feature amounts global_ore ′ [i] of the area images M and N of the HDR image of the previous frame and a pixel position to be processed in the HDR image of the current frame. The corresponding local feature value local_ore 'of the pixel of the HDR image one frame before is merged by a formula described later, and the total feature values ore [0] and ore [i] of the HDR image one frame before are generated.
Then, based on the total feature amounts ore [0] and ore [i] of the HDR image one frame before, the gain output circuit 50 sets a plurality of (i) global feature amounts global_ore ′ [i] and a plurality ( A plurality (i) of gains [i] corresponding to (i + 1) target luminances ore_y [0] and ore_y [i] are calculated. The total feature amount ore [0] represents the lower limit of the total feature amount ore [i], that is, 0, and the target luminance ore_y [0] represents the lower limit of the output luminance of the display. .

  For example, when a threshold th1 that specifies 10% of the number of pixels in the HDR image is input as a threshold for calculating the global feature amount global_ore '[i], the target luminance ore_y [1] is set to a luminance of 10% of the maximum luminance of the display. Similarly, when threshold values th2 and th3 of 50% and 90% are input, the target luminance ore_y [2] and ore_y [3] are set to 50% and 90% of the maximum luminance of the display. .

  FIG. 6 is an example graph illustrating gain applied to pixels of an HDR image. The horizontal axis of the graph shown in FIG. 6 represents the total feature amounts ore [0] and ore [i], and the vertical axis represents the target luminance ore_y [0] and ore_y [i] of the display. The total feature values ore [0] and ore [i] are increased toward the right side with ore [0] = 0 at the left end on the horizontal axis of the graph, and the maximum value is at the right end. Further, the target luminance ore_y [0] and ore_y [i] are set such that the target luminance at the lower end is 0 on the vertical axis of the graph, and the target luminance increases toward the upper side, and the upper end has the maximum value.

  On the horizontal axis of the graph, the total feature amounts ore [0] and the total feature amounts ore [1] and ore [2] corresponding to 10%, 50%, and 90% of the total number of pixels of the HDR image are shown. ] And ore [3] are shown. On the vertical axis of the graph, target luminance ore_y [0], total feature amounts ore [1], ore [2] and ore [3], that is, 10% of the total number of pixels of the HDR image, 50% Target luminances ore_y [1], ore_y [2], and ore_y [3] representing the output luminance of the display corresponding to the number of pixels of% and 90% are shown.

  The gain calculation circuit 18 calculates the gain gain [i] by, for example, the following equation.

gain [i] = (ore_y [i]-ore_y [i-1]) / (ore [i]-ore [i-1])
ore [i] = c * (a * local_ore '+ (1-a) * global_ore' [i])

  Here, i represents the number of thresholds, and in the present embodiment, i = 1, 2, and 3. a is a weighting coefficient input from outside, and its value can be adjusted from the outside in the range of 0 ≦ a ≦ 1. With the weighting coefficient a, the local feature value local_ore ′ and global feature value global_ore ′ [i] of the pixel of the HDR image one frame before corresponding to the pixel position to be processed can be weighted and adjusted. If a = 1, it means that the global feature amount global_ore '[i] is not added to the gain gain [i]. input represents pixel data of an input HDR image. c is a coefficient, and the maximum value of the total feature amount ore [i] and the maximum value of the pixel data input are matched by the coefficient c.

  That is, the feature amount merging circuit 48 weights and adds the local feature amount local_ore 'of the pixel of the HDR image one frame before corresponding to the pixel position to be processed and each of the plurality of global feature amounts global_ore' [i]. By doing so, a plurality of total feature amounts ore [0] and ore [i] corresponding to a plurality of global feature amounts global_ore '[i] are calculated. The gain output circuit 50 calculates the difference between two adjacent target luminances (ore_y [i] -ore_y [i-1]) and the difference between two adjacent total feature amounts (ore [i] -oree [i]. -1]) to calculate a plurality of gains gain [i].

  According to the above equation, the total feature amounts ore [0], ore [1], ore [2], ore [3] and the target luminance ore_y [0], ore_y [1], ore_y [2], Three gains gain [1], gain [2], and gain [3] corresponding to ore_y [3] are calculated as the slopes of the graph shown in FIG. As shown in the graph of FIG. 6, the target luminance of the largest total feature amount ore [3], that is, the target luminance of the region having the number of pixels higher than 90% of the total number of pixels of the HDR image, is Target luminance ore_y [3], which is the maximum value, is clipped to limit.

Subsequently, the data RGB image of the pixels of the HDR image is input from the outside to the linear space conversion circuit 20 every pixel time.
The linear space conversion circuit 20 converts the pixel data RGB image of the HDR image of the current frame into the data input of the pixel in the linear space using an electro-optical conversion function (EOTF).

  When the format of the HDR image is HDR10, the electro-optical conversion function is an inverse conversion function of the PQ curve, and is defined by the following equation.

Here, input is input pixel data RGB image, and output is output pixel data. max (s1, s2) is a function that outputs the larger value of the two values s1, s2. The values of m1, m2, c1, c2, and c3 are as follows.
m1 = 0.1593017578
m2 = 78.84375
c1 = 0.8359375 = c3-c2 + 1
c2 = 18.8515625
c3 = 18.6875

  When the format of the HDR image is HLG, the electro-optical conversion function is an inverse conversion function of the HLG curve, and is defined by the following equation.

Here, input is input pixel data, and output is output pixel data. The knee is pixel data at a knee point at which the curve shape of the conversion function switches, and is a value determined by the metadata of the image content. vmax is the maximum value of output. min (s1, s2) is a function that outputs the smaller value of the two values s1, s2. The values of a, b, and c are as follows.
a = 0.17883277
b = 0.28466892
c = 0.55991073

  In the case of the present embodiment, the RGB data RGB image of the pixel of the HDR image is input to the linear space conversion circuit 20. The linear space conversion circuit 20 can be realized by using three look-up tables (LUTs) having the same contents corresponding to RGB. In general, the look-up table stores values such that the output bit width is larger than the input bit width or the input fixed width and output floating point format.

Subsequently, the data input of the pixel in the linear space is input from the linear space conversion circuit 20 to the gain application circuit 22 every pixel time, and the total characteristic of each pixel of the HDR image is input from the gain calculation circuit 18. The quantities ore [0], ore [i] and a plurality of gains gain [i] of each block image of the HDR image are input.
The gain application circuit 22 calculates, for the data input of the pixel to be processed in the linear space, from a plurality of gains gain [i] for the block image of the HDR image one frame before corresponding to the pixel position to be processed. The gain gain [i] corresponding to the data input of the pixel to be processed in the linear space and the total feature values ore [0] and ore [i] are applied to calculate the pixel output data after the gain application. That is, tone mapping is performed.

  For example, as shown in FIG. 7, the gain application circuit 22 applies a gain gain [i] to pixel data input in a linear space by the following equation.

That is, when the data input of the pixel to be processed is smaller than the largest total feature amount ore [3], the gain application circuit 22 subtracts the total feature amount ore [i] from the data input of the pixel to be processed. Then, the result of the subtraction is multiplied by the gain gain [i] corresponding to the total feature amount ore [i], and the multiplication result is multiplied by the target luminance ore_y [i] corresponding to the total feature amount ore [i]. The pixel data output after the gain application is calculated by the addition.
On the other hand, if the data input of the pixel to be processed is greater than or equal to the largest total feature amount ore [3], the data output of the pixel after gain application corresponds to the largest total feature amount ore [3]. The maximum target luminance ore_y [3] = limit is output.

Then, the pixel data output of the HDR image of the current frame after gain application is input from the gain application circuit 22 to the gamma correction circuit 24 every pixel time.
The gamma correction circuit 24 performs gamma correction on the pixel output data after the gain application in accordance with the luminance characteristics of the display on which the HDR image after the gain application is displayed.

  The gamma correction circuit 24 performs gamma correction by non-linearly converting the data output of the gain-applied pixel using, for example, an inverse conversion curve of 2.2 squared gamma curve which is a standard gamma value γ of the display. Do.

  Next, the operation of the tone mapping circuit 10 will be described.

  In the tone mapping circuit 10, the position information extraction circuit 30 extracts the address x, y of the pixel of the HDR image as the position information based on the pixel data RGB image of the HDR image of the current frame.

  In addition, the data aggregating circuit 12 aggregates data of three channels of RGB input as data RGB image of one pixel of the HDR image of the current frame, and outputs an average value as the aggregated data.

Subsequently, the global feature value calculation circuit 14 specifies three threshold values th1, th2, and th3 for the total number of pixels of the HDR image based on the aggregated data of all the pixels of the HDR image of the current frame. For example, the median value of the aggregated data included in the three aggregated data regions corresponding to the pixel numbers of 10%, 50%, and 90% is global feature amount global_ore [1], global_ore [2], global_ore [3] And is held for one frame time.
Also, the global feature amounts global_ore [1], global_ore [2], global_ore [3] of the HDR image one frame before are output as global feature amounts global_ore '[1], global_ore' [2], global_ore '[3]. You.

Further, the local feature calculating circuit 16 calculates the block images m and n based on the aggregate data of all the pixels of the block images m and n and the addresses x and y of the pixels to be processed for each of the block images m and n. As the local feature value local_ore [m, n], for example, an average value of aggregate data of all pixels of the block images m and n is calculated. The local feature value local_ore [m, n] of the block images m and n is held for one frame time.
Further, the block image m, n of the HDR image one frame before corresponding to the pixel position to be processed and the local feature value local_ore [m, n] of the surrounding block image are interpolated to obtain the pixel position to be processed. Is calculated, the local feature value local_ore 'of the pixel of the block image m, n of the HDR image one frame before is calculated. The local feature value local_ore 'of a pixel has a feature that changes spatially more slowly than a change in data of the pixel.

  Subsequently, the gain calculation circuit 18 calculates three global feature values global_ore '[1], global_ore' [2], global_ore '[3] of the HDR image one frame before and one frame before the pixel position to be processed. Block image of the HDR image one frame before based on the local feature amount local_ore 'of the pixel of the HDR image and the target luminance ore_y [0], ore_y [1], ore_y [2] and ore_y [3]. , Three gains gain [1], gain [2], and gain [3] are calculated.

  On the other hand, the linear space conversion circuit 20 converts the pixel data RGB image of the HDR image of the current frame into the pixel data input of the linear space using the lightning conversion function.

  Subsequently, the gain application circuit 22 applies a plurality of gains gain [i] for the block image of the HDR image one frame before corresponding to the pixel position to be processed, to the data input of the pixel to be processed in the linear space. From among them, the gain gain [i] corresponding to the data input of the pixel to be processed in the linear space is applied, and the data output of the pixel after the gain application is calculated. That is, different gains gain [i] are applied to the dark, central, and light portions. For example, a larger gain is applied to a dark portion than to the central portion, and a smaller gain is applied to a bright portion than to the central portion. Thus, an unnatural appearance due to clipping of the dynamic range can be prevented.

  Then, the gamma correction circuit 24 performs gamma correction on the pixel data output after the gain application in accordance with the luminance characteristics of the display, and displays an image corresponding to the gamma corrected pixel data on the display.

  As described above, the calculation of the aggregated data, the global feature amount global_ore '[i], the local feature amount local_ore', and the gain gain [i] are pipelined, and in parallel with this, the conversion to linear space and the gain gain Application of [i] is also pipelined.

  The tone mapping circuit 10 has a wide dynamic range including a moving image in which a bright scene and a dark scene exist, and a still image in which both a bright portion and a dark portion exist in one frame even in a display having a narrow dynamic range. Can be expressed without loss of continuity of the detailed part and the output image by tone mapping. In particular, high-luminance objects in the sky such as clouds and celestial bodies, shapes of illumination and light reflection, and dark portions in night scenes and objects therein can be expressed in detail.

  The circuit scale and consumption are calculated by calculating the global feature amount global_ore '[i], the local feature amount local_ore', and the gain gain [i] using the aggregated data, not the data RGB image of all the pixels of the HDR image. Power and cost can be significantly reduced, and the computation time for it can be significantly reduced.

  Further, a global feature amount global_ore '[i] is calculated using the pixel data of the HDR image of the current frame, and when this is applied to the HDR image of the current frame, the global feature amount global_ore' [i] is calculated. Requires one frame time, so a memory for holding the pixel data of the HDR image of the current frame for one frame time is required. On the other hand, in the tone mapping circuit 10, in order to apply the global feature amount global_ore '[i] one frame before to the HDR image of the current frame, the pixel data of the HDR image of the current frame is held for one frame time. No additional memory is required. Therefore, the circuit scale can be significantly reduced, and power consumption and cost can be significantly reduced.

  In addition, since the local feature amounts local_ore [m, n] of a plurality of block images in the HDR image are stored instead of the local feature amounts of all the pixels of the HDR image, the memory capacity required for storing the local feature amounts is also large. It can be significantly reduced. Therefore, the circuit scale can be significantly reduced, and power consumption and cost can be significantly reduced.

  The gain application circuit 22 can apply the gain gain [i] to data in another color space such as a color space including a Y component, without being limited to data in the RGB color space of the pixels of the HDR image.

  FIG. 8 is a block diagram of another embodiment showing the configuration of the tone mapping circuit according to the present invention. The tone mapping circuit 11 shown in FIG. 8 further includes an image space conversion circuit 26 and an image space inverse conversion circuit 28 in the tone mapping circuit 10 shown in FIG.

  That is, the tone mapping circuit 11 includes a position information extraction circuit 30, a data aggregation circuit 12, a global feature calculation circuit 14, a local feature calculation circuit 16, a gain calculation circuit 18, a linear space conversion circuit 20, A gain application circuit 22, a gamma correction circuit 24, an image space conversion circuit 26, and an image space inverse conversion circuit 28 are provided. To the tone mapping circuit 11, data RGB image in the RGB color space is input as pixel data of the HDR image.

  Hereinafter, the image space conversion circuit 26 and the image space inverse conversion circuit 28 will be described, and the description of the other components will not be repeated.

In the tone mapping circuit 11, data input of the RGB color space of the pixels in the linear space is input from the linear space conversion circuit 20 to the image space conversion circuit 26 every pixel time.
The image space conversion circuit 26 converts the data input in the RGB color space of the pixels in the linear space into the data input ′ in the color space including the Y component. The image space conversion circuit 26 of the present embodiment converts data in the RGB color space into data in the YUV color space.

Subsequently, the pixel space output of the HDR image after the gain application is input from the gain application circuit 22 to the image space inverse conversion circuit 28 every pixel time.
The image space inverse conversion circuit 28 converts the data output of the color space including the Y component of the pixel after the gain application into the data output ′ of the RGB color space. The image space inverse conversion circuit 28 of the present embodiment converts data in the YUV color space into data in the RGB color space.

In the tone mapping circuit 11, the image space conversion circuit 26 converts the data input in the RGB color space of the pixel in the linear space into the data input ′ in the YUV color space.
Subsequently, the gain application circuit 22 applies the gain gain [i] to the Y component data input ′ of the pixel to be processed in the linear space in the YUV color space, and calculates the data output in the YUV color space. .
Subsequently, the image space inverse conversion circuit 28 converts the data output in the YUV color space of the pixel after the gain application to the data output ′ in the RGB color space. That is, the color space is inversely transformed.
Subsequently, the gamma correction circuit 24 performs gamma correction on the data output ′ of the RGB color space of the pixel to which the gain has been applied converted by the image space inverse conversion circuit 28.

  Note that, in the tone mapping circuit 10, the position information extraction circuit 30, the data aggregation circuit 12, the linear space conversion circuit 20, and the gamma correction circuit 24 are not essential components.

  For example, when the position information extracting circuit 30 is not provided, the global feature amount calculating circuit 14 does not refer to the addresses x and y of the pixels to be processed, and based on the aggregated data of all the pixels of the HDR image of the current frame. Thus, a plurality of global feature values can be calculated. Similarly, the local feature calculating circuit 16 can calculate the local feature of the block image based on the aggregated data of all the pixels of the HDR image of the current frame. Even in this case, the correction value can be calculated by merging two different feature amounts.

  When the data aggregating circuit 12 is not provided, the global feature amount calculating circuit 14 can calculate a plurality of global feature amounts based on the data of all the pixels of the HDR image of the current frame instead of the aggregated data. Similarly, the local feature calculating circuit 16 can calculate the local feature of the block image based on the data of all the pixels of the block image.

  When the data aggregation circuit 12 is not provided, as shown in FIG. 9, linear space pixel data input from the linear space conversion circuit 20 is input to the global feature amount calculation circuit 14 and the local feature amount calculation circuit 16. Is also good. In this case, the global feature value calculation circuit 14 and the local feature value calculation circuit 16 can calculate a plurality of global feature values and local feature values of the block image based on the pixel data input in the linear space.

  When the linear space conversion circuit 20 is not provided, the gain application circuit 22 calculates the position of the pixel to be processed with respect to the data RGB image of the pixel to be processed in the HDR image of the current frame, instead of the data input of the pixel in the linear space. Of the block image of the HDR image one frame before corresponding to the data of the pixel to be processed, from among the plurality of gains gain [i] corresponding to the pixel data to be processed, output can be calculated.

  As described above, the present invention has been described in detail. However, the present invention is not limited to the above-described embodiment, and various improvements and modifications may be made without departing from the gist of the present invention.

10, 11 tone mapping circuit 12 data aggregation circuit 14 global feature calculation circuit 16 local feature calculation circuit 18 gain calculation circuit 20 linear space conversion circuit 22 gain application circuit 24 gamma correction circuit 26 image space conversion circuit 28 image space inverse conversion circuit Reference Signs List 30 position information extraction circuit 32 area global feature calculation circuit 34, 42 buffer 36 area image selection circuit 38 global feature output circuit 40 block local feature calculation circuit 44 block image selection circuit 46 pixel local feature output circuit 48 feature merge Circuit 50 Gain output circuit

Claims (17)

Based on the data of all the pixels of the HDR image of the current frame, a plurality of global feature amounts, which are the feature amounts of the HDR image, are calculated and held for one frame time, and the global feature amount of the HDR image of the previous frame is calculated. A global feature amount calculation circuit that outputs
For each of a plurality of block images obtained by dividing the HDR image of the current frame in the horizontal direction and the vertical direction, a local feature amount which is a feature amount of the block image is calculated based on data of all pixels of the block image. By interpolating between the local feature values of the block image of the HDR image one frame before and the surrounding block image corresponding to the pixel position to be processed of the HDR image of the current frame, and A local feature calculating circuit that calculates a local feature of a pixel of the block image of the HDR image one frame before corresponding to the pixel position to be processed;
A plurality of global feature values of the HDR image of the previous frame, local feature values of pixels of the HDR image of the previous frame corresponding to the pixel position to be processed, and an output of a display corresponding to the plurality of global feature values A plurality of gains corresponding to the plurality of global feature amounts and the plurality of target brightnesses for each block image of the HDR image one frame before, based on a plurality of target brightnesses representing brightness; and the plurality of global features. A gain calculation circuit for calculating a plurality of total feature amounts corresponding to the amount and the local feature amount of the pixel of the HDR image one frame before corresponding to the pixel position to be processed;
For the data of the pixel to be processed, a gain corresponding to the data of the pixel to be processed is selected from among a plurality of gains for the block image of the HDR image one frame before corresponding to the pixel position of the pixel to be processed. And a gain application circuit that calculates pixel data after gain application by applying the feature amount of (1).   The global feature amount calculation circuit creates a histogram representing the number of pixels belonging to each area of the pixel data of the HDR image from the data of all pixels of the HDR image, and uses the histogram to generate the plurality of global regions. The tone mapping circuit according to claim 1, wherein the tone mapping circuit calculates a feature amount.   The global feature amount calculation circuit creates a cumulative histogram representing the cumulative number of pixels belonging to each region of the pixel data of the HDR image from the data of all the pixels of the HDR image, and uses the cumulative histogram to generate the cumulative histogram. 2. The tone mapping circuit according to claim 1, wherein a plurality of global feature values are calculated.   The global feature amount calculation circuit uses the cumulative histogram to determine, as the plurality of global feature amounts, a plurality of ratios of the number of pixels specified by a plurality of thresholds with respect to the total number of pixels of the HDR image. 4. The tone mapping circuit according to claim 3, wherein a representative value of data included in a plurality of corresponding data areas is output.   The global feature amount calculation circuit outputs a representative value of data included in a data area corresponding to a median pixel number of all pixels of the HDR image as one of the plurality of global feature amounts. Item 5. The tone mapping circuit according to Item 4.   The global feature amount calculating circuit, for each of a plurality of area images obtained by dividing the HDR image of the current frame in a horizontal direction and a vertical direction, a feature amount of the area image based on data of all pixels of the area image. Are sequentially calculated and held for one frame time, and correspond to the pixel position of the processing target from among the area global feature amounts of all the area images of the HDR image of the previous frame. The tone mapping circuit according to any one of claims 1 to 5, wherein an area global feature amount of an area image of the HDR image one frame before is sequentially output as the global feature amount.   The local feature value calculation circuit calculates an average value, a median value, a maximum value, a minimum value, or a representative pixel value of data of all pixels of the block image as the local feature value of the block image. 7. The tone mapping circuit according to claim 6.   The gain calculation circuit weights and adds a local feature amount of a pixel of the HDR image one frame before corresponding to the pixel position to be processed and each of the plurality of global feature amounts, thereby obtaining the plurality of global features. The tone mapping circuit according to any one of claims 1 to 7, further comprising a feature amount merging circuit that calculates a plurality of total feature amounts corresponding to the feature amounts.   The gain calculation circuit further includes a gain output circuit that calculates the plurality of gains by dividing a difference between two adjacent target luminances by a difference between two adjacent total feature amounts. 9. The tone mapping circuit according to 8. When the data of the pixel to be processed is smaller than the largest total feature amount, the gain application circuit subtracts the total feature amount from the data of the pixel to be processed, and the result of the subtraction and the total Multiplying by the gain corresponding to the feature amount, calculating the data of the pixel after the gain application by adding the multiplication result and the target luminance corresponding to the total feature amount,
When the data of the pixel to be processed is equal to or greater than the largest total feature amount, the largest target luminance corresponding to the largest total feature amount is output as the pixel data after the gain application. The tone mapping circuit according to claim 8. The pixel data of the HDR image is data of an RGB color space,
The image processing apparatus further includes an image space conversion circuit configured to convert data of the RGB color space of the pixel of the HDR image into data of a color space including a Y component.
The gain application circuit applies the gain to data of a Y component of a color space including a Y component of a pixel to be processed of the HDR image, and a color space including the Y component of the pixel after the gain application. Calculate the data of
The tone mapping according to any one of claims 1 to 10, further comprising an image space inverse conversion circuit configured to convert data in a color space including a Y component of the pixel after the gain application into data in the RGB color space. circuit. Further, based on the data of the pixel of the HDR image of the current frame, a position information extraction circuit for extracting the horizontal and vertical addresses of the pixel to be processed as the position information of the pixel,
The method according to any one of claims 1 to 11, wherein the global feature value calculating circuit and the local feature value calculating circuit calculate the plurality of global feature values and the local feature value of the block image based on an address of the pixel to be processed. The tone mapping circuit according to claim 1. A linear space conversion circuit configured to convert pixel data of the HDR image of the current frame into linear space pixel data using an electro-optical conversion function;
The gain application circuit performs processing on the linear space from among a plurality of gains for the block image of the HDR image one frame before corresponding to the pixel position on the processing target for the data of the pixel to be processed in the linear space. The tone mapping circuit according to any one of claims 1 to 12, wherein a gain corresponding to data of a target pixel is applied to calculate data of the pixel after the gain is applied. A data aggregation circuit that outputs aggregated data in which a plurality of data input as pixel data of one pixel of the HDR image of the current frame is aggregated into one data;
The global feature amount calculation circuit calculates the plurality of global feature amounts based on aggregated data of all pixels of the HDR image of the current frame,
The tone mapping circuit according to claim 1, wherein the local feature calculation circuit calculates a local feature of the block image based on aggregated data of all pixels of the block image.   15. The tone mapping according to claim 14, wherein the data aggregation circuit outputs, as the aggregated data, an average value, a maximum value, a minimum value, or a luminance value of a plurality of pieces of data input as the pixel data of the one pixel. circuit.   The tone mapping circuit according to claim 1, further comprising a gamma correction circuit configured to perform gamma correction on data of the pixel after the gain is applied in accordance with a luminance characteristic of a display. A global feature amount calculating circuit calculates a plurality of global feature amounts, which are feature amounts of the HDR image, based on data of all pixels of the HDR image of the current frame, holds the global feature amount for one frame time, and one frame before. Outputting global features of the HDR image of
For each of the plurality of block images obtained by dividing the HDR image of the current frame in the horizontal direction and the vertical direction, a local feature amount calculation circuit calculates a feature amount of the block image based on data of all pixels of the block image. A local feature value is calculated and held for one frame time, and the local feature value of the block image of the HDR image one frame before and the surrounding block image corresponding to the pixel position to be processed of the HDR image of the current frame. Calculating the local feature amount of the pixel of the block image of the HDR image one frame before corresponding to the pixel position of the processing target by interpolating
A gain calculating circuit configured to calculate a plurality of global feature amounts of the HDR image of the previous frame, a local feature amount of a pixel of the HDR image of the previous frame corresponding to the pixel position to be processed, and the plurality of global feature amounts; A plurality of gains corresponding to the plurality of global features and the plurality of target luminances, for each block image of the HDR image one frame before, based on a plurality of target luminances representing output luminances of a corresponding display; Calculating a plurality of total feature values corresponding to the plurality of global feature values and local feature values of pixels of the HDR image one frame before corresponding to the pixel position to be processed;
The gain application circuit converts the data of the pixel to be processed from the plurality of gains of the block image of the HDR image one frame before corresponding to the pixel position of the pixel to be processed to the data of the pixel to be processed. Calculating the data of the pixel after applying the gain by applying the corresponding gain and the total feature amount.