JP5486791B2 - Image processing device - Google Patents

Description

The present invention relates to an image processing apparatus using a histogram generation circuit that generates a histogram indicating the frequency distribution of all luminance values based on image data indicating the luminance values of a plurality of pixels forming a screen of a predetermined area. is there.

  One of the disadvantages of LCD as an image display device is that the contrast ratio is lower than that of other image display devices such as CRT and PDP. This is due to the fact that the liquid crystal cannot completely block the light of the backlight. For example, when displaying an image with a black background, the liquid crystal cannot completely shield the light, so that the black is blurred (white floating), resulting in a sharp image. Therefore, in order to improve the contrast of the LCD, a technique for suppressing the brightness of the backlight in a dark scene has been proposed. In addition, when the backlight is simply darkened in a dark scene, the bright portion is also darkened accordingly, so that the value (luminance value) of the image data is also changed.

  As described above, in order to improve contrast, correction (backlight brightness correction and image data value correction) is performed when displaying on the image display device. The optimum conditions for the correction are as follows. Depending on the distribution pattern of the luminance values of a plurality of pixels forming the screen. Therefore, in order to perform the optimum correction, it is necessary to analyze the image luminance value distribution pattern, and a histogram is used for the analysis method.

  As an example, Patent Document 1 describes that a histogram is created from image data and the value of the image data is changed based on the histogram or the brightness of the backlight is reduced. In this way, in a dark image, it is possible to reduce power consumption while enhancing gradation expression.

  However, in a scene where the histogram is biased and the luminance of the entire screen gradually shifts, the shape of the histogram may change abruptly. This is because the horizontal axis of the histogram is divided into specific luminance value ranges and divided into a plurality of classes (less than the number of luminance value resolutions), and the frequency of the luminance value of the input image data is cumulatively added to the corresponding class. Yes, this can happen when the luminance values contained within a class change uniformly across class boundaries.

  For this reason, when control such as image correction is performed based on such a histogram, a scene that should change continuously becomes discontinuous due to a sudden change in the histogram. For example, when the entire scene gradually becomes brighter in a dark scene, the screen suddenly becomes brighter due to a sharp change of the histogram at a certain brightness.

  “Conventional example 1” in FIG. 3 is obtained when the entire luminance of the original image (upper stage) is uniformly reduced by 6 (middle stage), and when it is further reduced by 6 (total 12) (lower stage). Changes in the histogram are shown. Here, the image data is 8 bits, the width of each class is 16, and the number of classes is 16. Since the overall brightness of the original image has only been reduced to -6, -12, the histogram created should ideally shift to the dark side (left side) with the same shape. However, in the “conventional example 1” in FIG. 3, in the image in which the overall brightness is reduced by 6 with respect to the histogram of the original image, the histogram does not change so much, but the overall brightness is 12%. If it is decreased only by a large amount, the histogram changes drastically.

  Specifically, in both the histogram of the original image and the histogram of the image in which the brightness of the entire image is reduced by 6, a high frequency is included in the class 10 which is the highest class in which the frequency exists. . However, in the histogram of the image in which the overall luminance is reduced by 12, the frequency of class 10 is 0. Therefore, when the correction condition is set with the highest class having a frequency equal to or greater than a certain threshold as one of the objects to be considered, the correction condition greatly changes when the overall luminance is reduced by 12.

  One solution to this problem is to perform luminance correction using information on past frames in order to increase the continuity of the histogram. In this method, a complicated circuit for comparison and calculation is used. Becomes necessary, and the circuit scale becomes large. Another solution is to increase the number of classes. This is to create a gentle and continuous histogram by narrowing the class width to reflect the information of the lower bits. However, this also increases the number of classes, which increases the amount of calculation and increases the circuit scale.

Patent Document 2 proposes a technique for weighting the luminance value of image data in the vicinity of a threshold value for setting the width of each class of the histogram in order to remove the instability factor of the histogram purification due to noise. . As shown in FIG. 4, when thresholds for determining the width of one class are A and B, this is an increasing function in the range from “A−W” to “A + W” in the vicinity of the threshold A, and in the vicinity of the threshold B. In the range of “B−W” to “B + W”, a cumulative addition is performed using a weight function which is a decreasing function.
JP 2007-36728 A JP 2000-513904 gazette

  However, in the method shown in FIG. 4, the weighting range is only in the vicinity of the threshold values A and B, and the luminance values in the class section (“A + W” to “B−W”) excluding the vicinity of the threshold value are used. , 100% of the frequency will be distributed to the class. Therefore, with this method, as in the conventional case, in a scene where the histogram is biased and the luminance of the entire screen gradually shifts, the histogram may change drastically.

  A “conventional example 2” in FIG. 3 shows a histogram created by the method of Patent Document 2. The number of bits of image data, the class width, the number of classes, and the luminance change of the entire image are the same as those in “Conventional Example 1”. Here, the range W = 4. In this example as well, the histogram does not change so much in the image in which the overall brightness is reduced by 6 with respect to the histogram of the original image, but in the image in which the overall brightness is reduced by 12, the class 10 The frequency becomes 0, and the histogram changes drastically.

An object of the present invention is to perform image processing using a histogram generation circuit in which a histogram does not change drastically even in a scene where the histogram is biased and the luminance of the entire screen gradually shifts. Is to provide a device.

To achieve the above object,請 image processing apparatus Motomeko according to first invention, a plurality of image data input consisting of upper bits and lower bits indicating the respective luminance values of a plurality of pixels forming the screen of a predetermined area And a histogram generation circuit for generating a histogram indicating the frequency distribution of the luminance value for each class divided by the value of the upper bit, and the screen is displayed on the display device according to the histogram generated by the histogram generation circuit. A calculation circuit for generating conditions for correction processing to be performed, and the histogram generation circuit receives the lower bits and generates three distribution frequencies according to the values of the lower bits; , enter the upper bits, to each of the distribution based on class and class and below determined in accordance with the value of the upper bits, the frequency generating circuit The corresponding distribution frequency of the generated the three distribution frequencies by accumulating, and a cumulative addition circuit which generates the frequency of each class, the frequency generating circuit, the luminance value is the distribution source hierarchy In any of the positions, a distribution frequency exceeding 0 that is distributed to at least one of the upper and lower classes is generated.
Such invention in claim 2, the image processing apparatus according to claim 1, wherein the frequency generating circuit, when the value of the lower bit indicates the minimum value is distributed to the class under the distribution source class distribution When the frequency is maximized and the distribution frequency distributed to the upper class of the distribution source class is minimized, and the value of the lower bit indicates the maximum value, the distribution frequency distributed to the lower class of the distribution source class The distribution frequency to be distributed to each of the upper and lower classes of the distribution source class when the distribution frequency is maximized and the value of the lower bit indicates an intermediate value. The frequency is set to exceed 0.
According to a third aspect of the present invention, in the image processing apparatus according to the first or second aspect, a distribution frequency distributed to the distribution source class, a distribution frequency distributed to a class above the distribution source class, and the distribution source It is characterized in that the total value of the distribution frequencies distributed to the lower classes is set to a constant value .

According to the present invention, there is a bias in histogram, even a scene such as brightness of the entire screen is shifted gradually, never histogram changes sharply increases. For this reason, conditions for performing display correction using the obtained histogram can be appropriately generated.

  In the present invention, it is noted that (a) information of lower-order bits of image data is also reflected in the histogram result, (b) the number of classes is not increased as much as possible, and (c) the total frequency is not different depending on the image pattern. Pay.

  In the conventional histogram generation, for example, if the image data is 8 bits, the resolution is 256, and if the class width is 16, the number of classes is 16, so that a specific pixel (hereinafter referred to as “distribution target”). In accordance with, for example, the upper 4 bits of image data of “pixel”, it is determined to which class the distribution target pixel belongs, and the frequency is only 1 for the class to which it belongs (hereinafter referred to as “distribution source class”). Cumulative addition. At this time, the lower 4 bits of the image data are unnecessary.

  On the other hand, in the present invention, in order to solve (a) and (b), the distribution source class to which the distribution target pixel belongs is determined according to, for example, the upper 4 bits of the image data of the distribution target pixel. The frequency of the distribution target pixel is distributed to three frequencies weighted according to the lower 4 bits of the image data, and the three distribution frequencies are distributed and accumulated in the distribution source class and the upper and lower classes. I decided to add. As a result, the number of classes is increased by two from the conventional level. In order to solve (c), the total value of the three distribution frequencies generated from the lower 4 bits is assumed to be constant, and the total value of the frequencies of all classes changes only depending on the image size.

  By adopting this method, the number of classes increases by two, but the continuity of the histogram is improved, so that the histogram does not change drastically in a scene where the brightness gradually changes.

  The histogram calculation is performed in the following order. In the following, it is assumed that the input image data is 8 bits and the number of histogram classes is “16 + 2”.

(1) Based on the upper 4 bits [7: 4] of the image data, a class for accumulating the frequencies is determined. There are three classes, Histo [n-1], Histo [n], and Histo [n + 1]. Histo [n] indicates the source class, Histo [n-1] indicates the class below it, and Histo [n + 1] indicates the class above it. n is a value determined by the upper 4 bits [7: 4] of the 8-bit image data, and is derived by the following equation (1).
n = [7: 4] +1 (1)
Here, n is 1-16.

  (2) Based on the lower 4 bits [3: 0] of the image data, three distribution frequency values to be added are determined from the LUT, and the three classes Histo [n-1], Histo [n], Histo [n Add to +1]. For one image data, the total value of the frequencies added to each class Histo [n−1], Histo [n], Histo [n + 1] is a constant value (α). FIG. 1 shows the relationship between the lower 4 bits and the distribution frequency added to each class Histo [n-1], Histo [n], Histo [n + 1]. Here, α = 100. This content is set in the LUT. The distribution frequency down_val added to the lower class Histo [n−1] is the minimum (0) when the value of the lower 4 bits is the maximum (15), and increases as the value of the lower 4 bits decreases, and the maximum 50 ( = Α / 2). The distribution frequency up_val to be added to the upper class Histo [n + 1] is the minimum (0) when the value of the lower 4 bits is the minimum (0), and increases as the value of the lower 4 bits is increased to the maximum 50 (= α / 2). The distribution frequency middle_val added to the distribution source class Histo [n] is the largest when the value of the lower 4 bits is intermediate (7 or 8), but is not 100, and the remaining frequencies are the upper and lower classes Histo [n-1]. , Histo [n + 1] distribution frequency is divided into down_val and up_val.

  (3) The histogram result is corrected by the screen size (total number of pixels) to generate a final histogram.

  In the above example, the image data indicating the luminance value of each pixel is 8 bits, the distribution source class is determined by the upper 4 bits, and the distribution frequency is determined by the lower 4 bits. The width and α described above can be set arbitrarily. For example, when the image data is 10 bits, the distribution source class is determined with the most significant 4 bits [9: 6] as the upper bits, and the distribution frequency is determined with the lower 5 bits [5: 2] as the lower bits. The least significant 1 bit [0] may not be used.

<Example>
FIG. 2 shows the configuration of an image processing apparatus according to one embodiment of the present invention. The input image data is 8 bits. A histogram generation circuit 10 includes a frequency generation circuit 11 and a cumulative addition circuit 12.

  The frequency generation circuit 11 inputs the lower 4 bits [3: 0] of the image data, and adjoins three classes Histo [n-1], Histo [n], Histo by the LUT having the characteristics shown in FIG. Create distribution frequencies down_val, middle_val, and up_val to be distributed to [n + 1]. For example, if the value of the lower 4 bits is 4, the distribution frequencies are down_val = 22, middle_val = 73, and up_val = 5.

The accumulative addition circuit 12 receives the upper 4 bits [7: 4] of the image data, thereby obtaining n from the equation (1) and determining the distribution source class Histo [n]. This three classes Histo determined by [n-1], Histo [ n], the Histo [n + 1], min Hyde number down_val input from the frequency generation circuit 11, middle_val, up_val is cumulatively added The For example, if n = 12 is determined in the upper 4 bits [7: 4], the distribution frequency of 22 is added to Histo [11], 73 is added to Histo [12], and 5 is added to Histo [13].

  By generating the distribution frequency in the frequency generation circuit 11 and the cumulative addition in the cumulative addition circuit 12 with respect to the image data of each pixel constituting the screen, and further correcting by the screen size, A histogram is generated.

  An arithmetic circuit 20 generates a condition for performing correction (image correction and backlight brightness correction) according to the obtained histogram.

  Reference numeral 30 denotes an image processing unit, which corrects input image data based on conditions for performing image correction processing obtained by the arithmetic circuit 20 on the input 8-bit image data.

  Reference numeral 40 denotes an LCD module as an image display device, which includes an LCD panel 41 and a backlight control unit 42, and controls the backlight based on the correction conditions generated by the arithmetic circuit 20.

  The “present example” in FIG. 3 shows a histogram created by the present example. The number of bits of the image data, the class width, and the luminance change of the entire image are the same as those of “Conventional example 1” in FIG. However, the number of classes is “16 + 2”. Further, the value of α is different from that in FIG. In this embodiment, both the histogram of the image in which the overall brightness of the screen is reduced by 6 and the histogram of the image in which the overall brightness is reduced by 12 are greatly changed from the histogram of the original image. Absent.

  Specifically, the peak position moves from class 9 to class 8 as the overall luminance decreases by 6 (middle) and 12 (lower) from the histogram of the original image (upper). However, the overall shape of the histogram has not changed significantly. The power of class 10, which contained a large power in the histogram of the original image, gradually decreased as the overall luminance decreased by 6 and 12, and “conventional example 1”. Or, unlike “Conventional example 2”, it does not suddenly become zero. Therefore, it is necessary to set the calculation in the arithmetic circuit 20 so that the correction condition also gradually changes.

It is a characteristic diagram of the frequency which distributes to three classes Histo [n-1], Histo [n], and Histo [n + 1] which are adjacent by the lower 4 bits of image data. 1 is a block diagram illustrating a configuration of an image processing apparatus according to one embodiment of the present invention. It is a characteristic view which shows an example of the histogram of "this Example", "conventional example 1", and "conventional example 2". It is explanatory drawing of the conventional histogram creation.

Explanation of symbols

10: Histogram generation circuit, 11: Frequency generation circuit, 12: Cumulative addition circuit, 20: Arithmetic circuit, 30: Image processing unit, 40: LCD module, 41: LCD panel, 42: Backlight control unit

Claims (3)

A plurality of image data composed of upper bits and lower bits indicating the luminance values of a plurality of pixels forming a screen in a predetermined area are inputted, and the distribution of luminance values for each class divided by the values of the upper bits is shown. A histogram generation circuit for generating a histogram;
An arithmetic circuit for generating conditions for correction processing performed when the screen is displayed on a display device according to the histogram generated by the histogram generation circuit;
With
The histogram generation circuit includes:
A frequency generation circuit that inputs the lower bits and generates three distribution frequencies according to the value of the lower bits;
The upper bits are input, and the distribution frequency corresponding to the distribution frequency generated by the frequency generation circuit is assigned to each of the distribution source class determined according to the value of the upper bits and the upper and lower classes. A cumulative addition circuit that performs cumulative addition and generates a frequency for each class;
The frequency generation circuit generates a distribution frequency exceeding 0 distributed to at least one of the upper and lower classes when the luminance value is in any position in the distribution source class. Equipment . The frequency generation circuit includes:
When the value of the lower bit indicates a minimum value, the distribution frequency distributed to the lower class of the distribution source class is maximized, and the distribution frequency distributed to the upper class of the distribution source class is minimized,
When the value of the lower bit indicates the maximum value, the distribution frequency distributed to the lower class of the distribution source class is minimized and the distribution frequency distributed to the upper class of the distribution source class is maximized,
The image processing apparatus according to claim 1, wherein when the value of the lower bit indicates an intermediate value, the distribution frequency distributed to each of the upper and lower classes of the distribution source class exceeds zero . Distribution frequency distributed to the distribution source class and distribution frequency distributed to the class above the distribution source class
And the distribution frequency distributed to the lower class of the distribution source class is a constant value.
The image processing apparatus according to claim 1 or 2, wherein

Images