JP2007259148A - Frequency threshold setting apparatus and method for histogram projection processing, and recording medium recorded with program therefor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To obtain an image with a high contrast intensification effect in a low-contrast image. <P>SOLUTION: A video signal input unit outputs an input video signal from the outside to a histogram calculation unit and a level conversion unit. The histogram calculation unit calculates a luminance histogram for a luminance level average among a plurality of frames of each of pixels in the image signal and outputs the luminance histogram to a frequency threshold determination unit and a level conversion function calculation unit. The frequency threshold determination unit performs frequency division on the luminance histogram, calculates an average value of signal level differences within divided sections, and calculates differentials (ΔL2-ΔL1, ΔL3-ΔL2, ...) between adjacent sections of these signal level difference average values. If the differential of average values of signal level differences investigated from up to down is a fixed value RTH or smaller continuously for N differentials and a differential ΔLN+1-ΔLN exceeds the fixed value RTH in an (N+1)th frequency divided section, this image is determined as an image having a peak and the frequency at a time when the differential exceeds the fixed value RTH is calculated as a frequency threshold for the peak. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a histogram projection processing frequency threshold setting apparatus and method capable of displaying a low-contrast image more clearly, and a recording medium on which the program is recorded.

  Conventionally, an image acquired by a camera or the like often has a low contrast image. Since low-contrast images are difficult to see, various methods have been developed to increase contrast by using image processing techniques and display them more clearly. One of the image processes is a histogram projection (hereinafter referred to as HP) process.

  In the HP processing, a histogram representing the relationship between the signal level of the input video signal and the frequency of the signal level within one frame is generated, and the input video signal level is assigned so that only a signal level having a frequency threshold or more is assigned to the display gradation. This is a method of converting the display signal level.

6A and 6B are diagrams more specifically showing the conventional HP process. The equation for calculating this can be expressed by the following equation.
image_out (i) = hplevel (image_in (i))
Here, “image_out (i)” indicates the output level after processing of the i-th pixel, and “image_in (i)” indicates the input level of the i-th pixel. The coefficient i represents i = 1, 2,... N, and N represents the total number of pixels.

Further, “hlevel (x)” is a function for converting the level of a pixel whose input level is x into a display gradation, and the expression hplevel (x) = GSdisp / GStotal × (GS (x) −1)
Is required. Here, GSdisp is the number of display gradations, and GStotal is all gradations after truncation. GS (x) = Σ [j = 0-x] (cnt (j))
Represents the number of gradations in which the number of pixels is less than or equal to x and exceeds HPth (frequency threshold), and pop (y) is the number of pixels of the input level y,
cnt (y) = 1 | pop (y) ≧ HPth
= 0 | pop (y) <HPth
Represents a flag indicating whether or not the number of pixels of the input level y exceeds the frequency threshold HPth.

  The input level “image_in (i)” of the i-th pixel follows the function “hlevel (x)” for converting the input level x to the display gradation, and the output level “image_out (i)” after the processing of the i-th element. Is converted to

  “Hplevel (x)” is a value obtained by subtracting 1 from the number of gradations “GS (x)” in which the number of pixels exceeds the frequency threshold HPth at the input level x or less, and the total number of gradations “GStotal” after rounding down. Is multiplied by the ratio so as to be effectively assigned to the display gradation number “GSdisp”.

  “GS (x)” is a flag “cnt (y)” indicating whether or not the number of pixels at the input level y exceeds the frequency threshold HPth when the number of pixels “pop (y)” at the input level y is equal to or greater than HPth. Is set to 1, and the count value “Σ [j = 0 to x] (cnt (j))” is counted up.

The frequency threshold HPth in the above-described HP processing is a fixed value for each scene of the input image, and is normally set to a low value so as to leave many input video signal levels.
FIG. 6 (a) shows the relationship between the frequency threshold when the frequency threshold of the luminance histogram is set to a low value and the luminance histogram when the luminance histogram is an image having a peak. FIG. 6 is a diagram showing an output level of an image subjected to HP processing based on the frequency threshold value in FIG.

  As shown in Fig. 6 (a), when the frequency threshold is always fixed to a low value, the output level of an image having a peak in the luminance histogram is shown in Fig. 6 (b). In this case, the peak portion is not utilized, and the proportion of gradations with no significant difference increases, and the contrast of the entire image decreases.

As a method for enhancing the contrast of an image, a method has been proposed in which a unique process is performed on a histogram of image data to enable a clear and sharp display even in an image with a narrow dynamic range. (For example, refer to Patent Document 1.)
JP 2004-212809 (Abstract, paragraphs 0028-0030, FIG. 1)

    As described in paragraphs 0026 to 0030, the technique of Patent Document 1 is a process of cutting off gradation levels only on the low luminance pixel side and the high luminance pixel side and converting them to display gradations, and there are few intermediate luminance pixels. In some cases, the gradation level of the intermediate luminance cannot be cut off, so that it cannot be effectively converted to the display gradation, and the contrast enhancement effect is reduced. Further, in the conventional HP processing method, as described above, since the frequency threshold is preset for each scene image, it cannot cope with a frequent change in the luminance level accompanying a sudden change in the scene. In addition, since the set frequency threshold is generally a low value, an image having a peak in the luminance histogram cannot be clearly and clearly displayed.

  In view of the above-described conventional situation, the problem of the present invention is that the frequency threshold of the luminance distribution is changed and set to an appropriate value according to the shape of the histogram of each scene, so that the contrast enhancement effect can be obtained more clearly even for low-contrast images. It is an object to provide a histogram projection processing frequency threshold setting apparatus and method capable of obtaining a high image, and a recording medium recording the program.

  First, a recording medium according to a first aspect of the present invention is a luminance histogram calculation function for calculating a luminance histogram with respect to an average luminance level between a plurality of frames of each pixel of input image data, and the generated luminance histogram has a certain peak or more. A peak presence / absence determination function for determining whether or not the peak has a certain peak or more, a peak rising frequency acquisition function for determining the frequency of the rising portion of the peak, and the peak is one. When the frequency of the rising portion of the peak is plural, and when the number of the peaks is plural, the minimum frequency among the frequencies of the rising portions of the respective peaks is set as the frequency threshold for the histogram projection processing of the input image data. The frequency threshold setting function and when there is no peak above a certain level, the frequency is accumulated in the order of the most frequent gradation. A program for executing a peakless frequency threshold setting function for setting the addition frequency when the predetermined ratio of the total number of pixels exceeds a frequency threshold for histogram projection processing of the input image data to be read by a computer is recorded. Has been.

  Next, a histogram threshold value setting method for histogram projection processing according to a second aspect of the present invention is a luminance histogram calculation step for calculating a luminance histogram with respect to an average luminance level between a plurality of frames of each pixel of input image data, A peak presence / absence determination step for determining whether or not the luminance histogram has a peak above a certain level, and a peak rising frequency acquisition step for determining the frequency of the rising portion of the peak when the luminance histogram has a peak above a certain level, When the number of peaks is one, the frequency of the rising portion of the peak, and when the number of peaks is plural, the minimum frequency among the frequencies of the rising portion of each peak is used for histogram projection processing of the input image data. The peak frequency threshold setting process to set as the frequency threshold and the peak above a certain level If there is no peak, the frequency is accumulated in order of the most frequent gray levels, and the addition frequency when the predetermined ratio of the total number of pixels is exceeded is set as the frequency threshold for histogram projection processing of the input image data. And comprising.

  Further, the histogram threshold value setting device for histogram projection processing according to the third aspect of the present invention is a luminance histogram calculation means for calculating a luminance histogram with respect to an average luminance level between a plurality of frames of each pixel of the input image data, and the generated luminance Peak presence / absence determining means for determining whether or not the histogram has a certain peak or higher, peak rising frequency obtaining means for obtaining the frequency of the rising portion of the peak when the luminance histogram has a certain or higher peak, and the above When there is one peak, the frequency at the rising portion of the peak, and when there are a plurality of peaks, the minimum frequency among the frequencies at the rising portion of each peak is used for histogram projection processing of the input image data. There is a peak frequency threshold setting means to set as a frequency threshold, and peaks above a certain level If there is no peak, the frequency is accumulated in the order of the most frequent gradations, and the addition frequency when the predetermined ratio of the total number of pixels is exceeded is set as a frequency threshold for histogram projection processing of the input image data. And is configured.

  According to the present invention, the frequency threshold of the luminance distribution is changed and set to an appropriate value depending on the shape of the histogram of each scene, so that an image with a high contrast enhancement effect can be obtained even for a low-contrast image. It is possible to provide a histogram projection processing frequency threshold setting apparatus and method, and a recording medium on which the program is recorded.

  Embodiments of the present invention will be described below with reference to the drawings.

  FIG. 1 is a block diagram of a histogram projection (hereinafter referred to as HP) processing system according to an embodiment. As shown in the figure, the HP processing system includes a video signal input unit 1, a histogram calculation unit 2, a level conversion unit 3, a frequency threshold determination unit 4, and a level conversion function calculation unit 5.

A video signal is input from the outside to the video signal input unit 1 for each frame. The video signal input unit 1 outputs the input video signal to the histogram calculation unit 2 and the level conversion unit 3.
The histogram calculation unit 2 calculates a luminance histogram with respect to an average luminance level between a plurality of frames of each pixel of the image signal input from the video signal input unit 1, and uses the calculated luminance histogram as a frequency threshold determination unit 4 and a level conversion function Output to the calculation unit 5.

The calculation of the luminance histogram performed by the histogram calculation unit 2 is a process performed by a normal HP process.
The frequency threshold determination unit 4 uses the luminance histogram input from the histogram calculation unit 2 to perform frequency division, a difference check, peak determination, and threshold setting (calculation) described later.

  FIG. 2 is a diagram illustrating an example of peak determination and frequency threshold calculation for an image having a peak in the luminance histogram. In the figure, the horizontal axis indicates gradation and the vertical axis indicates frequency. The gradation, that is, the signal level is represented by 256 gradations having a value of 0 to 255 in an 8-bit configuration in this example.

  The frequency threshold value determination unit 4 first extracts a global change by applying a low-pass filter to the luminance histogram. Next, the frequency maximum value of the extracted luminance histogram is calculated. Then, the maximum frequency value is divided by a certain fixed numerical value as shown in FIG.

  Then, an average value ΔL1 within the distribution interval of the signal level difference is calculated from the frequency division section at the top of the luminance histogram, and then within the distribution section of the signal level difference of the second frequency division section from the frequency division section at the top. The average value (ΔL 1, ΔL 2, ΔL 3,...) Of the signal level difference is sequentially calculated up to the lowest frequency division section of the luminance histogram so that the average value ΔL 2 is calculated.

  For example, in the first frequency division section, assuming that the signal level of the maximum frequency is a, the minimum frequency is b, and b is A level and B level (B> A)) (assuming that the A level is near the gradation 118) , B level is in the vicinity of the gradation 138), and the signal level difference in b is (B−A). Next, the signal level difference in (b + 1) is obtained and added to (B−A). This is repeated up to a, and the value obtained by dividing the total added value by (ab) is the average value ΔL1 of the signal level difference.

  The result is assumed to be 138-118 = 20. Further, when the average value ΔL1 of the signal level difference in the first frequency division section is smaller than a predetermined value, there may be a peak here.

Subsequent to the above, the difference (ΔL 2 −ΔL 1, ΔL 3 −ΔL 2,...) Between adjacent sections of the calculated signal level difference average value is sequentially calculated.
For example, in the second frequency division section, assuming that the maximum frequency is c, the minimum frequency is d, and the signal level of d is C level and D level (D> C)) (assuming that C level is gradation 113). The signal level difference between d and d is (D−C). Next, the signal level difference in (d + 1) is obtained and added to (D−C). This is repeated up to c, and the value obtained by dividing the total added value by (cd) is the average value ΔL2 of the signal level difference. The result is assumed to be 143-113 = 30.

  Then, the difference “ΔL2−ΔL1” (the provisional value “30-20 = 10”) of the average value of the signal level difference between the frequency division sections adjacent above and below is higher than a constant value RTH (for example, 20). Checking whether or not it is large is sequentially performed below the frequency division section.

  Then, the difference of the average value of the examined signal level differences is N consecutively less than or equal to a certain value RTH, and the signal level difference between the N + 1th frequency division section and the Nth frequency division section When the difference between the average values of these exceeds a certain value RTH, it is determined that the image of one frame is an image having a peak. Then, the frequency when the constant value RTH is exceeded is calculated as the frequency threshold for the peak.

  For example, in the Nth frequency division section, if the maximum frequency is i, the minimum frequency is j, and the signal level is j and the signal level is I level and J level (J> I), the I level is assumed to be near the gradation 85. , The D level is in the vicinity of the gradation 166), and the signal level difference at j is (J-I). Next, the signal level difference in (j + 1) is obtained and added to (J−I). This is repeated up to i times, and the value obtained by dividing the total sum by (i−j) is the average value ΔLN of the signal level difference. The result is assumed to be 166-85 = 81.

  Similarly, for example, in the (N + 1) th frequency division section, if the maximum frequency is m, the minimum frequency is n, and the signal level of n is M level and N level (N> M)) (assuming that M level is In the vicinity of gradation 64, the N level is assumed to be near gradation 188), and the signal level difference at n is (N−M). Next, a signal level difference in (n + 1) is obtained and added to (NM). This is repeated up to m, and the value obtained by dividing the total sum by (mn) is the average value ΔLN + 1 of the signal level difference. The result is assumed to be 188−64 = 124.

  If the difference “ΔLN + 1−ΔLN” (provisional value “124−81 = 43”) of the average signal level difference between the adjacent frequency division sections is larger than a predetermined constant value RTH. The frequency exceeding this fixed value RTH (the intermediate value between the Nth frequency division section and the N + 1th frequency division section) is calculated as the frequency threshold value for the peak.

Note that the frequency at which “ΔLN + 1−ΔLN” exceeds a certain value RTH indicates the frequency at which the peak of the luminance histogram starts to rise as seen from the lower gradation.
FIG. 3 (a) is a diagram in which the frequency threshold is set by the above-described method with respect to the input luminance histogram shown in FIG. 6 (a), and FIG. 3 (b) shows the resulting output luminance histogram. FIG.

  The frequency threshold calculated as described above is output from the frequency threshold determination unit 4 to the level conversion function calculation unit 5 in FIG. The level conversion function calculation unit 5 calculates a level conversion function based on the luminance histogram input from the histogram calculation unit 2 and the frequency threshold input from the frequency threshold determination unit 4.

  In the calculation of the level conversion function, first, as shown in FIG. 3A, the frequencies of the gradation portions a, b, c below the frequency threshold are cut off by the frequency threshold. The gradation portions a, b, and c with the frequencies truncated are unused gradation portions that are not used as they are. A level conversion function is a function for allocating the frequencies of the remaining gradation portions d and e to the remaining gradation portions.

  This level conversion function is output to the level conversion unit 3 in FIG. The level conversion unit 3 converts the signal level of the input video signal input from the video signal input unit 1 based on the level conversion function input from the level conversion function calculation unit 5 and outputs the signal level to the display device.

  FIG. 3B shows the output luminance histogram. As described above, according to the present embodiment, the frequency threshold is set according to the scene image. For example, for a scene image having a peak in the luminance histogram, the frequency threshold is set so as to leave a peak portion. Since it is set high, an image with a high contrast enhancement effect can be obtained.

  Also, since the luminance histogram is calculated for the average level between a plurality of frames of each pixel and the above processing is performed, the scene of the input video changes rapidly and the luminance level of the entire screen changes frequently. Even in such a case, the luminance level can be gradually changed so as to be linked to the preceding and following scenes, and an easy-to-view image can be obtained even when the luminance level of the input video frequently changes.

  In the example shown in FIG. 3 (a), in order to easily explain the processing for an image having a peak in the luminance histogram, in the case where there are two peaks, the frequency that happens to be calculated from the distribution change of the two peaks. Although the case where the threshold values are the same is taken up, the case where the frequency threshold value is different for each peak will be described below by taking a case where there are typically three peaks as an example.

  FIG. 4 is a diagram showing a setting example of the frequency threshold value of an image having a plurality of peaks (three in the example in the figure) in the luminance histogram. As shown in the figure, as a result of performing the processing of the present invention described in FIG. 2 and FIGS. 3A and 3B on an image having three peaks 6, 7, and 8 in the luminance histogram, 3 different for each. Two frequency thresholds Fth1, Fth2, and Fth3 are calculated, and it is assumed that Fth1 <Fth2 <Fth3.

  In this case, the frequency threshold Fth1 calculated for the first peak 6 as the minimum frequency threshold is set as the frequency threshold of this image. Thereby, the enhancement effect of an image having a plurality of peaks can be enhanced, and more input video signal levels can be left.

Next, setting of the frequency threshold for an image having no peak in the luminance histogram will be described.
FIG. 5 is a diagram illustrating an example of setting the frequency threshold of an image having no peak in the luminance histogram. In the case of an image having no peak in the luminance histogram, the frequency threshold value determination unit 4 in FIG. 1 first accumulates the frequencies in the order of the gradations with the highest frequency (descending order). Then, the addition frequency when R% of the total number of pixels is exceeded is set as the frequency threshold. In addition, R is set in the range of 90-100.

  As described above, in this example, in a scene image in which the luminance histogram does not have a peak, the frequency threshold is set low so as to leave many input video signal levels.

1 is a block diagram of a histogram projection (HP) processing system in one embodiment. FIG. It is a figure which shows the peak determination in the case of the image which has a peak in a brightness | luminance histogram, and the calculation example of a frequency threshold value. (a) is a diagram in which a frequency threshold is set for a peaked input luminance histogram by the method of the present invention, and (b) is a diagram showing an output luminance histogram as a result. It is a figure which shows the example of a setting of the frequency threshold value of the image which has a some peak in a brightness | luminance histogram. It is a figure which shows the example of a setting of the frequency threshold value of the image without a peak in a brightness | luminance histogram. (a), (b) is a figure which shows the conventional HP process more concretely.

Explanation of symbols

1 Video signal input unit 2 Histogram calculation unit 3 Level conversion unit 4 Frequency threshold determination unit 5 Level conversion function calculation unit 6, 7, 8 Peak of luminance histogram

Claims (3)

A luminance histogram calculation function for calculating a luminance histogram with respect to an average luminance level between a plurality of frames of each pixel of the input image data;
A peak presence / absence determination function for determining whether or not the generated luminance histogram has a certain peak or more;
When the brightness histogram has a peak above a certain level, a peak rising frequency acquisition function for determining the frequency of the rising portion of the peak;
When the number of peaks is one, the frequency of the rising portion of the peak, and when there are a plurality of peaks, the minimum frequency of the frequencies of the rising portion of each peak is used for histogram projection processing of the input image data. Peak frequency threshold setting function to set as the frequency threshold,
When there is no peak above a certain level, the frequency is accumulated in the order of the most frequent gradations, and the addition frequency when a predetermined ratio of the total number of pixels is exceeded is set as a frequency threshold for histogram projection processing of the input image data None frequency threshold setting function,
A recording medium recorded with a computer readable program. A luminance histogram calculating step of calculating a luminance histogram for the average luminance level between a plurality of frames of each pixel of the input image data;
A peak presence / absence determination step for determining whether or not the generated luminance histogram has a certain peak or more;
When the luminance histogram has a peak above a certain level, a peak rising frequency obtaining step for obtaining the frequency of the rising portion of the peak; and
When the number of peaks is one, the frequency of the rising portion of the peak, and when there are a plurality of peaks, the minimum frequency of the frequencies of the rising portion of each peak is used for histogram projection processing of the input image data. A peak presence frequency threshold setting step for setting as a frequency threshold,
When there is no peak above a certain level, the frequency is accumulated in the order of the most frequent gradations, and the addition frequency when a predetermined ratio of the total number of pixels is exceeded is set as a frequency threshold for histogram projection processing of the input image data None frequency threshold setting step,
A frequency threshold value setting method for histogram projection processing. A luminance histogram calculating means for calculating a luminance histogram with respect to an average luminance level between a plurality of frames of each pixel of the input image data;
Peak presence / absence determining means for determining whether or not the generated luminance histogram has a certain peak or more;
When the luminance histogram has a peak above a certain level, a peak rising frequency acquisition means for determining the frequency of the rising portion of the peak;
When the number of peaks is one, the frequency of the rising portion of the peak, and when there are a plurality of peaks, the minimum frequency of the frequencies of the rising portion of each peak is used for histogram projection processing of the input image data. A peak presence frequency threshold setting means for setting as a frequency threshold;
When there is no peak above a certain level, the frequency is accumulated in the order of the most frequent gradations, and the addition frequency when a predetermined ratio of the total number of pixels is exceeded is set as a frequency threshold for histogram projection processing of the input image data None frequency threshold setting means,
A frequency threshold setting device for histogram projection processing characterized by comprising: