CN106686320A - A Tone Mapping Method Based on Number Density Equalization - Google Patents

Abstract

The invention discloses a tone mapping method based on numerical density balance, comprising the steps of acquiring a high-dynamic image; dividing a brightness range of the high-dynamic image into multiple brightness levels; when the high-dynamic image is a dark scene image, dividing the brightness range of the high-dynamic image into a first brightness range and a second brightness range, mapping the first brightness range to the first brightness range of a low-dynamic image, and mapping the second brightness range to the second brightness range of low-dynamic image; when the high-dynamic image is a non-dark scene image, mapping the brightness range of the high-dynamic image to the brightness range of the low-dynamic image. The tone mapping method has high computing speed, is simple to implement and is widely applicable.

Description

A Tone Mapping Method Based on Number Density Equalization

technical field

The invention relates to the field of image and video data processing, in particular to a tone mapping method based on number density equalization.

Background technique

High dynamic range images can reveal details of extremely dark and extremely bright areas that may be lost by traditional low dynamic images, but can be perceived by the human visual system. It correctly represents real-world brightness ranges. However, the dynamic range of traditional image sensors is generally only 102 orders of magnitude. Therefore, due to the limitation of display devices, high dynamic images cannot be displayed on conventional devices, so it is very meaningful to perform tone mapping on high dynamic images.

Tone mapping compresses the brightness of high dynamic images to the acceptable range of traditional display devices, while retaining the details of the original image as much as possible, and finally achieves the effect that the image displayed on the device is consistent with the scene observed by human eyes. Existing tone mapping algorithms mainly include global algorithms and local algorithms.

The global algorithm performs the same transformation algorithm on all pixels in the high dynamic image, usually using an S-curve. The global algorithm is fast and easy to implement, but it greatly reduces the image contrast, easily loses detailed information, cannot simulate the real visual corresponding characteristics well, and is not effective for processing complex scenes and images with extremely high dynamic range.

Local algorithms perform different transformations for different regions of the image. When adjusting the gray value of a point in an image, the spatial position of that point is taken into account. Usually a method based on analog photography is used. The calculation amount of the local algorithm is complex, the efficiency is low, and the application range is relatively narrow.

Contents of the invention

In order to overcome the above-mentioned defects of the prior art, the present invention provides a tone mapping method based on number density equalization, which utilizes low dynamic range (LDR) display devices to reproduce high dynamic range (HDR) image content in tone.

The tone mapping method based on number density equalization proposed by the present invention comprises the steps of: obtaining a high dynamic image; dividing the brightness range of the high dynamic image into multiple brightness levels; when the high dynamic image is relatively dark During the scene image, the brightness range of the high dynamic image is divided into a first brightness range and a second brightness range, the first brightness range is mapped to the first brightness range of the low dynamic image, and the second brightness range is Mapping to the second brightness range of the low dynamic image; when the high dynamic image is a non-dark scene image, mapping the brightness range of the high dynamic image to the brightness range of the low dynamic image.

The tone mapping method based on number density equalization of the present invention has the following advantages: 1. The calculation speed is fast and the implementation is simple. The present invention uses the number density equalization method for tone mapping, which has a small amount of calculation and fast operation speed; 2. It is suitable for a wide range of scenarios, The present invention processes images of different scenes in different ways, which can not only adapt to scenes with sufficient light, but also adapt to darker scenes, effectively retain detailed information, and can well simulate real visual corresponding characteristics. Images with dynamic range are ideal.

Description of drawings

In order to further illustrate the technical method involved in the present invention. We provide a detailed algorithm flow chart, and also give a specific processing example, and the processing process is carried out in strict accordance with the steps described in this disclosure.

FIG. 1 is an algorithm flow chart of a tone mapping method based on number density equalization.

Figure 2 is an example of a brightness-number density histogram of a high dynamic image.

Figure 3 is an underexposed image (left) and an overexposed image (right).

Figure 4 is a practical example of a brightness-number density histogram of a high dynamic image.

Figure 5 is a practical example of the brightness-number density histogram of the low dynamic image after the mapping is completed.

Figure 6 is the low dynamic image after the mapping is completed.

detailed description

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be described in further detail below in conjunction with specific embodiments and with reference to the accompanying drawings.

Fig. 1 is the algorithm flowchart of the tone mapping method based on number density equalization of the present invention, with reference to Fig. 1, this method comprises steps:

Step 100, first obtain a high dynamic range (HDR) image.

The image can be obtained directly from a digital device (camera, video camera, etc.) or synthesized through a certain algorithm. For color images, you can convert them to black and white first. Use I to represent a high dynamic image, its brightness range is [0, HB _max ], and HB _max is the maximum brightness of the high dynamic image.

Step 200, counting the total number of pixels at each brightness level to obtain the "brightness-number density" histogram of image I, that is, counting the total number of pixels at each brightness level. Figure 2 shows an example diagram of the brightness-number density histogram of a high dynamic image. This step further includes:

Step 201: Divide the brightness range [0, HB _max ] into M parts to obtain a new brightness range [0, M-1] of the high dynamic image, M is greater than 255, and the larger the M, the better the effect. According to the present invention, the preferred , M can choose a value greater than 1000. [0, HB _max ] is divided into M parts, so M subintervals are obtained.

Step 202: Count the number of pixels in each sub-interval, the pixel count of the i-th sub-interval is denoted as N _i , and Obtain the "brightness-number density" histogram of image I.

Step 300, calculate the average brightness of the high dynamic image I, and judge whether the image I is a dark scene image or a non-dark scene image according to the calculated average brightness.

The average brightness of the high dynamic picture I can pass Obtained by calculation, where i=1,2,...M, N _i is the pixel count of the ith sub-interval, and N is the total number of image pixels. Set the threshold for judging whether it is a dark scene image. The threshold can be set by the user. When the average brightness of the image is greater than the threshold, it is a non-dark scene image, otherwise it is a dark scene image.

In this step, according to the judgment result, if the image I is a dark scene image, execute step 400, and if it is not a dark scene image, otherwise execute step 500.

Step 400, when image I is a darker scene image, according to step 201, wherein the brightness range [0, HB _max ] of the high dynamic image is divided into M parts, and a new brightness range [0, M-1] of the high dynamic image is obtained . Tone map the HDR image to a low dynamic LDR image, specifically, map the brightness range [0,M-1] to [LB _min ,255], 0≤LB _min <255, where LB _min is the low Brightness minimum for dynamic (LDR) images. That is, divide [0,M-1] into Set the low dynamic image brightness threshold after tone mapping as th, {LB _min <th<255}, the threshold can be set by the user, and the corresponding high dynamic image brightness threshold is M _th . Step 400 further includes:

Step 401: In the brightness range of the high dynamic image, when the brightness is less than or equal to M _th , map the brightness range [0, M _th ] of the high dynamic image to the low dynamic image [LB _min , th]. Specifically, the mapping is performed in the following way, and the counting threshold is set Map [0,M ₀ ] intervals to LB _min , the counts of these intervals are just over NA , counts of excess NA are counted into the next interval, map [ _{M 0} +1, _{M 1} ] intervals to LB _min +1 , the counts of these intervals plus the excess count of the previous interval just exceed N _A . And so on, until [0,M _th ] is mapped

Step 402, in the brightness range of the high dynamic image, when the brightness is greater than M _th , the high dynamic image I brightness range [M _th +1, M-1] is mapped to the low dynamic image brightness range [th+1, 255], according to The mapped pixel number rest_pixel and the mapped low dynamic brightness level k are used to adjust the N _A value, N _A =rest_pixel/(256-k-LB _min ), k=th,th+1,...255- LB _min , map [M _th +1,M _th+1 ] intervals to LB _min +th+1, the counts of these intervals just exceed N _A , and the counts of excess N _A are included in the next interval; and so on, until The mapping of [M _th +1,M] is completed.

In step 500, the image I is a non-dark scene image, and the brightness range [0, HB _max ] is mapped to [LB _min , 255]. Set count threshold Map [0,M ₀ ] intervals to LB _min , the counts of these intervals are just over NA , counts of excess NA are counted into the next interval, map [ _{M 0} +1, _{M 1} ] intervals to LB _min +1 , the counts of these intervals plus the redundant counts of the previous interval just exceed N _A , and so on until [0,M] is mapped.

An example is used below to describe in detail.

Step 100: Synthesize a high dynamic image according to the underexposed image and the overexposed image in FIG. 3 .

Step 200: Count the total number of pixels of each brightness level to obtain the "brightness-number density" histogram of image I, that is, count the total number of pixels of each brightness level. Figure 4 shows an example diagram of its high dynamic image brightness-number density histogram ,Specific steps are as follows:

Step 201: Divide the brightness range into 5000 parts, so 5000 sub-intervals are obtained.

Step 202: Count the number of pixels in each sub-interval, the pixel count of the i-th sub-interval is denoted as N _i , and Obtain the "brightness-number density" histogram of image I, as shown in Figure 5.

In step 300, set the threshold value of whether it is a darker scene image to 2000, and calculate the average brightness of the high dynamic image I as 1255 through the formula N/totoal_pixel. Therefore, the image is a darker scene image, and step 400 is executed.

In step 400, according to step 201, a new brightness range [0, 4999] of the high dynamic image is obtained. Tone map an HDR image to a low dynamic LDR image, specifically, map the luminance range [0,4999] to [0,255]. That is to say, divide [0,4999] into 256 parts, set the low dynamic image brightness threshold after tone mapping to 30, the threshold can be adjusted, and the corresponding high dynamic image brightness threshold is M _th .

Step 401: In the brightness range of the high dynamic image, when the brightness is less than or equal to M _th , map the brightness range [0, M _th ] of the high dynamic image to the low dynamic image [0, 30]. Specifically, the mapping is performed in the following way, and the counting threshold is set Map [0,M ₀ ] interval to low dynamic brightness 0, M ₀ is determined as the count in [0,M ₀ ] interval just exceeds N _A , the count of excess N _A is included in the next interval, and [M ₀ +1,M ₁ ] intervals to which the counts plus the excess count from the previous interval just exceed N _A . And so on, until [0,M _th ] is mapped

Step 402, in the brightness range of the high dynamic image, when the brightness is greater than th, the high dynamic image I brightness range [M _th +1, 4999] is mapped to the low dynamic image brightness range [th+1, 255], according to the unmapped The number of pixels rest_pixel and the mapped low dynamic brightness level k are used to adjust the N _A value, N _A =rest_pixel/(256-k), k=th,th+1,...255, [M _th +1 ,M _th+1 ] intervals are mapped to th+ ₁ , the counts of these intervals just exceed NA, and the counts of excess NA are included in the next interval; and so on until [ _{M th} +1, 4999] is mapped. Figure 5 and Figure 6 are the "brightness-number density" histogram after mapping and the image after tone mapping respectively.

The specific embodiments described above have further described the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above descriptions are only specific embodiments of the present invention, and are not intended to limit the present invention. Within the spirit and principles of the present invention, any modifications, equivalent replacements, improvements, etc., shall be included in the protection scope of the present invention.

Claims (9)

1. A tone mapping method based on number density equalization, the method comprising steps: Step 100, obtaining a high dynamic image; Step 200, dividing the brightness range of the high dynamic image into multiple brightness levels; Step 400, when the high dynamic image is a dark scene image, divide the brightness range of the high dynamic image into a first brightness range and a second brightness range, and map the first brightness range to the low dynamic image a first brightness range, mapping the second brightness range to a second brightness range of a low dynamic image; Step 500, when the high dynamic image is a non-dark scene image, map the brightness range of the high dynamic image to the brightness range of the low dynamic image. 2. The method of claim 1, wherein, In step 100, the brightness range of the high dynamic HDR image is [0, HB _max ], where HB _max is the maximum brightness value of the high dynamic image; In step 200, the brightness range [0, HB _max ] is divided into M parts to obtain a new brightness range [0, M-1] of the high dynamic image, M is greater than 255, and the number of pixels in the M sub-intervals is counted, the i-th sub-interval The pixel count of is denoted N _i , and 3. The method of claim 2, wherein, In step 400, the first luminance range of the high dynamic image is [0,M _th ], the second luminance range is [M _th +1,M-1], and the first luminance range of the low dynamic image is [LB _min ,th ], the second brightness range of the low dynamic image is [th+1,255], the brightness range of the low dynamic image is [LB _min ,255], 0≤LB _min <255, wherein LB _min is the low dynamic image The minimum brightness value, the low dynamic image brightness threshold is th, and the corresponding high dynamic image brightness threshold is M _th . 4. The method according to claim 3, wherein step 400 further comprises: Step 401: In the brightness range of the high dynamic image, when the brightness is less than or equal to M _th , map the brightness range [0, M _th ] of the high dynamic image to the low dynamic image [LB _min , th]; Step 402, in the brightness range of the high dynamic image, when the brightness is greater than th, map the brightness range [M _th +1, M-1] of the high dynamic image to the brightness range [th+1, 255] of the low dynamic image. 5. The method of claim 4, wherein, In step 401, divide [0, M-1] into [0, M ₀ ], [M ₀ +1, M ₁ ], ..., [M _th-1 +1, M _th ], ..., Set count threshold Map the [0,M ₀ ] interval to LB _min , the pixel count in this interval just exceeds N _A , and the pixel count more than N _A is counted into the next interval, map the [M ₀ +1,M ₁ ] interval to LB _min +1, the pixel count of this interval plus the count of the previous interval more than N _A just exceeds N _A , and so on until [0,M _th ] is mapped; In step 402, in the brightness range of the high dynamic image, when the brightness is greater than M _th , the high dynamic image I brightness range [M _th +1, M-1] is mapped to the low dynamic image brightness range [th+1, 255], according to The unmapped pixel number rest_pixel and the mapped low dynamic brightness level k are used to adjust the N _A value, N _A =rest_pixel/(256-k-LB _min ), k=th,th+1,...255 -LB _min , map the [M _th +1,M _th+1 ] interval to LB _min +th+1, the pixel count in this interval just exceeds N _A , and counts more than N _A are included in the next interval, so that By analogy, until [M _th +1, M-1] is mapped. 6. The method of claim 2, wherein, In step 500, the brightness range [0, HB _max ] of the high dynamic image is mapped to the brightness range [LB _min , 255] of the low dynamic image, LB _min is the brightness minimum value of the low dynamic image, and the low dynamic image brightness threshold is th, and the corresponding high dynamic image brightness threshold is M _th , where [0, M-1] is divided into [0, M ₀ ], [M ₀ +1, M ₁ ], ..., [M _{th- 1} +1, _Mth ],..., Set count threshold Map the [0,M ₀ ] interval to LB _min , the pixel count in this interval just exceeds N _A , and counts more than N _A are included in the next interval, map the [M ₀ +1,M ₁ ] interval to LB _min +1, the pixel count of this interval plus the count of the previous interval more than N _A just exceeds N _A , and so on until [0,M-1] is mapped. 7. The method according to any one of claims 1-6, further comprising: Step 300, calculating the average brightness of the high dynamic image, and judging whether the image is a dark scene image or a non-dark scene image according to the calculated average brightness. 8. The method according to claim 7, wherein the average brightness of the high dynamic image is passed The method is calculated and obtained, where i=1,2,...M, N _i is the pixel count of the i-th subinterval, N is the total number of image pixels, and the threshold for judging whether it is a darker scene image is set. When the average brightness of the image is greater than the When the threshold is set, it is a non-darker scene image, otherwise it is a darker scene image. 9. The method according to any one of claims 2-6, characterized in that M is greater than 1000.