WO2021217463A1

WO2021217463A1 - Image processing algorithm device, image processing method, and camera

Info

Publication number: WO2021217463A1
Application number: PCT/CN2020/087625
Authority: WO
Inventors: 廖文山; 李静; 郭浩铭; 卢庆博
Original assignee: 深圳市大疆创新科技有限公司
Priority date: 2020-04-28
Filing date: 2020-04-28
Publication date: 2021-11-04

Abstract

Disclosed in embodiments of the present application is an image processing algorithm device. The device comprises a first image processing layer and a second image processing layer, wherein the first image processing layer comprises a common algorithm module, and the common algorithm module is an algorithm module corresponding to a same image processing function in at least two image processing algorithms; the second image processing layer comprises other algorithm modules of different image processing algorithms; the first image processing layer is used for processing an obtained first image by using the common algorithm module to obtain an image data set corresponding to the first image; and the second image processing layer is used for processing, by using other algorithm modules, corresponding target data obtained from the image data set. The device provided in the embodiments of the present application solves the technical problem of the waste of computing resources caused by repeatedly executing common steps in different image processing algorithms when an image is processed by using the image processing algorithms.

Description

Image processing algorithm device, image processing method and camera

Technical field

This application relates to the field of image processing, and in particular to an image processing algorithm device, an image processing method, a camera, and a computer-readable storage medium.

Background technique

In the field of image processing, there are a variety of image processing algorithms used to achieve different image processing effects. At the software level, each image processing algorithm usually corresponds to an algorithm module. During application, the required algorithm module can be easily called to process the image.

However, each image processing algorithm contains multiple image processing steps, and some steps in different image processing algorithms may be the same, or in other words, the image processing functions corresponding to these steps are the same. These steps can be referred to as shared steps. Thus, when the algorithm modules corresponding to these image processing algorithms are used to process images, each algorithm module independently executes each step in their respective algorithm, then in the entire image processing process, there are common steps Will be repeatedly executed, resulting in a waste of computing resources.

Summary of the invention

In view of this, embodiments of the present application provide an image processing algorithm device, an image processing method, a camera, and a computer-readable storage medium, which are used to solve the common steps in different image processing algorithms when image processing algorithms are used to process images. It is repeatedly executed, causing a technical problem of wasting computing resources.

The first aspect of the embodiments of the present application provides an image processing algorithm device for processing an image using at least two image processing algorithms, each of the image processing algorithms includes at least two algorithm modules, and each of the algorithm modules corresponds to Different image processing functions in the image processing algorithm; the device includes a first image processing layer and a second image processing layer, the first image processing layer includes a common algorithm module, the common algorithm module is at least two kinds of Algorithm modules corresponding to the same image processing function in the image processing algorithm; the second image processing layer includes various other algorithm modules of the image processing algorithm;

The first image processing layer is configured to use the common algorithm module to process the acquired first image to obtain an image data set corresponding to the first image;

The second image processing layer is configured to use the other algorithm module to process the corresponding target data obtained from the image data set.

The second aspect of the embodiments of the present application provides an image processing method for processing an image by using at least two image processing algorithms, each of the image processing algorithms includes at least two algorithm modules, and each of the algorithm modules corresponds to The different image processing functions in the image processing algorithm; the method includes:

The shared algorithm module is used to process the acquired first image to obtain the image data set corresponding to the first image; wherein the shared algorithm module is at least two of the image processing algorithms corresponding to the same image processing function Module

Obtain target data corresponding to other algorithm modules of each of the image processing algorithms from the image data set, and use the other algorithm modules to process the corresponding target data.

A third aspect of the embodiments of the present application provides a camera, including: a body, a lens provided on the body, an image sensor, an ISP chip, and a DSP chip provided in the body;

The image sensor is used to collect an original image through the lens;

The ISP chip is used to obtain the original image from the image sensor, and process the original image to obtain a first image;

The DSP chip is used to obtain a first image from the ISP chip, and use a common algorithm module to process the first image to obtain an image data set corresponding to the first image; wherein, the common algorithm module is at least Algorithm modules corresponding to the same image processing function in the two image processing algorithms; respectively obtain target data corresponding to other algorithm modules of each of the image processing algorithms from the image data set, and use the other algorithms The module processes the corresponding target data.

The fourth aspect of the embodiments of the present application provides a computer-readable storage medium having a computer program stored on the computer-readable storage medium. Image processing method.

In the image processing algorithm device provided by the embodiment of the present application, the first image processing layer includes a shared algorithm module, and the shared algorithm module is an algorithm module shared by at least two image processing algorithms, that is, in the device provided by the embodiment of the present application, Common steps in different image processing algorithms can be separated as an algorithm module, and the algorithm module can be used as an algorithm module shared by these image processing algorithms at the same time. Then, in application, executing the shared algorithm module once is equivalent to executing the steps corresponding to the shared algorithm module once in each image processing algorithm. The image data set obtained by executing the shared algorithm module can be provided to each image. The sharing of processing algorithms realizes that without affecting the functional realization of various image processing algorithms, the number of executions of shared steps is reduced, and computing resources and memory resources are saved.

Description of the drawings

In order to more clearly describe the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings that need to be used in the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained from these drawings without creative labor.

FIG. 1 is a schematic structural diagram of an image processing algorithm device provided by an embodiment of the present application.

Fig. 2 is a schematic structural diagram of a Gaussian pyramid provided by an embodiment of the present application.

Fig. 3 is a flowchart of a fusion algorithm provided by an embodiment of the present application.

Fig. 4 is a flowchart of an image processing method provided by an embodiment of the present application.

Fig. 5 is a schematic structural diagram of a camera provided by an embodiment of the present application.

Detailed ways

The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of this application.

In the field of image processing, there are a variety of image processing algorithms used to achieve different image processing effects, such as image noise reduction algorithms, image contrast enhancement algorithms, image de-purple fringing algorithms, dead pixel correction algorithms, automatic white balance algorithms, etc. At the software level, because each image processing algorithm is independently developed, and the algorithm and the algorithm are independent of each other, each image processing algorithm can be considered as a large algorithm module. Based on the reusability of the algorithm module, it can be used in application. , You can conveniently call the algorithm module you want to call according to your needs. For example, if you want to get an image that has been denoised, de-purpled, and contrast-enhanced, you can match the algorithm module corresponding to the image serial input image noise reduction algorithm, the algorithm module corresponding to the image contrast enhancement, and the image de-purplening algorithm. Algorithm module.

However, the applicant found that although each image processing algorithm is independent of each other, each image processing algorithm contains multiple image processing steps. Some steps in different image processing algorithms may be the same, or , The image processing functions corresponding to these steps are the same. These steps with the same or corresponding image processing functions can be referred to as shared steps. Then, when the algorithm modules corresponding to these image processing algorithms are used to process images, each algorithm module independently executes each step in each algorithm. From the point of view of the image processing process, the common steps will be repeated, resulting in a waste of computing resources. Moreover, since each algorithm runs independently, the data obtained by executing the common steps will also be stored independently, which causes a waste of memory resources.

In order to solve the above problems, the embodiments of the present application provide a solution, which can merge the image processing algorithms that need to be used in image processing, and merge the common steps in the image processing algorithms, thereby reducing the number of times the common steps are executed. Realize the saving of computing resources and memory resources.

To integrate the common steps in several image processing algorithms, it is necessary to extract these common steps from the image processing algorithm first, or in other words, to separate these common steps from other steps in the image processing algorithm. However, when image processing algorithms are developed, the entire algorithm is usually developed as a module. Based on the selected framework, the code corresponding to each step in the algorithm is highly coupled and cannot be implemented without affecting the implementation of the entire algorithm. Directly separate the codes corresponding to the common steps.

Based on the foregoing problems, an embodiment of the present application provides an image processing algorithm device. Refer to FIG. 1, which is a schematic structural diagram of an image processing algorithm device provided by an embodiment of the present application.

The device provided by the embodiment of the application can be used to process images by using at least two image processing algorithms, each of which can include at least two algorithm modules, and each algorithm module corresponds to a different image in the image processing algorithm. Processing function.

To facilitate understanding, a simple example can be given. For example, Algorithm A is an image processing algorithm. Algorithm A includes four steps, namely step a1, step a2, step a3, and step a4. In the device provided by the embodiment of this application, the A algorithm can be decomposed into two algorithm modules. The first algorithm module can correspond to an image processing function, which includes steps a1 and a2, and the second algorithm module can correspond to another This image processing function includes step a3 and step a4. For the image processing function corresponding to the algorithm module, it is necessary to pay attention to the difference in the function corresponding to the algorithm as a whole. For example, the A algorithm is an image noise reduction algorithm, and the A algorithm as a whole corresponds to the noise reduction function, which can also be said to be the noise reduction effect. The image processing function corresponding to each algorithm module in the A algorithm can be a lower level function. For example, the first algorithm module can correspond to the frequency separation function, and the second algorithm module can correspond to the filtering function.

The device provided by the embodiment of the present application may include two image processing layers, that is, a first image processing layer and a second image processing layer. Wherein, the first image processing layer may include a common algorithm module. The shared algorithm module can be an algorithm module shared by at least two image processing algorithms, that is, a shared algorithm module can belong to several image processing algorithms at the same time, for example, a shared algorithm module can be an algorithm module in algorithm A, or it can be at the same time An algorithm module in an algorithm. It is not difficult to understand that the image processing function corresponding to the shared algorithm module is shared by at least two image processing algorithms. Corresponding to the above example, if algorithm A includes a step corresponding to frequency separation, and algorithm B also includes a step corresponding to frequency separation, then A shared algorithm module is set, and its corresponding image processing function is frequency separation. The shared algorithm module can be located in the first image processing layer of the device, and it serves as an algorithm module in the A algorithm and the B algorithm at the same time.

As for the algorithm modules other than the common algorithm module of the image processing algorithm, it can be set in the second image processing layer of the device. For example, the above example can be used. The algorithm module corresponding to the frequency separation function in the A algorithm is a common algorithm module and is set in the first image processing layer, and another algorithm module corresponding to the filtering function can be set in the second image processing layer.

The first image processing layer can be used to process the acquired first image using a common algorithm module to obtain the image data set corresponding to the first image, and the second image processing layer can be used to process the target data using other algorithm modules, where The target data is the data required by the other algorithm modules, and the target data can be obtained from the image data set corresponding to the first image.

It should be noted that the image data set can contain the image data of multiple images, or it can only contain the image data of one image, and it can also contain part of the image data of one image. In short, for the image data set, it can contain any image type. The data contained in the data depends on what kind of processing the shared algorithm module has performed on the image, which is not limited in this application.

For the first image, it can also be called a to-be-processed image, which can be acquired in various ways, for example, it can be an image acquired from an external electronic device, or an image acquired by the image sensor of the electronic device of the machine, or It can be an image generated by an electronic device of the machine, and so on. In an implementation, the first image may be an image output by an ISP chip (image signal processing chip), that is, the image may be pre-processed by the ISP chip, and then input to the machine for a series of post-processing.

In the image processing algorithm device provided by the embodiment of the present application, the first image processing layer includes a shared algorithm module, and the shared algorithm module is an algorithm module shared by at least two image processing algorithms, that is, in the device provided by the embodiment of the present application, Common steps in different image processing algorithms can be separated as an algorithm module, and the algorithm module can be used as an algorithm module shared by these image processing algorithms at the same time. Then, in application, executing the shared algorithm module once is equivalent to executing the steps corresponding to the shared algorithm module once in each image processing algorithm. The image data set obtained by executing the shared algorithm module can be provided to each image. The sharing of processing algorithms reduces the number of executions of shared steps and saves computing resources and memory resources on the basis of not affecting the functional realization of various image processing algorithms.

In an optional implementation manner, the shared algorithm module may include a first algorithm module, and the image processing function corresponding to the first algorithm module may be to decompose second images in different frequency bands corresponding to the first image. Correspondingly, based on the processing of the first image by the first algorithm module, the obtained image data set may be a set of image data of multiple second images.

The image contains information of different frequencies. In the process of image processing, different processing can be performed on the information of different frequencies of the image. In this way, better results can be obtained than processing only on the original image. For example, the image noise reduction algorithm can get a better noise reduction effect if the information of different frequencies of the image is processed. Another example is the image contrast enhancement algorithm. If the image information of different frequencies is processed, it can also get better Contrast enhancement effect. Therefore, the above-mentioned first algorithm module can be applied to various image processing algorithms, that is, it can be used as an algorithm module shared by various image processing algorithms.

When implementing the frequency division processing of the first image, there are many feasible implementation manners, for example, Fourier transform can be performed on the image to determine the frequency distribution of the image, and the frequency distribution of the image can also be quantitatively measured by wavelet transform. In the embodiment of the present application, an optional implementation manner is provided, and the second image in different frequency bands corresponding to the first image can be obtained by performing Gaussian pyramid decomposition on the first image.

The process of Gaussian pyramid decomposition can include two steps. The first step is to perform Gaussian smoothing on the image. Specifically, it can be to filter the check image generated by the Gaussian function; the second step is to filter the image after Gaussian smoothing. Perform downsampling. By processing the first image (the image at the bottom of the pyramid) through these two steps, the image of the second layer in the pyramid can be obtained, and then by processing the image of the second layer through these two steps, the image of the third layer in the pyramid can be obtained. image……

Refer to FIG. 2, which is a schematic structural diagram of a Gaussian pyramid provided by an embodiment of the present application. It can be seen that the Gaussian pyramid can include multiple layers, and each layer corresponds to an image, and the images of each layer correspond to different resolutions, scales, and frequency bands. Since the process of Gaussian smoothing can be considered as the process of low-pass filtering, the process of decomposing the image by Gaussian pyramid is also the process of dividing the image. In the Gaussian pyramid, the higher the level of the image, the number of low-pass filtering The more, the lower the frequency in the corresponding frequency band. That is to say, in the Gaussian pyramid, high-level images correspond to low-frequency information, and low-level images correspond to high-frequency information.

The above-mentioned first algorithm module can be applied to various image processing algorithms. As an example, the image processing algorithm may specifically be any of the following: an image noise reduction algorithm, an image de-purplening algorithm, and an image contrast enhancement algorithm.

In one implementation, the image processing algorithm may include an image noise reduction algorithm. The corresponding processing effect of the image noise reduction algorithm is to remove the noise in the image and improve the image signal-to-noise ratio. Image noise reduction algorithms usually run when shooting at high ISO (sensitivity), or in other words, usually process images shot at high ISO (images captured at high ISO are usually noisy). There are multiple image noise reduction algorithms, and the device provided in the embodiment of the present application uses an optional image noise reduction algorithm. The optional image noise reduction algorithm includes multiple algorithm modules, including a first algorithm module (common algorithm module) in the first image processing layer and other algorithm modules in the second image processing layer.

The first algorithm module of the image noise reduction algorithm has been described in the foregoing, and will not be repeated here. Other algorithm modules of the image noise reduction algorithm may include a luminance noise reduction algorithm module and a chrominance noise reduction algorithm module. Among them, the image processing function corresponding to the luminance noise reduction algorithm module may include noise reduction processing on luminance data, and chrominance noise reduction algorithm. The image processing function corresponding to the module may include noise reduction processing on the chrominance data.

The aforementioned image noise reduction algorithm includes noise reduction processing on luminance data and noise reduction processing on chrominance data, where both luminance data and chrominance data are image data. There are many color coding methods for image data, such as RGB, YUV, etc. In the embodiment of the present application, the image data may adopt YUV color coding, where Y corresponds to brightness data and UV corresponds to chromaticity data.

When performing noise reduction processing on brightness data, there are multiple specific implementation manners. The embodiment of the present application provides an alternative. In this optional implementation manner, performing noise reduction processing on brightness data may include the following steps:

Filtering the high-frequency information and low-frequency information of the brightness data of the first image separately;

Combine the filtered high-frequency information with the low-frequency information to obtain the brightness data of the first image after noise reduction.

When noise reduction is performed on the brightness data, the brightness data of the first image can be frequency separated to separate the high-frequency information and low-frequency information of the brightness data of the first image. After that, the high-frequency information and the low-frequency information can be separately Filter to get a better noise reduction effect. Wherein, the frequency separation of the brightness data of the first image can be performed by using the brightness data of the second image corresponding to the original frequency band and the brightness data of the second image corresponding to the designated frequency band.

The second image corresponding to the original frequency band may also be referred to as the second image corresponding to the original scale or original resolution. In the Gaussian pyramid, it corresponds to the bottom layer image in the Gaussian pyramid. The second image corresponding to the original frequency band has not undergone Gaussian smoothing, so there is no loss of high-frequency information. The corresponding frequency band is the most complete and original, that is, the second image corresponding to the original frequency band is actually the first image.

For the second image corresponding to the designated frequency band, the designated frequency band can be any frequency band other than the original frequency band. Because the second image corresponding to the designated frequency band is obtained after at least one Gaussian smoothing process (low-pass filtering) is performed on the first image, Therefore, it corresponds to low-frequency information. In an implementation, the brightness data of the second image corresponding to the designated frequency band can be directly used as the low-frequency information of the brightness data of the first graphic. After the low-frequency information is determined, the high-frequency information can be easily separated.

Specifically, it can correspond to the Gaussian pyramid in Figure 2. The Gaussian pyramid in Figure 2 includes four levels, namely the first layer, the second layer, the third layer, and the fourth layer. Each layer corresponds to a frequency band. In an example, the second image corresponding to the designated frequency band may be the image d4 corresponding to the third layer, that is, the brightness data of the image d4 of the third layer may be selected as the low-frequency information. Of course, the brightness data of the second layer image d2 can also be selected as the low-frequency information, but it needs to be considered that the resolution of the second layer image d2 is four times that of the third layer image d4. Therefore, if the second layer image is selected d2, the calculation amount will be correspondingly increased by four times. Therefore, choosing the image d4 corresponding to the third layer is a choice after balancing the calculation amount and the noise reduction effect.

Further, after the filtered high-frequency information is combined with the low-frequency information, the brightness data of the noise-reduced first image obtained by the combination may also be brightness sharpened, so as to improve the definition of the brightness edge.

When performing noise reduction processing on chrominance data, there may also be multiple implementation manners. The embodiment of the present application provides one of the optional ones. In this optional implementation manner, performing noise reduction processing on chrominance data may include the following steps:

Filtering the chrominance data of the second image corresponding to the low frequency band;

The chrominance data corresponding to the filtered low frequency band is adaptively fused with the chrominance data corresponding to other frequency bands other than the low frequency band to obtain the chrominance data of the first image after noise reduction.

It can be seen from the foregoing that the second images corresponding to different frequency bands are actually obtained by low-pass filtering the first image. Therefore, the more low-pass filtering passes, the lower the frequency band of the corresponding second image. The above-mentioned low-frequency frequency band may be the frequency band whose highest frequency is lower than the preset frequency. Corresponding to the Gaussian pyramid shown in Fig. 2, the second image corresponding to the low-frequency frequency band may be the image corresponding to the highest several layers. The second image corresponding to the fixed low-frequency band is the image corresponding to the highest two layers, and the second image corresponding to the low-frequency band includes the image d4 corresponding to the third layer and the image d8 corresponding to the fourth layer.

The chrominance data of the second image corresponding to the low frequency band can be filtered, and the specific filtering method can be bilateral filtering. After filtering, the chrominance data corresponding to each frequency band can be fused to obtain the chrominance data of the first image after noise reduction. In the above scheme, the chrominance data corresponding to other frequency bands other than the low frequency band may not be filtered. The reason for this is that the noise of the chrominance data mainly exists in the low frequency part. Therefore, filtering the chrominance data corresponding to the low frequency band is only necessary. The noise reduction of chrominance data can be achieved.

In one implementation, when filtering the chrominance data corresponding to the low-frequency frequency band, the luminance data corresponding to the reference frequency band can also be used as a guide to filter the chrominance data corresponding to the low-frequency frequency band. The lower frequency band. The Gaussian pyramid shown in Figure 2 can be used as an example. If the second image corresponding to the low-frequency band includes the image d4 corresponding to the third layer and the image d8 corresponding to the fourth layer, then the chrominance data corresponding to the third layer When filtering d4_uv, you can refer to the brightness data d8_y of the image d8 corresponding to the 4th layer (the image d8 of the 4th layer is obtained by low-pass filtering the image d4 of the 3rd layer, which is higher in frequency than the image d4 of the 3rd layer. One level lower). Similarly, when filtering the chrominance data d8_uv corresponding to the fourth layer, you can refer to the brightness data d16_y of the image corresponding to the fifth layer.

In the above example, the image d8 corresponding to the fourth layer is obtained through the most low-pass filtering, so its corresponding frequency band is the frequency band with the lowest frequency, which can be called the lowest frequency band. It is easy to understand that the lowest frequency band is one of the low frequency frequency bands. When filtering the chrominance data corresponding to the lowest frequency band, it is necessary to refer to the brightness data corresponding to the frequency band that is one level lower in frequency than the lowest frequency band. For example, the chrominance data d8_uv corresponding to the fourth layer in the above example needs to be filtered with reference to the brightness data d16_y corresponding to the fifth layer. Regarding the acquisition method of the brightness data d16_y corresponding to the fifth layer, in the first implementation, when the first image is decomposed by the Gaussian pyramid, the first image can be decomposed into 5 layers of images, that is, the image d0 and the image can be decomposed. d2, image d4, image d8, image d16, and in the second implementation, the brightness data d16_y corresponding to the fifth layer can be obtained by down-sampling the brightness data d8_y corresponding to the fourth layer. The second implementation is less computationally intensive than the first implementation.

When adaptively fusing the chrominance data corresponding to each frequency band, the edge fusion weight can be further referenced. The edge fusion weight may be determined by performing edge detection on the chrominance data corresponding to other frequency bands other than the low frequency frequency band.

The applicant found that in some scenes, the noise reduction effect of chrominance data by the above-mentioned method is still not ideal. Considering that the noise of chrominance data is mainly concentrated in the low frequency part, it is Before the chrominance data is filtered, the chrominance data corresponding to the lowest frequency band can be further de-saturated, thereby improving the noise reduction effect of the chrominance data.

In an implementation, the image processing algorithm may also include an image contrast enhancement algorithm. The corresponding processing effect of the image contrast enhancement algorithm is to improve the contrast of the image. There are also multiple image contrast enhancement algorithms. The device provided in this embodiment of the present application uses an optional image contrast enhancement algorithm. The optional image contrast enhancement algorithm includes multiple algorithm modules, including a first algorithm module in the first image processing layer (the first algorithm module can be shared with the image noise reduction algorithm) and in the second image processing layer Other algorithm modules.

Similarly, the first algorithm module of the image contrast enhancement algorithm has been described in the foregoing, and will not be repeated here. Other algorithm modules of the image contrast enhancement algorithm may include a brightness enhancement algorithm module and a chrominance enhancement algorithm module. Among them, the image processing function corresponding to the brightness enhancement algorithm module may include the contrast enhancement of the brightness data, and the image corresponding to the chrominance noise reduction algorithm module. The processing function may include contrast enhancement of the chrominance data.

When performing contrast enhancement processing on brightness data, there are multiple specific implementation manners. The embodiment of the present application provides one of the options. In this optional implementation manner, performing contrast enhancement processing on brightness data may include the following steps:

Performing contrast enhancement on the high-frequency information and the low-frequency information of the brightness data of the first image respectively;

Combine the contrast-enhanced high-frequency information with the low-frequency information to obtain the brightness data of the contrast-enhanced first image.

For the description of the high-frequency information and low-frequency information of the brightness data of the first image, please refer to the corresponding content of the image noise reduction algorithm in the foregoing, which will not be repeated here.

When performing contrast enhancement on low-frequency information, the low-frequency information can be specifically decomposed by a pyramid to obtain the Gaussian pyramid and Laplacian pyramid corresponding to the low-frequency information, and then enhance the low-frequency information of each layer in the pyramid, and finally use the enhanced Pyramid reconstruction is performed on the low-frequency information of each layer, so as to obtain contrast-enhanced low-frequency information.

The contrast enhancement of the chrominance data is carried out on the basis of the contrast enhancement of the brightness data. Specifically, because the image has been contrast enhanced in the brightness, in order to make the image more coordinated in the visual effect, the corresponding color matching is also required. The degree data is contrast enhanced, and the enhancement process can be based on the amount of change of the brightness data before and after the contrast enhancement.

Specifically, performing contrast enhancement on the chrominance data may include adjusting the chrominance data of the second image corresponding to the original frequency band according to the amount of change in the low-frequency information of the brightness data of the first image before and after the contrast enhancement.

It can be seen from the foregoing description of the low-frequency information that, in some implementations, the low-frequency information may be the brightness data of the second image corresponding to the specified frequency band. According to the amount of change of the brightness data of the second image corresponding to the designated frequency band before and after the contrast enhancement, the chromaticity data of the second image corresponding to the original frequency band (that is, the chromaticity data of the first image) can be adjusted.

More specifically, the chroma mapping ratio can be calculated according to the amount of change in the low-frequency information before and after the contrast enhancement, and the chroma data corresponding to the original frequency band can be remapped according to the calculated chroma mapping ratio to obtain the first image after the contrast enhancement. Chromaticity data.

In an implementation, the image processing algorithm may also include an image de-purplening algorithm. The corresponding processing effect of the image de-purple fringing algorithm is to remove the purple fringing problem generated at the brightness boundary. There are also multiple image de-purple-removing algorithms. The device provided in the embodiment of the present application uses an optional image de-purple-removing algorithm. The optional image de-purplening algorithm includes multiple algorithm modules, including the first algorithm module in the first image processing layer (the first algorithm module can be shared with the image noise reduction algorithm and the image contrast enhancement algorithm) and the Other algorithm modules in the second image processing layer.

Similarly, the first algorithm module of the image de-purple fringing algorithm has been described in the previous section, and will not be repeated here. The image processing function corresponding to other algorithm modules of the image de-purple algorithm may include the following steps:

Determine the color information and location information corresponding to the purple fringe;

According to the determined color information and position information, the chromaticity data of the second image corresponding to the original frequency band is desaturated.

In order to remove the purple fringe in the image, you need to determine the color and position of the purple fringe first. After determining the color and position of the purple fringe, you can desaturate the purple fringe color at the position of the purple fringe in the image to eliminate the image The purple fringe. In the image de-purple fringing algorithm provided by the embodiment of the application, the color information and the position information can be determined according to the image data corresponding to different frequency bands. Specifically, the color information can be determined according to the chromaticity data of the second image corresponding to the first frequency band. , The location information may be determined according to the brightness data of the second image corresponding to the second frequency band. Corresponding to the Gaussian pyramid shown in Figure 2, the first frequency band can correspond to the highest layer in the Gaussian pyramid, that is, the first frequency band can correspond to the fourth layer, and the second frequency band can correspond to the middle layer in the Gaussian pyramid, such as the second frequency band Can correspond to the second layer.

When the position information is specifically determined, since the purple fringing phenomenon exists at the brightness boundary, edge detection can be performed on the brightness data corresponding to the second frequency band to determine the position information of the purple fringing.

After determining the color information and position information, when desaturation is performed according to the color information and the position information, the color information and the position information can be fused to obtain the purple fringing mask; and then according to the purple fringing mask, the color corresponding to the original frequency band The degree data is desaturated. In an implementation, the color information may specifically be color weights, and the position information may specifically be edge weights. The aforementioned purple fringe mask may be obtained by fusing the color weight and the edge weight.

The above is a detailed description of the image noise reduction algorithm, the image contrast enhancement algorithm, and the image depurple removal algorithm provided by the embodiments of the present application. Further analyze the image noise reduction algorithm and image contrast enhancement algorithm. In the image noise reduction algorithm, when the brightness data is denoised, the high-frequency information and low-frequency information of the brightness data of the first image are filtered separately. In the image contrast enhancement algorithm, when the brightness data is contrast-enhanced, the high-frequency information and the low-frequency information of the brightness data of the first image are also contrast-enhanced separately. It can be found that the above two algorithms both include the step of separating the high-frequency information and low-frequency information of the brightness data of the first image. Therefore, this step is a common step of the two algorithms. A common algorithm corresponding to the common step can be set. Module.

Specifically, a second algorithm module may be set in the first image processing layer of the device. The second algorithm module is also a shared algorithm module. The corresponding image processing function may include the brightness data of the second image corresponding to the original frequency band and the designated frequency band. The brightness data of the corresponding second image separates high-frequency information and low-frequency information of the brightness data of the first image. Through this second algorithm module, the steps of separating high frequency information and low frequency information shared by the image noise reduction algorithm and the image contrast enhancement algorithm can be combined, so that the high frequency information and low frequency information of the brightness data of the first image can be separated. The entire image processing process only needs to be executed once, which further saves computing resources.

Further, both the image noise reduction algorithm and the image contrast enhancement algorithm include the step of combining high-frequency information with low-frequency information. Therefore, this step is also a common step. A third algorithm module can also be set in the first image processing layer. The image processing function corresponding to the algorithm module includes combining the high-frequency information with the low-frequency information.

The foregoing is a detailed description of the image processing algorithm device provided by the embodiment of the present application. The image processing algorithm device provided by the embodiment of the present application divides the image processing algorithm into smaller granular modules, so that the modules corresponding to the same image processing function can be merged, and the image processing algorithms can be merged. Specifically, in the device provided by the embodiment of the present application, the first image processing layer includes a shared algorithm module, and the shared algorithm module is an algorithm module shared by at least two image processing algorithms, that is, in the device provided by the embodiment of the present application , The common steps in different image processing algorithms can be separated as an algorithm module, and the algorithm module can be used as the algorithm module shared by these image processing algorithms at the same time. Then, in application, executing the shared algorithm module once is equivalent to executing the steps corresponding to the shared algorithm module once in each image processing algorithm. The image data set obtained by executing the shared algorithm module can be provided to each image. The sharing of processing algorithms reduces the number of executions of shared steps and saves computing resources and memory resources on the basis of not affecting the functional realization of various image processing algorithms.

Refer to FIG. 3 below, which is a flowchart of a fusion algorithm provided by an embodiment of the present application. It should be noted that this fusion algorithm is only used as an example. It combines the device provided in the embodiment of the present application, and integrates the image noise reduction algorithm, the image contrast enhancement algorithm, and the image depurple removal algorithm.

In Figure 3, rounded boxes can represent data, and rectangular boxes can represent modules. The image data yuv_in of the first image can be input to the first algorithm module. Specifically, the first algorithm module can be Gaussian Pyramid Decomposition, which can be set in the first image processing layer of the device, and is a common algorithm module for image denoising algorithm, image de-purplening algorithm, and image contrast enhancement algorithm. . After the first image is processed by the first algorithm module, the first image can be decomposed into second images of multiple frequency bands. Specifically, as shown in Figure 3, the Gaussian pyramid obtained after decomposition corresponds to the Gaussian pyramid in Figure 2. An image is decomposed into second images corresponding to four different frequency bands, d0, d2, d4, and d8.

Correspondingly, the image data set obtained after decomposition includes image data of each second image, corresponding to FIG. 3, the image data set includes image data of four second images d0, d2, d4, and d8, and these image data specifically include The luminance data d0_y and chrominance data d0_uv of the second image d0, the luminance data d2_y and chrominance data d2_uv of the second image d2, the luminance data d4_y and chrominance data d4_uv of the second image d4, and the luminance data d8_y of the second image d8 And chromaticity data d8_uv.

Figure 3 can be divided into two parts: luminance data processing and chrominance data processing. Among them, in the processing of the brightness data, d0_y is the brightness data corresponding to the original frequency band, and the brightness data d4_y corresponding to the designated frequency band is input to the second algorithm module. The second algorithm module corresponds to the specific GENRATE_HF module in Figure 3, and its corresponding image processing function is to separate the first image from the brightness data d0_y of the second image corresponding to the original frequency band and the brightness data d4_y of the second image corresponding to the specified frequency band. The high-frequency information hf_src and the low-frequency information d4_y of the brightness data (d4_y itself can be used as the low-frequency information). The second algorithm module is also a shared algorithm module in the first image processing layer, which is specifically a shared algorithm module of the image noise reduction algorithm and the image contrast enhancement algorithm.

The HF_FILTER module and the LF_FILTER module are the high frequency filter module and the low frequency filter module respectively. The HF_FILTER module is used to filter high frequency information, and the LF_FILTER module is used to filter low frequency information. These two algorithm modules are algorithms in the image noise reduction algorithm. Module.

The HF_GAIN module and the LF_GAIN module are the high-frequency enhancement module and the low-frequency enhancement module respectively. The HF_GAIN module is used to enhance the contrast of high-frequency information, and the LF_GAIN module is used to enhance the contrast of the low-frequency information. These two algorithm modules are in the image contrast enhancement algorithm. Algorithm module.

The UP_COMBINE module corresponds to the third algorithm module, which is used to combine high-frequency information and low-frequency information, corresponding to Figure 3, which is specifically used to perform filtering and contrast-enhanced high-frequency information hf and low-frequency information lf_after Up-sampling is combined to obtain the brightness data d0_combine. The steps corresponding to this module are shared by the image noise reduction algorithm and image contrast enhancement. Therefore, this module can also be used as a shared algorithm module and set in the first image processing layer.

The LUMA_SHARPEN module is used to sharpen the brightness data, which is an algorithm module in the image noise reduction algorithm.

For the processing of chrominance data, the chroma denoising algorithm module (Chroma Denoising, CDNS), the image de-purple removal module (Purple Fringe Remove, PFR) and the chroma enhancement algorithm module (UV_REMAPPING) can be connected in series in sequence. Realize the fusion of the three algorithms of image noise reduction algorithm, image contrast enhancement algorithm and image de-purple algorithm in the processing of chrominance data.

For each algorithm module in FIG. 3, the corresponding specific steps and specific implementation manners can refer to the corresponding descriptions already in the previous text, which will not be repeated here.

It should be noted that the algorithm module shown above is only an example. When the algorithm module is specifically divided, a smaller module granularity can be used. For example, some algorithm modules can be further subdivided into sub-modules. The module may also be an algorithm module referred to in the embodiment of the present application.

Please refer to FIG. 4 below, which is a flowchart of an image processing method provided by an embodiment of the present application. The method is used to process images by using at least two image processing algorithms, each of the image processing algorithms includes at least two algorithm modules, and each of the algorithm modules corresponds to a different image processing function in the image processing algorithm; The methods include:

S401: Use a common algorithm module to process the acquired first image to obtain an image data set corresponding to the first image.

Wherein, the common algorithm module is an algorithm module corresponding to the same image processing function in at least two of the image processing algorithms;

S402: Obtain target data corresponding to other algorithm modules of each of the image processing algorithms from the image data set, and use the other algorithm modules to process the corresponding target data.

Optionally, the shared algorithm module includes a first algorithm module, and the image processing function corresponding to the first algorithm module includes decomposing second images of different frequency bands corresponding to the first image; the image data set includes multiple Image data of the second image.

Optionally, the second images of different frequency bands corresponding to the first image are obtained by performing Gaussian pyramid decomposition on the first image.

Optionally, the image data includes luminance data and chrominance data.

Optionally, the image processing algorithm includes any one of the following: an image noise reduction algorithm, an image de-purplening algorithm, and an image contrast enhancement algorithm.

Optionally, the other algorithm modules of the image noise reduction algorithm include a luminance noise reduction algorithm module and a chrominance noise reduction algorithm module, and the image processing function corresponding to the luminance noise reduction algorithm module includes noise reduction processing on luminance data The image processing function corresponding to the chrominance noise reduction algorithm module includes performing noise reduction processing on the chrominance data.

Optionally, the performing noise reduction processing on the brightness data includes:

Filtering high-frequency information and low-frequency information of the brightness data of the first image separately;

Combining the filtered high frequency information and the low frequency information to obtain the brightness data of the first image after noise reduction.

Optionally, after the combining the filtered high-frequency information with the low-frequency information, the method further includes:

Brightness sharpening is performed on the combined brightness data of the noise-reduced first image.

Optionally, the performing noise reduction processing on the chrominance data includes:

Filtering the chromaticity data of the second image corresponding to the low frequency frequency band; wherein the low frequency frequency band is a frequency band in which the highest frequency in the frequency band is lower than a preset frequency;

Adaptively fusing the filtered chrominance data corresponding to the low-frequency frequency band and the chrominance data corresponding to other frequency bands other than the low-frequency frequency band to obtain the chrominance data of the first image after noise reduction.

Optionally, the filtering of the chrominance data corresponding to the low-frequency frequency band is performed according to the brightness data corresponding to a reference frequency band; the reference frequency band is a frequency band whose frequency is one level lower than the low-frequency frequency band.

Optionally, the brightness data of the reference frequency band corresponding to the lowest frequency band is obtained by down-sampling the brightness data corresponding to the lowest frequency band.

Optionally, the adaptive fusion is performed according to edge fusion weights, and the edge fusion weights are determined by performing edge detection on the chrominance data corresponding to the other frequency bands.

Optionally, before filtering the chrominance data corresponding to the lowest frequency band, it also includes:

Desaturate the chrominance data corresponding to the lowest frequency band.

Optionally, the other algorithm modules of the image contrast enhancement algorithm include a brightness enhancement algorithm module and a chrominance enhancement algorithm module, and the image processing function corresponding to the brightness enhancement algorithm module includes contrast enhancement of brightness data, and the color The image processing function corresponding to the degree enhancement algorithm module includes the contrast enhancement of the chrominance data.

Optionally, the performing contrast enhancement on the brightness data includes:

Combining the contrast-enhanced high-frequency information and the low-frequency information to obtain the brightness data of the first image after the contrast-enhancement.

Optionally, the high-frequency information and low-frequency information of the brightness data of the first image are obtained from the brightness data of the second image corresponding to an original frequency band and the brightness data of the second image corresponding to a designated frequency band.

Optionally, the performing contrast enhancement on the chrominance data includes:

The chromaticity data of the second image corresponding to the original frequency band is adjusted according to the amount of change of the low-frequency information before and after the contrast enhancement.

Optionally, the adjusting the chromaticity data of the second image corresponding to the original frequency band according to the amount of change of the low-frequency information before and after the contrast enhancement includes:

Calculating the chromaticity mapping ratio according to the amount of change in the low-frequency information before and after the contrast enhancement;

Remapping the chrominance data corresponding to the original frequency band according to the chrominance mapping ratio to obtain the chrominance data of the first image after contrast enhancement.

Optionally, the image processing functions corresponding to the other algorithm modules of the image de-purplening algorithm include:

According to the color information and the position information, the chroma data of the second image corresponding to the original frequency band is desaturated.

Optionally, the color information is determined according to the chromaticity data of the second image corresponding to the first frequency band; the position information is determined according to the brightness data of the second image corresponding to the second frequency band.

Optionally, the location information is determined by performing edge detection on the brightness data corresponding to the second frequency band.

Optionally, desaturating the chroma data of the second image corresponding to the original frequency band according to the color information and the position information includes:

Fusing the color information and the position information to obtain a purple fringing mask;

Desaturate the chrominance data corresponding to the original frequency band according to the purple fringing mask.

Optionally, the shared algorithm module further includes a second algorithm module, and the image processing function corresponding to the second algorithm module includes the second image processing function corresponding to the specified frequency band through the brightness data of the second image corresponding to the original frequency band. The brightness data of the image separates high-frequency information and low-frequency information of the brightness data of the first image.

Optionally, the shared algorithm module further includes a third algorithm module, and the image processing function corresponding to the third algorithm module includes combining the high-frequency information with the low-frequency information.

Optionally, the first image is an image output by an ISP chip.

For the specific implementation of any image processing method provided above, refer to the relevant description of the image processing algorithm device provided in the embodiment of the present application, which will not be repeated here.

Please refer to FIG. 5 below. FIG. 5 is a schematic structural diagram of a camera provided by an embodiment of the present application. The camera includes: a body 501, a lens 502 arranged on the body 501, an image sensor 503, an ISP chip 504 and a DSP chip 505 arranged in the body 501;

The image sensor 503 is used to collect original images through the lens 502;

The ISP chip 504 is used to obtain the original image from the image sensor 503, and process the original image to obtain a first image;

The DSP chip 505 is configured to obtain a first image from the ISP chip 504, and process the first image by using a common algorithm module to obtain an image data set corresponding to the first image; wherein, the common algorithm module Are algorithm modules corresponding to the same image processing function in at least two of the image processing algorithms; respectively obtain target data corresponding to other algorithm modules of each of the image processing algorithms from the image data set, and use the Other algorithm modules process the corresponding target data.

Optionally, the image data includes luminance data and chrominance data.

Optionally, the DSP chip is further configured to filter the high frequency information and low frequency information of the brightness data of the first image separately; combine the filtered high frequency information with the low frequency information to obtain a reduction Luminance data of the first image after noise.

Optionally, the DSP chip is further configured to, after combining the filtered high-frequency information with the low-frequency information, compare the combined brightness data of the noise-reduced first image Perform brightness sharpening.

Optionally, the DSP chip is also used to filter the chrominance data of the second image corresponding to the low frequency frequency band; wherein the low frequency frequency band is a frequency band in which the highest frequency is lower than a preset frequency; The filtered chromaticity data corresponding to the low-frequency frequency band is adaptively fused with chromaticity data corresponding to other frequency bands other than the low-frequency frequency band to obtain the chromaticity data of the first image after noise reduction.

Optionally, the DSP chip is also used to desaturate the chromaticity data corresponding to the lowest frequency band before filtering the chromaticity data corresponding to the lowest frequency band.

Optionally, the DSP chip is further configured to perform contrast enhancement on the high-frequency information and low-frequency information of the brightness data of the first image respectively; combine the high-frequency information with the contrast-enhanced high-frequency information and the low-frequency information, Obtain brightness data of the first image after contrast enhancement.

Optionally, the DSP chip is further configured to adjust the chromaticity data of the second image corresponding to the original frequency band according to the amount of change in the low-frequency information before and after the contrast enhancement.

Optionally, the DSP chip is further configured to calculate a chrominance mapping ratio according to the amount of change in the low-frequency information before and after the contrast enhancement; and to remap the chrominance data corresponding to the original frequency band according to the chrominance mapping ratio , To obtain the chromaticity data of the first image after contrast enhancement.

Optionally, the DSP chip is further configured to determine the color information and position information corresponding to the purple fringing; according to the color information and the position information, reduce the chromaticity data of the second image corresponding to the original frequency band. saturation.

Optionally, the DSP chip is further configured to fuse the color information and the position information to obtain a purple fringing mask; and desaturate the chromaticity data corresponding to the original frequency band according to the purple fringing mask.

For the specific implementation of any camera provided above, refer to the related description of the image processing algorithm device provided in the embodiment of the present application, which will not be repeated here.

The embodiments of the present application also provide a computer-readable storage medium having a computer program stored on the computer-readable storage medium, and when the computer program is executed by a processor, any one of the image processing provided in the embodiments of the present application is implemented. method.

As long as there is no conflict or contradiction between the technical features provided in the above embodiments, those skilled in the art can combine the various technical features according to actual conditions to form various different embodiments. However, the document of this application is limited in length and does not describe various embodiments. However, it is understandable that various embodiments also belong to the scope of the disclosure of the embodiments of this application.

The embodiments of the present application may adopt the form of a computer program product implemented on one or more storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing program codes. Computer usable storage media include permanent and non-permanent, removable and non-removable media, and information storage can be realized by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to: phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, CD-ROM, digital versatile disc (DVD) or other optical storage, Magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media can be used to store information that can be accessed by computing devices.

It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply one of these entities or operations. There is any such actual relationship or order between. The terms "include", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements that are not explicitly listed. Elements, or also include elements inherent to such processes, methods, articles, or equipment. If there are no more restrictions, the element defined by the sentence "including a..." does not exclude the existence of other identical elements in the process, method, article, or equipment that includes the element.

The methods and devices provided in the embodiments of the application are described in detail above. Specific examples are used in this article to illustrate the principles and implementations of the application. The descriptions of the above embodiments are only used to help understand the methods and methods of the application. Core idea; At the same time, for ordinary technicians in the field, according to the idea of this application, there will be changes in the specific implementation and scope of application. In summary, the content of this specification should not be construed as a limitation to this application .

Claims

An image processing algorithm device, characterized in that it is used to process images by using at least two image processing algorithms, each of the image processing algorithms includes at least two algorithm modules, and each of the algorithm modules corresponds to the image processing Different image processing functions in the algorithm; the device includes a first image processing layer and a second image processing layer, the first image processing layer includes a common algorithm module, and the common algorithm module is at least two of the image processing algorithms Algorithm modules corresponding to the same image processing function in the image processing; the second image processing layer includes various other algorithm modules of the image processing algorithm;

The first image processing layer is configured to use the common algorithm module to process the acquired first image to obtain an image data set corresponding to the first image;

The second image processing layer is configured to use the other algorithm module to process the corresponding target data obtained from the image data set.
The image processing algorithm device according to claim 1, wherein the shared algorithm module includes a first algorithm module, and the image processing function corresponding to the first algorithm module includes decomposing different frequency bands corresponding to the first image The second image; the image data set includes a plurality of image data of the second image.
3. The image processing algorithm device according to claim 2, wherein the second image in different frequency bands corresponding to the first image is obtained by performing Gaussian pyramid decomposition on the first image.
The image processing algorithm device according to claim 2, wherein the image data includes luminance data and chrominance data.
The image processing algorithm device according to claim 4, wherein the image processing algorithm comprises any one of the following: an image noise reduction algorithm, an image de-purple fringing algorithm, and an image contrast enhancement algorithm.
The image processing algorithm device according to claim 5, wherein the other algorithm modules of the image noise reduction algorithm include a luminance noise reduction algorithm module and a chrominance noise reduction algorithm module, and the luminance noise reduction algorithm module corresponds to The image processing function includes performing noise reduction processing on luminance data, and the image processing function corresponding to the chrominance noise reduction algorithm module includes performing noise reduction processing on chrominance data.
The image processing algorithm device according to claim 6, wherein said performing noise reduction processing on brightness data comprises:

Filtering high-frequency information and low-frequency information of the brightness data of the first image separately;

Combining the filtered high frequency information and the low frequency information to obtain the brightness data of the first image after noise reduction.
8. The image processing algorithm device according to claim 7, wherein after said combining the filtered high frequency information with the low frequency information, the method further comprises:

Brightness sharpening is performed on the combined brightness data of the noise-reduced first image.
The image processing algorithm device according to claim 6, wherein said performing noise reduction processing on chrominance data comprises:

Filtering the chromaticity data of the second image corresponding to the low frequency frequency band; wherein the low frequency frequency band is a frequency band in which the highest frequency in the frequency band is lower than a preset frequency;

Adaptively fusing the filtered chrominance data corresponding to the low-frequency frequency band and the chrominance data corresponding to other frequency bands other than the low-frequency frequency band to obtain the chrominance data of the first image after noise reduction.
The image processing algorithm device according to claim 9, wherein the filtering of the chrominance data corresponding to the low frequency band is performed according to the brightness data corresponding to the reference frequency band; The frequency band one level lower than the frequency band.
The image processing algorithm device according to claim 10, wherein the brightness data of the reference frequency band corresponding to the lowest frequency band is obtained by down-sampling the brightness data corresponding to the lowest frequency band.
The image processing algorithm device according to claim 9, wherein the adaptive fusion is performed according to edge fusion weights, and the edge fusion weights are determined by edge detection of chrominance data corresponding to the other frequency bands. of.
The image processing algorithm device according to claim 9, wherein before filtering the chrominance data corresponding to the lowest frequency band, the method further comprises:

Desaturate the chrominance data corresponding to the lowest frequency band.
The image processing algorithm device according to claim 5, wherein the other algorithm modules of the image contrast enhancement algorithm include a brightness enhancement algorithm module and a chrominance enhancement algorithm module, and image processing corresponding to the brightness enhancement algorithm module The function includes the contrast enhancement of the luminance data, and the image processing function corresponding to the chrominance enhancement algorithm module includes the contrast enhancement of the chrominance data.
The image processing algorithm device according to claim 14, wherein said performing contrast enhancement on brightness data comprises:

Performing contrast enhancement on the high-frequency information and the low-frequency information of the brightness data of the first image respectively;

Combining the contrast-enhanced high-frequency information and the low-frequency information to obtain the brightness data of the first image after the contrast-enhancement.
The image processing algorithm device according to claim 7 or 15, wherein the high-frequency information and low-frequency information of the brightness data of the first image pass through the brightness data of the second image and the designated frequency band corresponding to the original frequency band. Corresponding to the brightness data of the second image.
The image processing algorithm device according to claim 15, wherein said performing contrast enhancement on chrominance data comprises:

The chromaticity data of the second image corresponding to the original frequency band is adjusted according to the amount of change of the low-frequency information before and after the contrast enhancement.
The image processing algorithm device according to claim 17, wherein the adjusting the chromaticity data of the second image corresponding to the original frequency band according to the amount of change of the low-frequency information before and after the contrast enhancement comprises:

Calculating the chromaticity mapping ratio according to the amount of change in the low-frequency information before and after the contrast enhancement;

Remapping the chrominance data corresponding to the original frequency band according to the chrominance mapping ratio to obtain the chrominance data of the first image after contrast enhancement.
The image processing algorithm device according to claim 5, wherein the image processing functions corresponding to the other algorithm modules of the image de-purplening algorithm include:

Determine the color information and location information corresponding to the purple fringe;

According to the color information and the position information, the chroma data of the second image corresponding to the original frequency band is desaturated.
The image processing algorithm device according to claim 19, wherein the color information is determined according to the chromaticity data of the second image corresponding to the first frequency band; and the position information is determined according to the second frequency band The brightness data of the second image is determined.
22. The image processing algorithm device of claim 20, wherein the position information is determined by performing edge detection on the brightness data corresponding to the second frequency band.
The image processing algorithm device according to claim 19, wherein, according to the color information and the position information, de-saturating the chroma data of the second image corresponding to the original frequency band comprises:

Fusing the color information and the position information to obtain a purple fringing mask;

Desaturate the chrominance data corresponding to the original frequency band according to the purple fringing mask.
The image processing algorithm device according to claim 2, wherein the shared algorithm module further comprises a second algorithm module, and the image processing function corresponding to the second algorithm module includes the second image corresponding to the original frequency band. The brightness data of and the brightness data of the second image corresponding to the designated frequency band are separated into high-frequency information and low-frequency information of the brightness data of the first image.
The image processing algorithm device according to claim 23, wherein the shared algorithm module further comprises a third algorithm module, and the image processing function corresponding to the third algorithm module includes combining the high-frequency information with the low-frequency information. Information integration.
The image processing algorithm device according to claim 1, wherein the first image is an image output by an ISP chip.
An image processing method, characterized in that it is used to process images by using at least two image processing algorithms, each of the image processing algorithms includes at least two algorithm modules, and each of the algorithm modules corresponds to the image processing algorithm The different image processing functions in the; the method includes:

The shared algorithm module is used to process the acquired first image to obtain the image data set corresponding to the first image; wherein the shared algorithm module is at least two of the image processing algorithms corresponding to the same image processing function Module

Obtain target data corresponding to other algorithm modules of each of the image processing algorithms from the image data set, and use the other algorithm modules to process the corresponding target data.
The image processing method according to claim 26, wherein the shared algorithm module includes a first algorithm module, and the image processing function corresponding to the first algorithm module includes decomposing different frequency bands corresponding to the first image. A second image; the image data set includes image data of a plurality of the second images.
28. The image processing method according to claim 27, wherein the second images of different frequency bands corresponding to the first image are obtained by performing Gaussian pyramid decomposition on the first image.
The image processing method according to claim 27, wherein the image data includes luminance data and chrominance data.
The image processing method according to claim 29, wherein the image processing algorithm comprises any one of the following: an image noise reduction algorithm, an image de-purple fringing algorithm, and an image contrast enhancement algorithm.
The image processing method according to claim 30, wherein the other algorithm modules of the image noise reduction algorithm include a luminance noise reduction algorithm module and a chrominance noise reduction algorithm module, and the luminance noise reduction algorithm module corresponds to The image processing function includes performing noise reduction processing on luminance data, and the image processing function corresponding to the chrominance noise reduction algorithm module includes performing noise reduction processing on chrominance data.
The image processing method according to claim 31, wherein said performing noise reduction processing on brightness data comprises:

Filtering high-frequency information and low-frequency information of the brightness data of the first image separately;

Combining the filtered high frequency information and the low frequency information to obtain the brightness data of the first image after noise reduction.
The image processing method according to claim 32, wherein after said combining the filtered high frequency information with the low frequency information, the method further comprises:

Brightness sharpening is performed on the combined brightness data of the noise-reduced first image.
The image processing method according to claim 31, wherein said performing noise reduction processing on chrominance data comprises:

Filtering the chromaticity data of the second image corresponding to the low frequency frequency band; wherein the low frequency frequency band is a frequency band in which the highest frequency in the frequency band is lower than a preset frequency;

Adaptively fusing the filtered chrominance data corresponding to the low-frequency frequency band and the chrominance data corresponding to other frequency bands other than the low-frequency frequency band to obtain the chrominance data of the first image after noise reduction.
The image processing method according to claim 34, wherein the filtering of the chrominance data corresponding to the low-frequency frequency band is performed according to the brightness data corresponding to the reference frequency band; The lower frequency band.
The image processing method according to claim 35, wherein the brightness data of the reference frequency band corresponding to the lowest frequency band is obtained by down-sampling the brightness data corresponding to the lowest frequency band.
The image processing method according to claim 34, wherein the adaptive fusion is performed according to edge fusion weights, and the edge fusion weights are determined by performing edge detection on the chrominance data corresponding to the other frequency bands. .
The image processing method according to claim 34, wherein before filtering the chrominance data corresponding to the lowest frequency band, the method further comprises:

Desaturate the chrominance data corresponding to the lowest frequency band.
The image processing method according to claim 30, wherein the other algorithm modules of the image contrast enhancement algorithm include a brightness enhancement algorithm module and a chrominance enhancement algorithm module, and image processing functions corresponding to the brightness enhancement algorithm module It includes performing contrast enhancement on brightness data, and the image processing function corresponding to the chrominance enhancement algorithm module includes performing contrast enhancement on chrominance data.
The image processing method according to claim 39, wherein said performing contrast enhancement on brightness data comprises:

Performing contrast enhancement on the high-frequency information and the low-frequency information of the brightness data of the first image respectively;

Combining the contrast-enhanced high-frequency information and the low-frequency information to obtain the brightness data of the first image after the contrast-enhancement.
The image processing method according to claim 32 or 40, wherein the high-frequency information and low-frequency information of the brightness data of the first image correspond to the specified frequency band by the brightness data of the second image corresponding to the original frequency band. Is obtained from the brightness data of the second image.
The image processing method according to claim 40, wherein said performing contrast enhancement on chrominance data comprises:

The chromaticity data of the second image corresponding to the original frequency band is adjusted according to the amount of change of the low-frequency information before and after the contrast enhancement.
The image processing method according to claim 42, wherein the adjusting the chrominance data of the second image corresponding to the original frequency band according to the amount of change of the low-frequency information before and after the contrast enhancement comprises:

Calculating the chromaticity mapping ratio according to the amount of change in the low-frequency information before and after the contrast enhancement;

Remapping the chrominance data corresponding to the original frequency band according to the chrominance mapping ratio to obtain the chrominance data of the first image after contrast enhancement.
The image processing method according to claim 30, wherein the image processing functions corresponding to the other algorithm modules of the image de-purplening algorithm comprise:

Determine the color information and location information corresponding to the purple fringe;

According to the color information and the position information, the chroma data of the second image corresponding to the original frequency band is desaturated.
The image processing method according to claim 44, wherein the color information is determined according to the chrominance data of the second image corresponding to the first frequency band; and the position information is determined according to the all corresponding to the second frequency band. The brightness data of the second image is determined.
The image processing method according to claim 45, wherein the position information is determined by performing edge detection on the brightness data corresponding to the second frequency band.
The image processing method according to claim 44, wherein, according to the color information and the position information, de-saturating the chrominance data of the second image corresponding to the original frequency band comprises:

Fusing the color information and the position information to obtain a purple fringing mask;

Desaturate the chrominance data corresponding to the original frequency band according to the purple fringing mask.
The image processing method according to claim 27, wherein the shared algorithm module further comprises a second algorithm module, and the image processing function corresponding to the second algorithm module includes the image processing function of the second image corresponding to the original frequency band. The brightness data and the brightness data of the second image corresponding to the designated frequency band separate high-frequency information and low-frequency information of the brightness data of the first image.
The image processing method according to claim 48, wherein the shared algorithm module further comprises a third algorithm module, and the image processing function corresponding to the third algorithm module includes combining the high-frequency information with the low-frequency information Combine.
The image processing method according to claim 26, wherein the first image is an image output by an ISP chip.
A camera, characterized by comprising: a body, a lens arranged on the body, an image sensor, an ISP chip, and a DSP chip arranged in the body;

The image sensor is used to collect an original image through the lens;

The ISP chip is used to obtain the original image from the image sensor, and process the original image to obtain a first image;

The DSP chip is used to obtain a first image from the ISP chip, and use a common algorithm module to process the first image to obtain an image data set corresponding to the first image; wherein, the common algorithm module is at least Algorithm modules corresponding to the same image processing function in the two image processing algorithms; respectively obtain target data corresponding to other algorithm modules of each of the image processing algorithms from the image data set, and use the other algorithms The module processes the corresponding target data.
The camera according to claim 51, wherein the shared algorithm module comprises a first algorithm module, and the image processing function corresponding to the first algorithm module comprises decomposing the second image in different frequency bands corresponding to the first image. Image; the image data set includes a plurality of image data of the second image.
The camera of claim 52, wherein the second images of different frequency bands corresponding to the first image are obtained by performing Gaussian pyramid decomposition on the first image.
The camera of claim 52, wherein the image data includes luminance data and chrominance data.
The camera according to claim 54, wherein the image processing algorithm comprises any one of the following: an image noise reduction algorithm, an image de-purplening algorithm, and an image contrast enhancement algorithm.
The camera according to claim 55, wherein the other algorithm modules of the image noise reduction algorithm include a luminance noise reduction algorithm module and a chrominance noise reduction algorithm module, and image processing corresponding to the luminance noise reduction algorithm module The function includes performing noise reduction processing on luminance data, and the image processing function corresponding to the chrominance noise reduction algorithm module includes performing noise reduction processing on chrominance data.
The camera according to claim 56, wherein the DSP chip is specifically used to filter high-frequency information and low-frequency information of the brightness data of the first image when the brightness noise reduction algorithm module is invoked. ; Combine the filtered high-frequency information with the low-frequency information to obtain the luminance data of the first image after noise reduction.
The camera according to claim 57, wherein the DSP chip is further configured to: after the filtering the high frequency information and the low frequency information are combined, the combined noise reduction Brightness sharpening is performed on the brightness data of the first image.
The camera according to claim 56, wherein the DSP chip is specifically used to filter the chromaticity data of the second image corresponding to the low frequency band when the chromaticity noise reduction algorithm module is invoked; wherein The low-frequency frequency band is a frequency band whose highest frequency is lower than a preset frequency; the filtered chromaticity data corresponding to the low-frequency frequency band and the chromaticity data corresponding to other frequency bands other than the low-frequency frequency band are adaptively fused To obtain the chromaticity data of the first image after noise reduction.
The camera of claim 59, wherein the filtering of the chrominance data corresponding to the low frequency band is performed according to the brightness data corresponding to the reference frequency band; the reference frequency band has a frequency lower than the low frequency band by one. Grade frequency band.
The camera of claim 60, wherein the brightness data of the reference frequency band corresponding to the lowest frequency band is obtained by down-sampling the brightness data corresponding to the lowest frequency band.
The camera of claim 59, wherein the adaptive fusion is performed according to edge fusion weights, and the edge fusion weights are determined by performing edge detection on chrominance data corresponding to the other frequency bands.
The camera of claim 59, wherein the DSP chip is further configured to desaturate the chrominance data corresponding to the lowest frequency band before filtering the chrominance data corresponding to the lowest frequency band.
The camera of claim 55, wherein the other algorithm modules of the image contrast enhancement algorithm include a brightness enhancement algorithm module and a chrominance enhancement algorithm module, and the image processing function corresponding to the brightness enhancement algorithm module includes The brightness data is contrast enhanced, and the image processing function corresponding to the chrominance enhancement algorithm module includes the contrast enhancement of the chrominance data.
The camera according to claim 64, wherein the DSP chip is specifically configured to perform contrast enhancement on high-frequency information and low-frequency information of the brightness data of the first image when the brightness enhancement algorithm module is invoked. ; Combining the contrast-enhanced high-frequency information with the low-frequency information to obtain the contrast-enhanced brightness data of the first image.
The camera according to claim 57 or 65, wherein the high-frequency information and low-frequency information of the brightness data of the first image are obtained from the brightness data of the second image corresponding to the original frequency band and the specified frequency band. The brightness data of the second image is obtained.
The camera according to claim 65, wherein when the DSP chip calls the chrominance enhancement algorithm module, it is specifically configured to determine the amount of change in the low-frequency information before and after the contrast enhancement to the original frequency band corresponding to the The chromaticity data of the second image is adjusted.
The camera of claim 67, wherein the DSP chip is further configured to calculate a chromaticity mapping ratio according to the amount of change of the low-frequency information before and after contrast enhancement; The chromaticity data corresponding to the frequency band is remapped to obtain the chromaticity data of the first image after contrast enhancement.
The camera of claim 55, wherein the DSP chip is specifically used to determine the color information and position information corresponding to the purple fringing when calling the other algorithm modules of the image de-purple fringing algorithm; The color information and the position information desaturate the chromaticity data of the second image corresponding to the original frequency band.
The camera according to claim 69, wherein the color information is determined according to the chromaticity data of the second image corresponding to the first frequency band; and the position information is determined according to the first image corresponding to the second frequency band. The brightness data of the second image is determined.
The camera of claim 70, wherein the position information is determined by performing edge detection on the brightness data corresponding to the second frequency band.
The camera according to claim 69, wherein the DSP chip is further configured to fuse the color information and the position information to obtain a purple fringing mask; according to the purple fringing mask, the original The chrominance data corresponding to the frequency band is desaturated.
The camera according to claim 52, wherein the shared algorithm module further comprises a second algorithm module, and the image processing function corresponding to the second algorithm module includes brightness data of the second image corresponding to the original frequency band. The brightness data of the second image corresponding to the designated frequency band separates high-frequency information and low-frequency information of the brightness data of the first image.
The camera according to claim 73, wherein the shared algorithm module further comprises a third algorithm module, and the image processing function corresponding to the third algorithm module includes combining the high frequency information with the low frequency information.
A computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the image processing method according to any one of claims 26 to 50 is realized .