WO2024031579A1 - 图像降噪方法、装置及芯片 - Google Patents

图像降噪方法、装置及芯片 Download PDF

Info

Publication number
WO2024031579A1
WO2024031579A1 PCT/CN2022/111921 CN2022111921W WO2024031579A1 WO 2024031579 A1 WO2024031579 A1 WO 2024031579A1 CN 2022111921 W CN2022111921 W CN 2022111921W WO 2024031579 A1 WO2024031579 A1 WO 2024031579A1
Authority
WO
WIPO (PCT)
Prior art keywords
image block
image
denoised
color channel
information
Prior art date
Application number
PCT/CN2022/111921
Other languages
English (en)
French (fr)
Inventor
尹佳琪
马昊辰
Original Assignee
上海玄戒技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 上海玄戒技术有限公司 filed Critical 上海玄戒技术有限公司
Priority to PCT/CN2022/111921 priority Critical patent/WO2024031579A1/zh
Priority to CN202280004092.3A priority patent/CN116391202B/zh
Publication of WO2024031579A1 publication Critical patent/WO2024031579A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration

Definitions

  • the present application relates to the field of image processing technology, and in particular to an image noise reduction method, device and chip.
  • Image sensors with Color Filter Array are commonly used in digital cameras to obtain color scene images. Since each pixel can only record one color component, the interpolation method needs to be used to complete the missing color at the corresponding pixel position, and the digital image is expanded from a single-channel grayscale image to a three-channel color image, that is, demosaic.
  • noise will inevitably be introduced during the acquisition process of digital images. If demosaicing is performed directly, structural noise will be introduced and noise from different color channels will be mixed. It is difficult for subsequent image processing methods to deal with these problems, which is serious. affects the quality of the image.
  • this application provides an image noise reduction method, device and chip.
  • the main purpose is to improve the technical problem that the current traditional image noise reduction method cannot be directly applied to CFA images, thereby affecting the noise reduction effect of CFA images. .
  • this application provides an image denoising method, which includes: dividing the image to be denoised into blocks to obtain several image blocks; and obtaining the brightness information and weight of the image block according to the mean and variance of the image blocks in each color channel.
  • Information determine the color difference information of the image block in each color channel according to the brightness information of the image block and the mean value of the image block in each color channel; according to the brightness information of the image block and the color difference information in each color channel, Perform discrete cosine transform (DCT) denoising on the image blocks; based on each image block after DCT denoising and the weight information corresponding to each image block, the corresponding denoising image corresponding to the image to be denoised is obtained.
  • DCT discrete cosine transform
  • obtaining the brightness information and weight information of the image block based on the mean and variance of the image block in each color channel includes: determining the mean value of the image block based on the mean value of the image block in each color channel, and based on The mean value of the image block determines the brightness information of the image block; based on the variance of the image block in each color channel, the variance of the image block is determined, and based on the variance of the image block, the weight of the image block is determined information.
  • determining the color difference information of the image block in each color channel based on the brightness information of the image block and the mean value of the image block in each color channel includes: converting the mean value of the image block in each color channel Compare it with the brightness information of the image block to obtain the color difference information of the image block in each color channel.
  • performing DCT noise reduction on the image block based on the brightness information of the image block and the color difference information in each color channel includes: determining the image block based on the calibrated noise distribution curve and the brightness information of the image block. noise intensity, and set a hard threshold according to the noise intensity; and, after removing the color difference of the corresponding color from all pixels in the image block, perform a two-dimensional DCT transformation to obtain the DCT coefficient of the image block; based on the DCT of the image block
  • the coefficients are processed by the hard threshold to obtain the denoised DCT coefficients, and then a two-dimensional inverse DCT transform is performed to obtain the target image block; the color difference of each color channel is supplemented to the target image block respectively to obtain the DCT denoised coefficients.
  • Image block includes: determining the image block based on the calibrated noise distribution curve and the brightness information of the image block. noise intensity, and set a hard threshold according to the noise intensity; and, after removing the color difference of the corresponding color from all pixels in the image block
  • processing each image block after denoising based on DCT and the weight information corresponding to each image block to obtain a denoised image corresponding to the image to be denoised includes: combining the weight information Each image block is aggregated to obtain the denoised image.
  • dividing the image to be denoised into blocks to obtain several image blocks includes: dividing the image to be denoised into several image blocks according to a preset step size, wherein each image block has pixels in the length direction The number of pixels and the number of pixels in the width direction are both even numbers.
  • the image to be denoised is a CFA image.
  • this application provides an image noise reduction device, including:
  • the blocking module is configured to divide the image to be denoised into blocks to obtain several image blocks;
  • the acquisition module is configured to obtain the brightness information and weight information of the image block based on the mean and variance of the image block in each color channel;
  • a determination module configured to determine the color difference information of the image block in each color channel based on the brightness information of the image block and the mean value of the image block in each color channel;
  • a noise reduction module configured to perform DCT noise reduction on the image block based on the brightness information of the image block and the color difference information in each color channel;
  • the processing module is configured to process, based on each image block after DCT denoising and the weight information corresponding to each image block, to obtain a denoised image corresponding to the image to be denoised.
  • the present application provides a computer-readable storage medium on which a computer program is stored.
  • the computer program is executed by a processor, the image noise reduction method described in the first aspect is implemented.
  • the present application provides an electronic device, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor.
  • the processor executes the computer program, the first aspect is implemented.
  • the image noise reduction method is implemented.
  • the application provides a chip, including one or more interface circuits and one or more processors; the interface circuit is used to receive signals from a memory of an electronic device and send the said signal to the processor.
  • a signal includes a computer instruction stored in a memory; when the processor executes the computer instruction, the electronic device is caused to perform the image noise reduction method described in the first aspect.
  • this application provides an image noise reduction method, device, electronic equipment and chip. Compared with the current traditional image noise reduction methods, this application proposes a CFA image reduction method based on extended multi-channel DCT. Noise method, the method makes full use of the correlation between different color channels, can be directly applied to CFA images, and can improve the noise reduction effect of CFA images.
  • the image to be denoised is first divided into blocks to obtain several image blocks; then the brightness information and weight information of the image blocks are obtained according to the mean and variance of the image blocks in each color channel; and based on the brightness information of the image blocks and the image blocks
  • the mean value of each color channel is used to determine the color difference information of the image block in each color channel; then DCT noise reduction is performed on the image block based on the brightness information of the image block and the color difference information in each color channel; finally, the image block can be denoised based on DCT
  • Each image block and the weight information corresponding to each image block are processed to obtain a denoised image corresponding to the image to be denoised.
  • the correlation between different color channels in the CFA image can be used to denoise the CFA image, which improves the performance of CFA image denoising and better maintains the texture detail information in the image
  • the variance of CFA image blocks is used to calculate the weight information of the aggregated image blocks, which effectively suppresses problems such as "artifacts” and “ringing" in the texture area, and effectively maintains the high-frequency detail information of the image.
  • Figure 1 shows a schematic flow chart of an image noise reduction method provided by an embodiment of the present application
  • Figure 2 shows a schematic flow chart of another image noise reduction method provided by an embodiment of the present application
  • Figure 3 shows a schematic diagram of an example of CFA distribution in the embodiment of the present application.
  • FIG. 4 shows a schematic structural diagram of an image noise reduction device provided by an embodiment of the present application.
  • This embodiment provides an image noise reduction method, as shown in Figure 1, which can be applied to the end-side of image processing equipment (such as smart phones, tablets, drones, smart robots and other smart terminals).
  • the method includes:
  • Step 101 Divide the image to be denoised into blocks to obtain several image blocks.
  • the image to be denoised can first be divided into blocks to obtain several image blocks, which can be divided according to actual needs. For example, the greater the number of blocks, the better the effect will be, but the corresponding calculation will be more time-consuming. The smaller the number of blocks, the faster the operation will be, but the accuracy of the calculation results will decrease. Therefore, performance and effect need to be weighed and evaluated according to appropriate
  • the blocking standard is used to divide the image to be denoised into blocks.
  • steps 102 to 104 can be performed for each image block.
  • Step 102 Obtain the brightness information and weight information of the image block based on the mean and variance of the image block in each color channel.
  • the brightness information of the image block can be calculated based on the mean value of the image block in each color channel, and the weight information of the image block can be calculated based on the variance of the image block in each color channel. Then, a smaller value can be used for areas containing edges and textures. weight, while flat areas use larger weights so that problems such as "artifacts" and "ringing" can be suppressed later.
  • Step 103 Determine the color difference information of the image block in each color channel based on the brightness information of the image block and the mean value of the image block in each color channel.
  • Step 104 Perform DCT noise reduction on the image block based on the brightness information of the image block and the color difference information in each color channel.
  • this embodiment can use DCT transformation and hard threshold processing to obtain the image block based on the brightness information of the image block. And the color difference information in each color channel, DCT noise reduction is performed on the image block.
  • this embodiment does not introduce problems such as color cast, does not need to perform noise reduction processing on different color channels separately, and can make full use of the correlation between different color channels.
  • Step 105 Based on each DCT denoised image block and the weight information corresponding to each image block, process to obtain a denoised image corresponding to the image to be denoised.
  • the CFA image denoising method proposed in this embodiment based on extended multi-channel DCT compared with the current traditional image denoising method, makes full use of the correlation between different color channels and can be directly applied to CFA images. , which can improve the noise reduction effect of CFA images.
  • the image to be denoised is first divided into blocks to obtain several image blocks; then the brightness information and weight information of the image blocks are obtained according to the mean and variance of the image blocks in each color channel; and based on the brightness information of the image blocks and the image blocks
  • the mean value of each color channel is used to determine the color difference information of the image block in each color channel; then DCT noise reduction is performed on the image block based on the brightness information of the image block and the color difference information in each color channel; finally, the image block can be denoised based on DCT
  • Each image block and the weight information corresponding to each image block are processed to obtain a denoised image corresponding to the image to be denoised.
  • the correlation between different color channels in the CFA image can be used to denoise the CFA image, which improves the performance of the CFA image denoising and better maintains the texture detail information in the image.
  • the variance of the CFA image blocks is used to calculate the weight information of the aggregated image blocks, which effectively suppresses problems such as "artifacts” and “ringing" in the texture area, and effectively maintains the high-frequency detail information of the image.
  • this embodiment takes the image to be denoised as a CFA image as an example, and provides a specific method as shown in Figure 2,
  • the method includes:
  • Step 201 Divide the CFA image to be denoised into several image blocks according to a preset step size.
  • the number of pixels in the length direction and the number of pixels in the width direction of each image block are both even numbers.
  • M can be the number of pixels in the length direction
  • N can be the number of pixels in the width direction.
  • M and N can both take an even number, such as usually 8 or 16.
  • Common CFA distributions include GRBG, GBRG, RGGB and BGGR.
  • the CFA of GRBG distribution can be shown in Figure 3. According to the prior information of natural images, it can be known that there is local smoothness in the image, that is, the color and brightness differences in local areas of the image are small. In CFA images, local smoothness also exists in different color channels. Therefore, it can be assumed that the difference between different color channels in the local area in the CFA image is fixed. This embodiment makes full use of the differences between different color channels, so that the CFA image noise reduction method based on extended multi-channel DCT can be directly applied to CFA images, effectively improving the CFA image noise reduction effect.
  • the DCT noise reduction method is frequency domain noise reduction and uses hard threshold processing, which will cause problems such as "artifacts” and "ringing" in image blocks containing edge information.
  • an average weighting method is used to calculate Noise reduction pixels.
  • this embodiment introduces weight information in the image aggregation process. Smaller weights are used for areas containing edges and textures, while larger weights are used for flat areas. Therefore, “artifacts” and “vibration” can be suppressed. Bell” and other issues.
  • steps 202 to 207 can be continued.
  • steps 202 to 206 may be processes implemented for each image block.
  • Step 202 Obtain the brightness information and weight information of the image block based on the mean and variance of the image block in each color channel.
  • step 202 may specifically include: determining the mean value of the image block based on the mean value of the image block in each color channel, and determining the brightness information of the image block based on the mean value of the image block; and based on the variance of the image block in each color channel. , determine the variance of the image block, and determine the weight information of the image block based on the variance of the image block.
  • any image block calculate the mean and variance of the four color channels respectively, and then perform a weighted average of the four color channels to obtain the mean L of the image block.
  • L is regarded as the brightness value of the image block; for the four
  • the variance of the color channel is weighted and averaged to obtain the variance V of the image block, and the variance V is used to calculate the weight coefficient of the image block, as shown in Formula 1:
  • weight is the weight coefficient of the image block
  • V is the variance of the image block
  • Step 203 Compare the mean value of the image block in each color channel with the brightness information of the image block to obtain the color difference information of the image block in each color channel.
  • the color difference B is calculated based on the difference between the mean value of the four color channels and the brightness information L.
  • Step 204 Determine the noise intensity of the image block based on the calibrated noise distribution curve and the brightness information of the image block, set a hard threshold according to the noise intensity, and remove the color difference of the corresponding color from all pixels in the image block, and then perform the second step.
  • Dimensional DCT transform to obtain the DCT coefficients of the image block.
  • the image block noise intensity ⁇ is estimated, and a hard threshold is set according to the noise intensity ⁇ .
  • a hard threshold is set according to the noise intensity ⁇ .
  • the two-dimensional DCT transformation is as shown in Formula 2:
  • c(u) and c(v) in formula 2 are both constant terms, respectively expressed as F(u,v) is the transformation coefficient vector obtained by two-dimensional DCT transformation of the image block, u and v are the two transform domain sequences of the image block, M is the number of pixels in the length direction of the image block, and N is the width direction of the image block.
  • the number of pixels, f (i, j) is the two-dimensional vector in the spatial domain of the image block, and i, j are the two spatial domain sequences of the image block respectively.
  • Step 205 Use hard threshold processing to obtain denoised DCT coefficients based on the DCT coefficients of the image block, and then perform two-dimensional inverse DCT transformation to obtain the target image block.
  • c(u) and c(v) in formula 3 are both constant terms, respectively expressed as F(u,v) is the transformation coefficient vector obtained by two-dimensional DCT transformation of the image block, u and v are the two transform domain sequences of the image block, M is the number of pixels in the length direction of the image block, and N is the width direction of the image block.
  • the number of pixels, f (i, j) is the two-dimensional vector in the spatial domain of the image block, and i, j are the two spatial domain sequences of the image block respectively.
  • Step 206 Supplement the color difference of each color channel to the target image block to obtain the image block after DCT noise reduction.
  • the color difference B of different color channels is added back to the denoised image block (ie, the target image block) to obtain the denoised CFA image block.
  • Step 207 Based on each DCT denoised image block and the weight information corresponding to each image block, process to obtain a denoised CFA image corresponding to the CFA image to be denoised.
  • step 207 may specifically include: aggregating various image blocks combined with weight information to obtain the denoised image. For example, all image blocks are denoised in the above manner (process shown in steps 202 to 206), multiplied by the corresponding weight coefficient weight, and then the denoised images are aggregated into a denoised CFA image.
  • This embodiment proposes a CFA image denoising method based on extended multi-channel DCT, which makes full use of the correlation between different color channels and can be directly applied to CFA images without introducing color cast problems.
  • Using different weight coefficients can suppress problems such as "ringing” and “artifacts” introduced by frequency domain noise reduction methods. It can suppress the noise in the CFA image and better maintain the texture details in the image. If this solution is applied to a camera image signal processor (ISP), the structure will be simpler and hardware resources will be saved.
  • ISP camera image signal processor
  • this embodiment provides an image noise reduction device.
  • the device includes: a blocking module 31, an acquisition module 32, and a determination module 33. , noise reduction module 34, processing module 35.
  • the blocking module 31 is configured to divide the image to be denoised into blocks to obtain several image blocks;
  • the acquisition module 32 is configured to acquire the brightness information and weight information of the image block according to the mean and variance of the image block in each color channel;
  • the determination module 33 is configured to determine the color difference information of the image block in each color channel based on the brightness information of the image block and the mean value of the image block in each color channel;
  • the noise reduction module 34 is configured to perform DCT noise reduction on the image block according to the brightness information of the image block and the color difference information in each color channel;
  • the processing module 35 is configured to process, based on each image block after DCT denoising and the weight information corresponding to each image block, to obtain a denoised image corresponding to the image to be denoised.
  • the acquisition module 32 is specifically configured to determine the mean value of the image block based on the mean value of the image block in each color channel, and determine the brightness information of the image block based on the mean value of the image block. ; Determine the variance of the image block based on the variance of the image block in each color channel, and determine the weight information of the image block based on the variance of the image block.
  • the determination module 33 is specifically configured to compare the mean value of the image block in each color channel with the brightness information of the image block, and obtain the mean value of the image block in each color channel. Color difference information.
  • the noise reduction module 34 is specifically configured to determine the noise intensity of the image block based on the calibrated noise distribution curve and the brightness information of the image block, and set a hard threshold according to the noise intensity; and, After removing the color difference of the corresponding colors from all pixels in the image block, a two-dimensional DCT transformation is performed to obtain the DCT coefficients of the image block; based on the DCT coefficients of the image block, the hard threshold processing is used to obtain the denoised DCT coefficients, and then the DCT coefficients are obtained.
  • the target image block is obtained by two-dimensional inverse DCT transformation; the color difference of each color channel is respectively supplemented to the target image block to obtain the image block after DCT noise reduction.
  • the processing module 35 is specifically configured to aggregate various image blocks combined with weight information to obtain the denoised image.
  • the blocking module 31 is specifically configured to divide the image to be denoised into several image blocks according to a preset step size, where the number of pixels in the length direction and the number of pixels in the width direction of each image block are The numbers are all even.
  • the image to be denoised is a CFA image.
  • this embodiment also provides a computer-readable storage medium on which a computer program is stored.
  • the computer program is executed by a processor, the above mentioned Figures 1 to 2 are implemented.
  • the technical solution of this application can be embodied in the form of a software product.
  • the software product can be stored in a non-volatile storage medium (can be a CD-ROM, U disk, mobile hard disk, etc.), including several Instructions are used to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method of each implementation scenario of this application.
  • a computer device which may be a personal computer, a server, or a network device, etc.
  • embodiments of the present application also provide an electronic device, such as a smart phone, a tablet computer, an unmanned device Intelligent terminals such as machines and intelligent robots, the device includes a storage medium and a processor; the storage medium is used to store computer programs; and the processor is used to execute the computer program to implement the above-mentioned methods shown in Figures 1 to 2.
  • an electronic device such as a smart phone, a tablet computer, an unmanned device Intelligent terminals such as machines and intelligent robots
  • the device includes a storage medium and a processor; the storage medium is used to store computer programs; and the processor is used to execute the computer program to implement the above-mentioned methods shown in Figures 1 to 2.
  • the above-mentioned physical devices may also include user interfaces, network interfaces, cameras, radio frequency (Radio Frequency, RF) circuits, sensors, audio circuits, WI-FI modules, etc.
  • the user interface may include a display screen (Display), an input unit such as a keyboard (Keyboard), etc.
  • the optional user interface may also include a USB interface, a card reader interface, etc.
  • Optional network interfaces may include standard wired interfaces, wireless interfaces (such as WI-FI interfaces), etc.
  • the above-mentioned physical device structure does not constitute a limitation on the physical device, and may include more or fewer components, or combine certain components, or arrange different components.
  • the storage medium may also include an operating system and a network communication module.
  • the operating system is a program that manages the hardware and software resources of the above-mentioned physical devices and supports the operation of information processing programs and other software and/or programs.
  • the network communication module is used to realize communication between components within the storage medium, as well as communication with other hardware and software in the information processing physical device.
  • this embodiment also provides a chip including one or more interface circuits and one or more processors; so
  • the interface circuit is configured to receive signals from a memory of an electronic device and send the signals to the processor, where the signals include computer instructions stored in the memory; when the processor executes the computer instructions, the The electronic device performs the above-mentioned method shown in FIGS. 1 to 2 .

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Image Processing (AREA)

Abstract

一种图像降噪方法、装置及芯片，涉及图像处理技术领域。通过应用本图像降噪方法，可利用CFA图像中不同颜色通道之间相关性对CFA图像进行降噪，提高了CFA图像降噪的性能，且较好地保持了图像中的纹理细节信息，并利用CFA图像块的方差计算聚合图像块的权重信息，有效地抑制了纹理区域的"伪影"和"振铃"等问题。

Description

图像降噪方法、装置及芯片 技术领域
本申请涉及图像处理技术领域，特别涉及一种图像降噪方法、装置及芯片。
背景技术
数码相机中普遍采用带有彩色滤波阵列(Color Filter Array，CFA)的图像传感器来获取带有颜色场景图片。由于每个像素只能记录一种颜色分量，需要利用插值方法将对应像素位置缺失的颜色进行补全，将数字图像从单通道灰度图像拓展成三通道的颜色图像即去马赛克(demosaic)。然而，数字图像在采集过程中，不可避免地会引入噪声，若直接进行去马赛克，会引入结构性噪声且会混合不同颜色通道的噪声等问题，后续的图像处理方法很难处理这些问题，严重影响了图像的质量。
目前，大部分传统的图像降噪方法可直接应用于灰度图、红绿蓝颜色系统(Red Yellow Blue system，RGB)图像和YUV图像，而CFA图像数据分布不符合常见图像的特点，因此传统的图像降噪方法并不能直接应用于CFA图像，进而影响了CFA图像的降噪效果。
发明内容
有鉴于此，本申请提供了一种图像降噪方法、装置及芯片，主要目的在于改善目前传统的图像降噪方法并不能直接应用于CFA图像，进而会影响CFA图像的降噪效果的技术问题。
第一方面，本申请提供了一种图像降噪方法，包括：对待降噪图像进行分块得到若干个图像块；根据图像块在各个颜色通道的均值和方差，获取图像块的亮度信息和权重信息；依据图像块的亮度信息和图像块在所述各个颜色通道的均值，确定图像块在所述各个颜色通道的色差信息；根据图像块的亮度信息和在所述各个颜色通道的色差信息，对图像块进行离散余弦变换(Discrete Cosine Transform，DCT)降噪；基于DCT降噪后的各个图像块以及与所述各个图像块各自对应的权重信息，处理得到与所述待降噪图像对应降噪后的图像。
可选的，所述根据图像块在各个颜色通道的均值和方差，获取图像块的亮度信息和权重信息，包括：依据图像块在各个颜色通道的均值，确定所述图像块的均值，并依据所述图像块的均值，确定所述图像块的亮度信息；依据图像块在各个颜色通道的方差，确定所述图像块的方差，并依据所述图像块的方差，确定所述图像块的权重信息。
可选的，所述依据图像块的亮度信息和图像块在所述各个颜色通道的均值，确定图像块在所述各个颜色通道的色差信息，包括：将图像块在所述各个颜色通道的均值与所述图像块的亮度信息进行比对，获得所述图像块在所述各个颜色通道的色差信息。
可选的，所述根据图像块的亮度信息和在所述各个颜色通道的色差信息，对图像块进行DCT降噪，包括：根据标定的噪点分布图曲线和图像块的亮度信息，确定图像块的噪声强度，并按照所述噪声强度设置硬阈值；及，将图像块中所有像素分别去掉对应颜色的色差后，再进行二维DCT变换得到图像块的DCT系数；基于所述图像块的DCT系数利用所述硬阈值处理得到降噪后的DCT系数，再进行二维 逆DCT变换得到目标图像块；将所述各个颜色通道的色差分别补充到所述目标图像块，得到DCT降噪后的图像块。
可选的，所述基于DCT降噪后的各个图像块以及与所述各个图像块各自对应的权重信息，处理得到与所述待降噪图像对应降噪后的图像，包括：将结合权重信息的各个图像块聚合得到所述降噪后的图像。
可选的，所述对待降噪图像进行分块得到若干个图像块，包括：按照预设步长将所述待降噪图像分成若干个图像块，其中，每个图像块的长度方向像素个数和宽度方向像素个数均取偶数。
可选的，所述待降噪图像为CFA图像。
第二方面，本申请提供了一种图像降噪装置，包括：
分块模块，被配置为对待降噪图像进行分块得到若干个图像块；
获取模块，被配置为根据图像块在各个颜色通道的均值和方差，获取图像块的亮度信息和权重信息；
确定模块，被配置为依据图像块的亮度信息和图像块在所述各个颜色通道的均值，确定图像块在所述各个颜色通道的色差信息；
降噪模块，被配置为根据图像块的亮度信息和在所述各个颜色通道的色差信息，对图像块进行DCT降噪；
处理模块，被配置为基于DCT降噪后的各个图像块以及与所述各个图像块各自对应的权重信息，处理得到与所述待降噪图像对应降噪后的图像。
第三方面，本申请提供了一种计算机可读存储介质，其上存储有计算机程序，所述计算机程序被处理器执行时实现第一方面所述的图像降噪方法。
第四方面，本申请提供了一种电子设备，包括存储介质、处理器及存储在存储介质上并可在处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现第一方面所述的图像降噪方法。
第五方面，本申请提供了一种芯片，包括一个或多个接口电路和一个或多个处理器；所述接口电路用于从电子设备的存储器接收信号，并向所述处理器发送所述信号，所述信号包括存储器中存储的计算机指令；当所述处理器执行所述计算机指令时，使得所述电子设备执行第一方面所述的图像降噪方法。
借由上述技术方案，本申请提供的一种图像降噪方法、装置、电子设备及芯片，与目前传统的图像降噪方法相比，本申请提出了一种基于拓展多通道DCT的CFA图像降噪方法，方法充分利用了不同颜色通道之间的相关性，可以直接应用于CFA图像，可提高CFA图像的降噪效果。具体的，首先对待降噪图像进行分块得到若干个图像块；再根据图像块在各个颜色通道的均值和方差，获取图像块的亮度信息和权重信息；并依据图像块的亮度信息和图像块在各个颜色通道的均值，确定图像块在各个颜色通道的色差信息；然后根据图像块的亮度信息和在各个颜色通道的色差信息，对图像块进行DCT降噪；最后可基于DCT降噪后的各个图像块以及与各个图像块各自对应的权重信息，处理得到与待降噪图像对应降噪后的图像。通过应用本申请的技术方案，可利用CFA图像中不同颜色通道之间相关性对CFA图像进行降噪，提高了CFA图像降噪的性能，且较好地保持了图像中的纹理细节信息，并利用CFA图像块的方差计算聚合图像块的权重信息，有效地抑制了纹理区域的“伪影”和“振铃”等问题，有效保持了图像的高频细节信息。
上述说明仅是本申请技术方案的概述，为了能够更清楚了解本申请的技术手段，而可依照说明书的内容予以实施，并且为了让本申请的上述和其它目的、特征和优点能够更明显易懂，以下特举本申请的具体实施方式。
附图说明
此处的附图被并入说明书中并构成本说明书的一部分，示出了符合本申请的实施例，并与说明书一起用于解释本申请的原理。
为了更清楚地说明本申请实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，对于本领域普通技术人员而言，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。
图1示出了本申请实施例提供的一种图像降噪方法的流程示意图；
图2示出了本申请实施例提供的另一种图像降噪方法的流程示意图；
图3示出了本申请实施例中CFA分布的一种示例的示意图；
图4示出了本申请实施例提供的一种图像降噪装置的结构示意图。
具体实施方式
下面将参照附图更详细地描述本申请的实施例。需要说明的是，在不冲突的情况下，本申请中的实施例及实施例中的特征可以相互组合。
为了改善目前传统的图像降噪方法并不能直接应用于CFA图像，进而会影响CFA图像的降噪效果的技术问题。本实施例提供了一种图像降噪方法，如图1所示，可应用于图像处理设备(如智能手机、平板电脑、无人机、智能机器人等智能终端)端侧，该方法包括：
步骤101、对待降噪图像进行分块得到若干个图像块。
在本实施例中，在待降噪图像输入进来，可先对该待降噪图像进行分块得到若干个图像块，具体可根据实际需求进行切分。例如，分块数量越多效果越好，但是相应的计算更耗时；而分块数量越少，运行越快，但是计算结果的精度会下降，因此需要对性能和效果进行权衡评估，按照合适的分块标准对待降噪图像进行分块切分。
在切分得到若干个图像块以后，针对每一个图像块，可执行步骤102至104所述的过程。
步骤102、根据图像块在各个颜色通道的均值和方差，获取图像块的亮度信息和权重信息。
例如，可根据图像块在各个颜色通道的均值，计算图像块的亮度信息，并根据图像块在各个颜色通道的方差，计算图像块的权重信息，进而可对于含有边缘和纹理区域，采用较小的权重，而平坦区域采用较大的权重，以便后续可抑制“伪影”和“振铃”等问题。
步骤103、依据图像块的亮度信息和图像块在各个颜色通道的均值，确定图像块在各个颜色通道的色差信息。
步骤104、根据图像块的亮度信息和在各个颜色通道的色差信息，对图像块进行DCT降噪。
因为噪声往往是高频部分，可以将图像转换到频域，再进行高频处理，将高频部分滤掉，因此本实施例可采用DCT变换盒硬阈值处理的方式，根据图像块的亮度信息和在各个颜色通道的色差信息，对 图像块进行DCT降噪。与常规的DCT降噪方式相比，本实施例不会引入偏色等问题，无需对不同颜色通道分别进行降噪处理，可充分利用不同颜色通道之间的相关性。
步骤105、基于DCT降噪后的各个图像块以及与各个图像块各自对应的权重信息，处理得到与待降噪图像对应降噪后的图像。
目前为了实现对CFA图像进行图像降噪，可以将CFA图像中不同颜色通道数据分开，得到不同颜色通道的子图像，并对子图像分别进行降噪处理，然后将子图像合并成降噪的CFA图像。这种方式可以有效降低CFA图像中的噪声，但是没有充分利用不同颜色通道之间的相关性，而通道之间相关性对减少最终图像中各种偏色和伪像等问题至关重要，因此有必要利用不同颜色通道之间的相关性进行去噪。
而本实施例提出的这种可基于拓展多通道DCT的CFA图像降噪方法，与目前传统的图像降噪方法相比，充分利用了不同颜色通道之间的相关性，可以直接应用于CFA图像，可提高CFA图像的降噪效果。具体的，首先对待降噪图像进行分块得到若干个图像块；再根据图像块在各个颜色通道的均值和方差，获取图像块的亮度信息和权重信息；并依据图像块的亮度信息和图像块在各个颜色通道的均值，确定图像块在各个颜色通道的色差信息；然后根据图像块的亮度信息和在各个颜色通道的色差信息，对图像块进行DCT降噪；最后可基于DCT降噪后的各个图像块以及与各个图像块各自对应的权重信息，处理得到与待降噪图像对应降噪后的图像。通过应用本实施例的技术方案，可利用CFA图像中不同颜色通道之间相关性对CFA图像进行降噪，提高了CFA图像降噪的性能，且较好地保持了图像中的纹理细节信息，并利用CFA图像块的方差计算聚合图像块的权重信息，有效地抑制了纹理区域的“伪影”和“振铃”等问题，有效保持了图像的高频细节信息。
进一步的，作为上述实施例的细化和扩展，为了完整说明本实施例方法的具体实现过程，本实施例以待降噪图像为CFA图像作为示例，提供了如图2所示的具体方法，该方法包括：
步骤201、按照预设步长将待降噪的CFA图像分成若干个图像块。
其中，每个图像块的长度方向像素个数和宽度方向像素个数均取偶数。例如，对于输入含噪声的CFA图像，按照设置的步长分成若干尺寸为M×N的图像块，M可为长度方向的像素个数，N可为宽度方向的像素个数，并且为了保证后续的图像降噪效果，M和N可均取偶数，如通常可取8或16等。
常见的CFA分布包括GRBG、GBRG、RGGB和BGGR，其中GRBG分布的CFA可如图3所示。根据自然图像的先验信息可知图像中存在局部平滑性，即图像局部区域的颜色和亮度差异较小。在CFA图像中，不同的颜色通道也存在局部平滑性。因此，可以假设在CFA图像中局部区域内，不同颜色通道均存在之间的差异是固定不变的。本实施例充分利用了不同颜色通道之间的差异，使得基于拓展多通道DCT的CFA图像降噪方法可以直接应用于CFA图像，有效提高了CFA图像降噪效果。
DCT降噪方法是频域降噪并采用硬阈值处理，会导致含有边缘信息的图像块中存在“伪影”和“振铃”等问题，而图像块聚合过程中，采用平均加权的方式计算降噪像素。本实施例为了抑制这些问题，在图像聚合过程中，引入权重信息，对于含有边缘和纹理区域，采用较小的权重，而平坦区域采用较大的权重，因此可以抑制“伪影”和“振铃”等问题。
本实施例为了实现上述这些目的，在通过步骤201实现将待降噪的CFA图像分成若干个图像块以后，可继续执行步骤202至207所示的过程。其中，步骤202至206可以是针对每一个图像块所实现的过程。
步骤202、根据图像块在各个颜色通道的均值和方差，获取图像块的亮度信息和权重信息。
可选的，步骤202具体可包括：依据图像块在各个颜色通道的均值，确定图像块的均值，并依据图像块的均值，确定图像块的亮度信息；以及依据图像块在各个颜色通道的方差，确定图像块的方差，并依据图像块的方差，确定图像块的权重信息。
例如，对于任意一个图像块，分别计算四个颜色通道的均值和方差，将四个颜色通道的均值再进行加权平均得到图像块的均值L，将L视为图像块的亮度值；对四个颜色通道的方差进行加权平均得到图像块的方差V，利用方差V计算该图像块的权重系数，如公式一所示：
Figure PCTCN2022111921-appb-000001
其中，weight为图像块的权重系数，V为图像块的方差。
步骤203、将图像块在各个颜色通道的均值与图像块的亮度信息进行比对，获得图像块在各个颜色通道的色差信息。
例如，根据四个颜色通道均值与亮度信息L差值计算色差B。
步骤204、根据标定的噪点分布图曲线和图像块的亮度信息，确定图像块的噪声强度，并按照噪声强度设置硬阈值，以及将图像块中所有像素分别去掉对应颜色的色差后，再进行二维DCT变换得到图像块的DCT系数。
例如，根据标定的噪点分布图曲线(noise profile)和图像块的亮度信息L，估计图像块噪声强度σ，并按照该噪声强度σ设置硬阈值。将图像块中的所有像素分别减去对应颜色的色差B。然后将减去色差的图像块进行二维DCT变换，得到图像块的DCT系数，其中二维DCT变换如公式二所示：
Figure PCTCN2022111921-appb-000002
其中，公式二中的c(u)和c(v)均是常数项，分别表示为
Figure PCTCN2022111921-appb-000003
F(u,v)为图像块经过二维DCT变换得到的变换系数向量，u、v为图像块的两个变换域序列，M为图像块长度方向的像素个数，N为图像块宽度方向的像素个数，f(i,j)是图像块的空间域二维向量，i,j分别为图像块的两个空间域序列。
步骤205、基于图像块的DCT系数利用硬阈值处理得到降噪后的DCT系数，再进行二维逆DCT变换得到目标图像块。
例如，在上述示例的基础上，利用硬阈值处理得到降噪的DCT系数，并进行二维逆DCT变换得到降噪后图像块(即目标图像块)，其中逆变换如公式三所示：
Figure PCTCN2022111921-appb-000004
其中，公式三中的c(u)和c(v)均是常数项，分别表示为
Figure PCTCN2022111921-appb-000005
F(u,v)为图像块经过二维DCT变换得到的变换系数向量，u、v为图像块的两个变换域序列，M为图像块长度方向的像素个数，N为图像块宽度方向的像素个数，f(i,j)是图像块的空间域二维向量，i,j分别为图像块的两个空间域序列。
步骤206、将各个颜色通道的色差分别补充到目标图像块，得到DCT降噪后的图像块。
例如，将不同颜色通道的色差B分别加回降噪后的图像块(即目标图像块)，得到降噪后的CFA图像块。
步骤207、基于DCT降噪后的各个图像块以及与各个图像块各自对应的权重信息，处理得到与待降噪的CFA图像对应降噪后的CFA图像。
可选的，步骤207具体可包括：将结合权重信息的各个图像块聚合得到所述降噪后的图像。例如，将所有的图像块均按照以上方式(步骤202至206所示的过程)进行降噪，并乘以对应权重系数weight，然后将降噪后的图像聚合成降噪的CFA图像。
本实施例提出了一种基于拓展多通道DCT的CFA图像降噪方法，充分利用了不同颜色通道之间的相关性，可以直接应用于CFA图像，且不引入偏色问题，结合对不同图像块采用不同的权重系数，可以抑制频域降噪方法引入的“振铃”和“伪影”等问题。可以抑制CFA图像中的噪声，较好地保持图像中纹理细节信息。本方案若应用于相机图像信号处理器(Image Signal Processor，ISP)中，结构更加简单，更节省硬件资源。
进一步的，作为图1至图2所示方法的具体实现，本实施例提供了一种图像降噪装置，如图4所示，该装置包括：分块模块31、获取模块32、确定模块33、降噪模块34、处理模块35。
分块模块31，被配置为对待降噪图像进行分块得到若干个图像块；
获取模块32，被配置为根据图像块在各个颜色通道的均值和方差，获取图像块的亮度信息和权重信息；
确定模块33，被配置为依据图像块的亮度信息和图像块在所述各个颜色通道的均值，确定图像块在所述各个颜色通道的色差信息；
降噪模块34，被配置为根据图像块的亮度信息和在所述各个颜色通道的色差信息，对图像块进行DCT降噪；
处理模块35，被配置为基于DCT降噪后的各个图像块以及与所述各个图像块各自对应的权重信息，处理得到与所述待降噪图像对应降噪后的图像。
在具体的应用场景中，获取模块32，具体被配置为依据图像块在各个颜色通道的均值，确定所述图像块的均值，并依据所述图像块的均值，确定所述图像块的亮度信息；依据图像块在各个颜色通道的方差，确定所述图像块的方差，并依据所述图像块的方差，确定所述图像块的权重信息。
在具体的应用场景中，确定模块33，具体被配置为将图像块在所述各个颜色通道的均值与所述图像块的亮度信息进行比对，获得所述图像块在所述各个颜色通道的色差信息。
在具体的应用场景中，降噪模块34，具体被配置为根据标定的噪点分布图曲线和图像块的亮度信息，确定图像块的噪声强度，并按照所述噪声强度设置硬阈值；及，将图像块中所有像素分别去掉对应颜色的色差后，再进行二维DCT变换得到图像块的DCT系数；基于所述图像块的DCT系数利用所述硬阈值处理得到降噪后的DCT系数，再进行二维逆DCT变换得到目标图像块；将所述各个颜色通道的色差分别补充到所述目标图像块，得到DCT降噪后的图像块。
在具体的应用场景中，处理模块35，具体被配置为将结合权重信息的各个图像块聚合得到所述降噪后的图像。
在具体的应用场景中，分块模块31，具体被配置为按照预设步长将所述待降噪图像分成若干个图像块，其中，每个图像块的长度方向像素个数和宽度方向像素个数均取偶数。
在具体的应用场景中，可选的，所述待降噪图像为CFA图像。
需要说明的是，本实施例提供的一种图像降噪装置所涉及各功能单元的其它相应描述，可以参考图1至图2中的对应描述，在此不再赘述。
基于上述如图1至图2所示方法，相应的，本实施例还提供了一种计算机可读存储介质，其上存储有计算机程序，该计算机程序被处理器执行时实现上述如图1至图2所示的方法。
基于这样的理解，本申请的技术方案可以以软件产品的形式体现出来，该软件产品可以存储在一个非易失性存储介质(可以是CD-ROM，U盘，移动硬盘等)中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本申请各个实施场景的方法。
基于上述如图1至图2所示的方法，以及图4所示的虚拟装置实施例，为了实现上述目的，本申请实施例还提供了一种电子设备，如智能手机、平板电脑、无人机、智能机器人等智能终端，该设备包括存储介质和处理器；存储介质，用于存储计算机程序；处理器，用于执行计算机程序以实现上述如图1至图2所示的方法。
可选的，上述实体设备还可以包括用户接口、网络接口、摄像头、射频(Radio Frequency，RF)电路，传感器、音频电路、WI-FI模块等等。用户接口可以包括显示屏(Display)、输入单元比如键盘(Keyboard)等，可选用户接口还可以包括USB接口、读卡器接口等。网络接口可选的可以包括标准的有线接口、无线接口(如WI-FI接口)等。
本领域技术人员可以理解，本实施例提供的上述实体设备结构并不构成对该实体设备的限定，可以包括更多或更少的部件，或者组合某些部件，或者不同的部件布置。
存储介质中还可以包括操作系统、网络通信模块。操作系统是管理上述实体设备硬件和软件资源的程序，支持信息处理程序以及其它软件和/或程序的运行。网络通信模块用于实现存储介质内部各组件之间的通信，以及与信息处理实体设备中其它硬件和软件之间通信。
基于上述如图1至图2所示的方法，以及图4所示的虚拟装置实施例，本实施例还提供了一种芯片，包括一个或多个接口电路和一个或多个处理器；所述接口电路用于从电子设备的存储器接收信号，并向所述处理器发送所述信号，所述信号包括存储器中存储的计算机指令；当所述处理器执行所述计算机指令时，使得所述电子设备执行上述如图1至图2所示的方法。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到本申请可以借助软件加必要的通用硬件平台的方式来实现，也可以通过硬件实现。通过应用本实施例的方案，提出了一种基于拓展多通道DCT的CFA图像降噪方法，充分利用了不同颜色通道之间的相关性，可以直接应用于CFA图像，且不引入偏色问题，结合对不同图像块采用不同的权重系数，可以抑制频域降噪方法引入的“振铃”和“伪影”等问题。可以抑制CFA图像中的噪声，较好地保持图像中纹理细节信息。本方案若应用于相机ISP中，结构更加简单，更节省硬件资源。
需要说明的是，在本文中，诸如“第一”和“第二”等之类的关系术语仅仅用来将一个实体或者操作与另一个实体或操作区分开来，而不一定要求或者暗示这些实体或操作之间存在任何这种实际的关系或者顺序。而且，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者设备所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括所述要素的过程、方法、物品或者设备中还存在另外的相同要素。
以上所述仅是本申请的具体实施方式，使本领域技术人员能够理解或实现本申请。对这些实施例的多种修改对本领域的技术人员来说将是显而易见的，本文中所定义的一般原理可以在不脱离本申请的精神或范围的情况下，在其它实施例中实现。因此，本申请将不会被限制于本文所述的这些实施例，而是要符合与本文所申请的原理和新颖特点相一致的最宽的范围。

Claims (11)

  1. 一种图像降噪方法，其特征在于，包括：
    对待降噪图像进行分块得到若干个图像块；
    根据图像块在各个颜色通道的均值和方差，获取图像块的亮度信息和权重信息；
    依据图像块的亮度信息和图像块在所述各个颜色通道的均值，确定图像块在所述各个颜色通道的色差信息；
    根据图像块的亮度信息和在所述各个颜色通道的色差信息，对图像块进行离散余弦变换DCT降噪；
    基于DCT降噪后的各个图像块以及与所述各个图像块各自对应的权重信息，处理得到与所述待降噪图像对应降噪后的图像。
  2. 根据权利要求1所述的方法，其特征在于，所述根据图像块在各个颜色通道的均值和方差，获取图像块的亮度信息和权重信息，包括：
    依据图像块在各个颜色通道的均值，确定所述图像块的均值，并依据所述图像块的均值，确定所述图像块的亮度信息；
    依据图像块在各个颜色通道的方差，确定所述图像块的方差，并依据所述图像块的方差，确定所述图像块的权重信息。
  3. 根据权利要求1所述的方法，其特征在于，所述依据图像块的亮度信息和图像块在所述各个颜色通道的均值，确定图像块在所述各个颜色通道的色差信息，包括：
    将图像块在所述各个颜色通道的均值与所述图像块的亮度信息进行比对，获得所述图像块在所述各个颜色通道的色差信息。
  4. 根据权利要求1所述的方法，其特征在于，所述根据图像块的亮度信息和在所述各个颜色通道的色差信息，对图像块进行DCT降噪，包括：
    根据标定的噪点分布图曲线和图像块的亮度信息，确定图像块的噪声强度，并按照所述噪声强度设置硬阈值；及，
    将图像块中所有像素分别去掉对应颜色的色差后，再进行二维DCT变换得到图像块的DCT系数；
    基于所述图像块的DCT系数利用所述硬阈值处理得到降噪后的DCT系数，再进行二维逆DCT变换得到目标图像块；
    将所述各个颜色通道的色差分别补充到所述目标图像块，得到DCT降噪后的图像块。
  5. 根据权利要求1所述的方法，其特征在于，所述基于DCT降噪后的各个图像块以及与所述各个图像块各自对应的权重信息，处理得到与所述待降噪图像对应降噪后的图像，包括：
    将结合权重信息的各个图像块聚合得到所述降噪后的图像。
  6. 根据权利要求1所述的方法，其特征在于，所述对待降噪图像进行分块得到若干个图像块，包括：
    按照预设步长将所述待降噪图像分成若干个图像块，其中，每个图像块的长度方向像素个数和宽度方向像素个数均取偶数。
  7. 根据权利要求1至6任一项所述的方法，其特征在于，所述待降噪图像为彩色滤波阵列CFA图像。
  8. 一种图像降噪装置，其特征在于，包括：
    分块模块，被配置为对待降噪图像进行分块得到若干个图像块；
    获取模块，被配置为根据图像块在各个颜色通道的均值和方差，获取图像块的亮度信息和权重信息；
    确定模块，被配置为依据图像块的亮度信息和图像块在所述各个颜色通道的均值，确定图像块在所述各个颜色通道的色差信息；
    降噪模块，被配置为根据图像块的亮度信息和在所述各个颜色通道的色差信息，对图像块进行离散余弦变换DCT降噪；
    处理模块，被配置为基于DCT降噪后的各个图像块以及与所述各个图像块各自对应的权重信息，处理得到与所述待降噪图像对应降噪后的图像。
  9. 一种计算机可读存储介质，其上存储有计算机程序，其特征在于，所述计算机程序被处理器执行时实现权利要求1至7中任一项所述的方法。
  10. 一种电子设备，包括存储介质、处理器及存储在存储介质上并可在处理器上运行的计算机程序，其特征在于，所述处理器执行所述计算机程序时实现权利要求1至7中任一项所述的方法。
  11. 一种芯片，其特征在于，包括一个或多个接口电路和一个或多个处理器；所述接口电路用于从电子设备的存储器接收信号，并向所述处理器发送所述信号，所述信号包括存储器中存储的计算机指令；当所述处理器执行所述计算机指令时，使得所述电子设备执行权利要求1至7任一项所述的方法。
PCT/CN2022/111921 2022-08-11 2022-08-11 图像降噪方法、装置及芯片 WO2024031579A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2022/111921 WO2024031579A1 (zh) 2022-08-11 2022-08-11 图像降噪方法、装置及芯片
CN202280004092.3A CN116391202B (zh) 2022-08-11 2022-08-11 图像降噪方法、装置及芯片

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2022/111921 WO2024031579A1 (zh) 2022-08-11 2022-08-11 图像降噪方法、装置及芯片

Publications (1)

Publication Number Publication Date
WO2024031579A1 true WO2024031579A1 (zh) 2024-02-15

Family

ID=86969837

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/111921 WO2024031579A1 (zh) 2022-08-11 2022-08-11 图像降噪方法、装置及芯片

Country Status (2)

Country Link
CN (1) CN116391202B (zh)
WO (1) WO2024031579A1 (zh)

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2011146893A (ja) * 2010-01-14 2011-07-28 Nikon Corp 画像処理装置、撮像装置及び画像処理プログラム
CN102647597A (zh) * 2012-05-02 2012-08-22 华南理工大学 一种基于多边形裁剪dct的jpeg图像压缩方法
CN104217416A (zh) * 2013-05-31 2014-12-17 富士通株式会社 灰度图像处理方法及其装置
CN105144232A (zh) * 2014-03-25 2015-12-09 展讯通信（上海）有限公司 图像去噪方法和系统
CN105279742A (zh) * 2015-11-19 2016-01-27 中国人民解放军国防科学技术大学 一种快速的基于分块噪声能量估计的图像去噪方法
CN108305222A (zh) * 2018-01-04 2018-07-20 浙江大华技术股份有限公司 一种图像的降噪方法、装置、电子设备和存储介质
CN109389567A (zh) * 2018-10-24 2019-02-26 山东大学 一种快速光学成像数据的稀疏滤波方法

Family Cites Families (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7437013B2 (en) * 2003-12-23 2008-10-14 General Instrument Corporation Directional spatial video noise reduction
US7965900B2 (en) * 2007-09-26 2011-06-21 Hewlett-Packard Development Company, L.P. Processing an input image to reduce compression-related artifacts
CN103544703B (zh) * 2013-10-19 2016-12-07 上海理工大学 数字图像拼接检测方法
US10032252B2 (en) * 2015-10-30 2018-07-24 Canon Kabushiki Kaisha Image processing apparatus, image capturing apparatus, image processing method, and non-transitory computer readable storage medium
CN105787893B (zh) * 2016-02-23 2018-11-02 西安电子科技大学 一种基于整数dct变换的图像噪声方差估计方法
WO2018179851A1 (en) * 2017-03-28 2018-10-04 Sharp Kabushiki Kaisha Systems and methods for determining a level of quantization
JP7182907B2 (ja) * 2017-06-15 2022-12-05 ブラックマジック デザイン ピーティーワイ リミテッド カメラの画像データ処理方法およびカメラ
CN111968057A (zh) * 2020-08-24 2020-11-20 浙江大华技术股份有限公司 图像降噪方法、装置、存储介质及电子装置
CN112435182B (zh) * 2020-11-17 2024-05-10 浙江大华技术股份有限公司 图像降噪方法及装置
CN114596229A (zh) * 2022-03-07 2022-06-07 上海义礼健康科技有限公司 一种图像降噪方法及系统

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2011146893A (ja) * 2010-01-14 2011-07-28 Nikon Corp 画像処理装置、撮像装置及び画像処理プログラム
CN102647597A (zh) * 2012-05-02 2012-08-22 华南理工大学 一种基于多边形裁剪dct的jpeg图像压缩方法
CN104217416A (zh) * 2013-05-31 2014-12-17 富士通株式会社 灰度图像处理方法及其装置
CN105144232A (zh) * 2014-03-25 2015-12-09 展讯通信（上海）有限公司 图像去噪方法和系统
CN105279742A (zh) * 2015-11-19 2016-01-27 中国人民解放军国防科学技术大学 一种快速的基于分块噪声能量估计的图像去噪方法
CN108305222A (zh) * 2018-01-04 2018-07-20 浙江大华技术股份有限公司 一种图像的降噪方法、装置、电子设备和存储介质
CN109389567A (zh) * 2018-10-24 2019-02-26 山东大学 一种快速光学成像数据的稀疏滤波方法

Also Published As

Publication number Publication date
CN116391202A (zh) 2023-07-04
CN116391202B (zh) 2024-03-08

Similar Documents

Publication Publication Date Title
US10719918B2 (en) Dynamically determining filtering strength for noise filtering in image processing
KR102163424B1 (ko) 인터리빙된 채널 데이터를 위한 구성가능한 컨볼루션 엔진
US9747514B2 (en) Noise filtering and image sharpening utilizing common spatial support
US9787922B2 (en) Pixel defect preprocessing in an image signal processor
US9479695B2 (en) Generating a high dynamic range image using a temporal filter
KR101183371B1 (ko) 디지털 컬러 이미지 처리 방법
US10467496B2 (en) Temporal filtering of independent color channels in image data
US9911174B2 (en) Multi-rate processing for image data in an image processing pipeline
US9413951B2 (en) Dynamic motion estimation and compensation for temporal filtering
US9514525B2 (en) Temporal filtering for image data using spatial filtering and noise history
US20170070692A1 (en) Correcting pixel defects based on defect history in an image processing pipeline
US9177367B2 (en) Image processing apparatus and image processing method
JP2010218110A (ja) ノイズ低減装置、ノイズ低減方法、ノイズ低減プログラム、記録媒体
EP2791898A2 (en) Method, apparatus and computer program product for capturing images
TWI703872B (zh) 影像色彩還原與增強電路
US8538189B2 (en) Image noise filter and method
WO2019090580A1 (en) System and method for image dynamic range adjusting
WO2022061879A1 (zh) 图像处理方法、装置和系统，计算机可读存储介质
US9858889B2 (en) Color compensation circuit, display apparatus, and color compensation method
CN111524074A (zh) 锐化图像的方法、电子设备及其图像处理器
CN110795659A (zh) 页面背景图的生成方法及其装置
WO2024031579A1 (zh) 图像降噪方法、装置及芯片
US9911177B2 (en) Applying chroma suppression to image data in a scaler of an image processing pipeline
CN113132562A (zh) 镜头阴影校正方法、装置及电子设备
CN116320792A (zh) 多帧图像中增强分辨率和降低噪声的方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22954545

Country of ref document: EP

Kind code of ref document: A1