WO2019091270A1 - 图像增强方法及系统 - Google Patents

图像增强方法及系统 Download PDF

Info

Publication number
WO2019091270A1
WO2019091270A1 PCT/CN2018/110702 CN2018110702W WO2019091270A1 WO 2019091270 A1 WO2019091270 A1 WO 2019091270A1 CN 2018110702 W CN2018110702 W CN 2018110702W WO 2019091270 A1 WO2019091270 A1 WO 2019091270A1
Authority
WO
WIPO (PCT)
Prior art keywords
value
brightness
input
image
pixel point
Prior art date
Application number
PCT/CN2018/110702
Other languages
English (en)
French (fr)
Inventor
孟国庆
汤峰峰
章勇
曹李军
陈卫东
Original Assignee
苏州科达科技股份有限公司
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 苏州科达科技股份有限公司 filed Critical 苏州科达科技股份有限公司
Priority to GB2007219.5A priority Critical patent/GB2581907B/en
Priority to DE112018005567.1T priority patent/DE112018005567T5/de
Publication of WO2019091270A1 publication Critical patent/WO2019091270A1/zh
Priority to US16/872,532 priority patent/US11127122B2/en

Links

Images

Classifications

    • G06T5/94—
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/20—Image enhancement or restoration by the use of local operators
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20004—Adaptive image processing
    • G06T2207/20012—Locally adaptive
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20024—Filtering details

Definitions

  • the present application relates to the field of digital image processing technologies, and in particular, to an image enhancement method and system based on a Retinex image enhancement algorithm.
  • the video capture process is affected by many factors, such as low light and low brightness under night or dark light shooting conditions; light occlusion makes the formed image part bright and the other part is dark, resulting in uneven illumination; reflective or strong light source makes the acquisition
  • the brightness of the image is unevenly distributed, and the details of the highlighted area are blurred. Insufficient illumination or uneven illumination may cause poor subjective effects on images, and it is difficult to meet the needs of people's visual senses. On the other hand, it will have a greater impact on subsequent image processing such as pattern recognition and target tracking. Therefore, image enhancement technology is needed. Enhance the quality by enhancing the uneven illumination image.
  • Image enhancement technology refers to the purpose of enhancing local or overall features according to image quality and different applications using signal processing techniques.
  • the commonly used algorithms are: grayscale transform method, homomorphic filtering method, wavelet transform enhancement, image enhancement algorithm based on Retinex theory.
  • Retinex image enhancement algorithm has many advantages such as color fidelity, detail enhancement and dynamic range compression. It is often combined with other algorithms to achieve better enhancement.
  • Multi-scale Retinex MSR, Multi-Scale Retinex
  • MSR Multi-Scales Retinex with Color Restoration
  • these two algorithms are for images with multiple light sources. The treatment effect is not good, and it is easy to produce halo artifacts. For darker images, the output image brightness is too high, although it looks more clear, but it has not been consistent with the real scene; especially when the image quality is poor, the output image is accompanied by noise, block effect, and the effect is worse.
  • an object of the present application is to provide an image enhancement method and system, which controls the output brightness so that the output image can be improved and conformed to the real environment.
  • An embodiment of the present application provides an image enhancement method, including the following steps:
  • An image of each pixel in the input image is enhanced according to the second brightness gain value to obtain an output image.
  • the embodiment of the present application further provides an image enhancement system, which is used to implement the image enhancement method, and the system includes:
  • the input brightness obtaining module is configured to obtain an input brightness value of each pixel in the input image
  • a brightness gain calculation module configured to: calculate a logarithmic mean value of the input brightness value; and process the input image based on a Retinex image enhancement algorithm to obtain a first brightness gain value of each pixel point;
  • a brightness control module configured to: calculate a brightness gain control factor according to a logarithmic mean value of the input brightness value, and calculate a second brightness gain value of each pixel point according to the brightness gain control factor;
  • an image enhancement module configured to enhance each pixel point in the input image according to the second brightness gain value to obtain an output image.
  • the present application uses the Retinex image enhancement algorithm to enhance the input image, and by using the brightness gain control factor, when the image brightness is low, the first brightness gain value is greatly compressed, and when the image brightness is large, the first brightness The gain value is compressed less, so that the brightness gain of the dark image is reduced; thus, the brightness of the output image is effectively controlled, so that the brightness of the output image can be improved and conformed to the real environment. .
  • FIG. 2 is a flow chart of calculating a first brightness gain value according to an embodiment of the present application
  • FIG. 3 is a flowchart of an image enhancement method according to an embodiment of the present application.
  • FIG. 4 is a schematic structural diagram of an image enhancement system according to an embodiment of the present application.
  • FIG. 5 is a schematic structural diagram of a luminance gain calculation module according to an embodiment of the present application.
  • the present application provides an image enhancement method, including the following steps:
  • S100 Acquire an input luminance value of each pixel in the input image, and calculate a logarithmic average value of the input luminance value
  • S200 processing an input image based on a Retinex image enhancement algorithm to obtain a first brightness gain value of each pixel point;
  • S300 Calculating, according to a logarithmic mean value of the input luminance values, a brightness gain control factor, where the brightness gain control factor is positively correlated with a logarithmic mean value of the input brightness value;
  • S400 Calculate, according to the brightness gain control factor and the first brightness gain value, a second brightness gain value of each pixel point;
  • S500 Enhance each pixel point in the input image according to the second brightness gain value to obtain an enhanced image.
  • the present application uses the Retinex image enhancement algorithm to enhance the input image and control the brightness gain by using a brightness gain control factor.
  • the brightness gain control factor is determined by the logarithmic average of the input brightness values, which can well reflect the overall brightness level of the input image, and control the brightness gain according to the overall brightness level of the input image, so that the brightness of the output image can be obtained. Improve and match the real environment.
  • the Retinex image enhancement algorithm may adopt a global adaptive algorithm, a local adaptive algorithm, or a combination of a global adaptive algorithm and a local adaptive algorithm, or other image enhancement based on Retinex.
  • the algorithms are all within the scope of protection of the present application.
  • the invention will be specifically described below with a specific embodiment.
  • the input image is preprocessed using a global tone mapping technique; then, the globally processed output is processed using local tone mapping; The output brightness of the Retinex algorithm is normalized, and the brightness of the Retinex algorithm is processed by the brightness gain control factor. Finally, the output image is obtained from the processed brightness value and the input chromaticity value.
  • image enhancement includes two parts of global processing and local processing.
  • step S200 may include the following steps:
  • S201 processing the input image according to a global adaptive algorithm to obtain a global output luminance value of each pixel point;
  • S202 Processing, according to a local adaptive algorithm, the global output luminance value, to obtain an output luminance value of each pixel point;
  • S203 Calculate a first brightness gain value of each pixel point according to the output brightness value and the input brightness value.
  • the local adaptive algorithm of step S202 is improved, and the Gaussian filter in the visual algorithm is replaced with a guide filter to remove the halo.
  • a contrast enhancement factor based on the scene brightness value is added to the local processing algorithm to control the contrast of the output image.
  • this embodiment also introduces the nonlinear strength of the logarithmic function to handle the adaptive nonlinear offset.
  • the image enhancement method of this embodiment may include the following steps:
  • the image enhancement method of this embodiment not only the contrast and brightness of the output image can be enhanced, but also halo artifacts can be eliminated.
  • the present application can perform adaptive enhancement for different types of images, ensuring real-time image enhancement.
  • L w (x, y) is the input luminance value
  • N is the total number of pixels in the input image
  • ⁇ is the second preset control value
  • L w (x, y) is the input luminance value of the pixel (x, y) .
  • L g (x, y) is the global output luminance value of the pixel point (x, y)
  • L w (x, y) is the input luminance value of the pixel point (x, y)
  • L wmax is the maximum of L w value
  • a local adaptive algorithm based on visual theory is used.
  • a pilot filter is used instead of the dual-edge filter in the traditional theory, both of which are filters with edge-preserving characteristics, and the guided filter has better performance near the edge.
  • the local adaptive formula is:
  • L g (x, y) is the global output luminance value of the pixel point (x, y)
  • H g (x, y) directs the output value of the filter
  • L l (x, y) is the pixel point (x, y) The local output brightness value.
  • the output value of the pilot filter is calculated according to the following formula:
  • ⁇ x , ⁇ y are neighborhood pixel coordinates
  • ⁇ (x, y) is a partial rectangular window with a radius r at the pixel point (x, y)
  • is a pixel within ⁇ (x, y)
  • a( ⁇ x , ⁇ y ) and b( ⁇ x , ⁇ y ) are linear coefficients, respectively, and can be calculated according to the following two formulas:
  • ⁇ ( ⁇ x , ⁇ y ) and ⁇ 2 ( ⁇ x , ⁇ y ) are the mean and variance of L g in ⁇ ( ⁇ x , ⁇ y ), respectively, ⁇ 2 ( ⁇ x , ⁇ y ) is In the mean of ⁇ ( ⁇ x , ⁇ y ), ⁇ is the regularization parameter.
  • the contrast enhancement factor is as follows:
  • Another factor is the adaptive nonlinear offset factor, which is:
  • L gmax is the maximum value of L g and ⁇ is the nonlinear control parameter. Is the logarithmic mean of L g .
  • L out (x, y) is the optimized local output luminance value of the pixel point (x, y).
  • L' out (x, y) is a normalized luminance value of L out (x, y), and L outmax and L outmin are respectively a maximum value and a minimum value of L out ;
  • the first brightness gain value is calculated according to the following formula:
  • G(x, y) is the first brightness gain value of the pixel point (x, y).
  • G(x, y) is directly used as the gain output image, the output luminance of the image cannot be effectively controlled.
  • the output image is too bright, although it looks sharper, but it does not match the real scene; especially when the image quality is poor, the output image is accompanied by noise, and the effect is worse.
  • a luminance gain control factor ⁇ is designed. ⁇ is adjusted on the basis of G(x, y).
  • ⁇ is the brightness gain control factor
  • G(x, y) and G'(x, y) are the first brightness gain value and the second brightness gain value of the pixel point (x, y), respectively.
  • the brightness gain control factor is calculated according to the following formula:
  • ⁇ is the luminance gain control factor
  • the first preset control value may be selected according to an empirical value, for example, selected as 0.5, 0.6, etc., to obtain an optimal brightness gain control effect.
  • the brightness gain control factor is 1, which has no effect on the first brightness gain value. Therefore, The value of the value is further preferably selected within the range of (0, 1).
  • ⁇ is the brightness gain control factor
  • the logarithmic average of the input brightness is input, which better reflects the brightness of the original image. Is the logarithmic mean of the input brightness values, and So, ⁇ [0,1].
  • the purpose of this is that when the brightness of the image is low, the first brightness gain value is compressed a lot, and when the brightness of the image is large, the first brightness gain value is compressed less, thereby controlling the output brightness to be effective.
  • the purpose of control is that when the brightness of the image is low, the first brightness gain value is compressed a lot, and when the brightness of the image is large, the first brightness gain value is compressed less, thereby controlling the output brightness to be effective. The purpose of control.
  • RGB1(x,y) RGB0(x,y)*G'(x,y)
  • G'(x, y) is the second luminance gain value of the pixel point (x, y)
  • RGB1(x, y) is the RGB value of the pixel point (x, y) in the output image
  • RGB0 (x, y) ) is the RGB value of the pixel point (x, y) in the input image.
  • an embodiment of the present application further provides an image enhancement system for implementing the image enhancement method described above, where the system includes an input brightness acquisition module 100, a brightness gain calculation module 200, a brightness control module 300, and image enhancement.
  • Module 400 wherein:
  • the brightness acquisition module 100 acquires an input brightness value of each pixel in the input image; the brightness gain calculation module 200 calculates a log average of the input brightness value; and processes the input image based on a Retinex image enhancement algorithm to obtain a first brightness gain value of each pixel; the brightness control module 300 calculates a brightness gain control factor according to a logarithmic mean value of the input brightness value, and calculates a second color of each pixel point according to the brightness gain control factor a brightness gain value; the image enhancement module 400 enhances each pixel point in the input image according to the second brightness gain value to obtain an output image.
  • the input image is enhanced using a Retinex image enhancement algorithm, and the brightness gain is controlled by using a brightness gain control factor.
  • the brightness gain control factor is determined by the logarithmic average of the input brightness values, which can well reflect the overall brightness level of the input image, and control the brightness gain according to the overall brightness level of the input image, so that the brightness of the output image can be obtained. Improve and match the real environment.
  • each of the modules can implement the fixed-point processing of the algorithm, and can be implemented in an FPGA (Field-Programmable Gate Array).
  • the luminance gain calculation module 200 may further include a global adaptation unit 201 that processes the image by using a global adaptive algorithm to obtain a global output luminance value; and the filtering unit 202 uses the bootstrap filter to globally output the luminance value.
  • Filtering; contrast control unit 203 calculates a contrast enhancement factor, and introduces a contrast enhancement factor into the local adaptive formula;
  • nonlinear offset unit 204 calculates an adaptive nonlinear offset factor, and applies an adaptive nonlinear offset The factor is introduced into the local adaptive formula.
  • the image enhancement method of this embodiment not only the brightness of the output image can be effectively controlled, but also the aperture effect of the image enhancement can be reduced by the guidance filter, and the global contrast of the image can be enhanced by the contrast enhancement factor through adaptive nonlinearity.
  • the offset factor maintains the accuracy of the local adaptive algorithm.
  • the present application uses the Retinex image enhancement algorithm to enhance the input image, and by using the brightness gain control factor, when the image brightness is low, the first brightness gain value is greatly compressed, and when the image brightness is large, the first brightness The gain value is compressed less, so that the brightness gain of the dark image is reduced; thus, the brightness of the output image is effectively controlled, so that the brightness of the output image can be improved and conformed to the real environment. .

Abstract

本申请提供了一种图像增强方法及系统，方法包括获取输入亮度值，并计算输入亮度值的对数平均值；基于Retinex图像增强算法处理输入图像，得到第一亮度增益值；根据输入亮度值的对数平均值，计算得到亮度增益控制因子；根据亮度增益控制因子和第一亮度增益值，计算得到第二亮度增益值；根据第二亮度增益值增强输入图像得到输出图像。本申请采用Retinex图像增强算法对输入图像进行增强，并且通过采用亮度增益控制因子，当图像亮度较低时，第一亮度增益值会被较大的压缩，当图像亮度较大时，第一亮度增益值被较小的压缩，这样就可以使暗的图像的亮度增益变小；实现了对输出图像的亮度进行有效控制，使输出图像的亮度既能得到提升，又能与真实环境相符合。

Description

图像增强方法及系统 技术领域
本申请涉及数字图像处理技术领域，尤其涉及一种基于Retinex图像增强算法的图像增强方法及系统。
背景技术
视频采集过程受诸多因素影响，如夜间或者暗光拍摄条件下表现为光照不足、亮度偏低；光线遮挡使得形成的图像一部分明亮另一部分较暗，造成光照不均；反光或强光源使得获取的图像亮度分布不均匀，高亮区域细节模糊。光照不足或者光照不均一方面会造成图像主观效果不佳，难以满足人们视觉感官的需要，另一方面对于后续的图像处理比如模式识别、目标跟踪都会造成较大影响，因此需要采用图像增强技术对光照不均图像进行增强处理提高质量。
图像增强技术指的是根据图像质量情况和不同的应用釆用信号处理技术手段达到增强局部或者整体特征的目的。针对光照不均匀图像的增强处理，常常采用的算法主要有：灰度变换方法、同态滤波方法、小波变换增强、基于Retinex理论的图像增强算法等。其中，Retinex图像增强算法具有颜色保真、细节增强和动态范围压缩等多个方面的优势，它常常与其他算法结合使用可以达到更好的增强效果。多尺度Retinex(MSR，Multi-Scale Retinex)和带色彩恢复的MSR(MSRCR，Multi-Scales Retinex with Color Restoration)是其中得到广泛应用的两种方法，然而这两种算法对有多个光源的图像处理效果不好，且易产生光晕伪影。对于较暗的图像，输出图像亮度过高，虽然看起来更加清晰，但已经与现实场景不相符合；特别是图像质量较差时，输出的图像伴有噪声，出现块效应，效果变差。
发明内容
针对现有技术中的问题，本申请的目的在于提供一种图像增强方法及系统，对输出亮度进行控制，使输出图像既能得到提升，又能与真实环境相符合。
本申请实施例提供一种图像增强方法，包括如下步骤：
获取输入图像中各个像素点的输入亮度值，并计算所述输入亮度值的对数平均值；
基于Retinex图像增强算法处理所述输入图像，得到各个像素点的第一亮度增益值；
根据所述输入亮度值的对数平均值，计算得到亮度增益控制因子，所述亮度增益控制因子与所述输入亮度值的对数平均值正相关；
根据所述亮度增益控制因子和所述第一亮度增益值，计算得到各个像素点的第二亮度增益值；
根据所述第二亮度增益值增强所述输入图像中的各个像素点，得到输出图像。
本申请实施例还提供一种图像增强系统，用于实现所述的图像增强方法，所述系统包括：
输入亮度获取模块，用于获取输入图像中各个像素点的输入亮度值；
亮度增益计算模块，用于：计算所述输入亮度值的对数平均值；基于Retinex图像增强算法处理所述输入图像，得到各个像素点的第一亮度增益值；
亮度控制模块，用于：根据所述输入亮度值的对数平均值计算得到亮度增益控制因子，以及根据所述亮度增益控制因子计算得到各个像素点的第二亮度增益值；
图像增强模块，用于根据所述第二亮度增益值增强所述输入图像中的各个像素点，得到输出图像。
本申请所提供的图像增强方法具有下列优点：
本申请采用Retinex图像增强算法对输入图像进行增强，并且通过采用亮度增益控制因子，当图像亮度较低时，第一亮度增益值会被较大的压缩，当图像亮度较大时，第一亮度增益值被较小的压缩，这样就可以使暗 的图像的亮度增益变小；从而实现了对输出图像的亮度进行有效控制，使输出图像的亮度既能得到提升，又能与真实环境相符合。
附图说明
通过阅读参照以下附图对非限制性实施例所作的详细描述，本申请的其它特征、目的和优点将会变得更明显。
图1是本申请的图像增强方法的流程图；
图2是本申请一实施例的计算第一亮度增益值的流程图；
图3是本申请一实施例的图像增强方法的流程图；
图4是本申请一实施例的图像增强系统的结构示意图；
图5是本申请一实施例的亮度增益计算模块的结构示意图。
具体实施方式
现在将参考附图更全面地描述示例实施方式。然而，示例实施方式能够以多种形式实施，且不应被理解为限于在此阐述的实施方式；相反，提供这些实施方式使得本申请将全面和完整，并将示例实施方式的构思全面地传达给本领域的技术人员。在图中相同的附图标记表示相同或类似的结构，因而将省略对它们的重复描述。
如图1所示，本申请提供了一种图像增强方法，包括如下步骤：
S100：获取输入图像中各个像素点的输入亮度值，并计算输入亮度值的对数平均值；
S200：基于Retinex图像增强算法处理输入图像，得到各个像素点的第一亮度增益值；
S300：根据输入亮度值的对数平均值，计算得到亮度增益控制因子，所述亮度增益控制因子与所述输入亮度值的对数平均值正相关；
S400：根据亮度增益控制因子和第一亮度增益值，计算得到各个像素点的第二亮度增益值；
S500：根据第二亮度增益值增强输入图像中的各个像素点，得到增强图像。
本申请采用Retinex图像增强算法对输入图像进行增强，并且通过采 用亮度增益控制因子对亮度增益进行控制。亮度增益控制因子由输入亮度值的对数平均值来决定，可以很好地反映输入图像的整体亮度水平，根据输入图像的整体亮度水平对亮度增益进行控制，从而使输出图像的亮度既能得到提升，又能与真实环境相符合。
在本申请中，Retinex图像增强算法可以采用全局自适应算法，也可以采用局部自适应算法，也可以采用全局自适应算法和局部自适应算法结合的方式，另外也可以采用基于Retinex的其他图像增强算法，均属于本申请的保护范围之内。
下面以一个具体实施例来具体说明该发明，该实施例中，以Retinex理论为基础，首先，使用全局色调映射技术预处理输入图像；然后，使用局部色调映射处理全局处理后的输出；然后，对Retinex算法输出亮度做归一化，采用亮度增益控制因子对Retinex算法输出亮度进行处理；最后从处理后的亮度值和输入色度值中获得输出图像。
如图2所示，在该实施例中，图像增强包括全局处理和局部处理两部分。具体地，上述步骤S200可以包括如下步骤：
S201：基于全局自适应算法处理所述输入图像，得到各个像素点的全局输出亮度值；
S202：基于局部自适应算法处理所述全局输出亮度值，得到各个像素点的输出亮度值；
S203：根据所述输出亮度值和输入亮度值，计算得到各个像素点的第一亮度增益值。
另外，在该实施例中，对步骤S202局部自适应算法进行了改进，将视觉算法中的高斯滤波器使用一个引导滤波器取代，能够去除光晕。局部处理算法中增加了一种基于场景亮度值的对比度增强因子，控制输出图像的对比度。此外，该实施例还引入了对数函数的非线性强度来处理自适应非线性偏移。
如图3所示，该实施例的图像增强方法可以包括如下步骤：
S1：输入图像数据，计算输入图像的输入亮度值L w和对数平均值
Figure PCTCN2018110702-appb-000001
S2：输入亮度值L w代入人类视觉系统感知亮度的对数函数，得到全局输出亮度值L g；
S3：对全局输出亮度值L g进行引导滤波，得到引导滤波器的输出H g，确定对比度控制参数η和非线性控制参数λ，计算局部亮度输出值L l；
S4：确定亮度增益控制因子τ，通过输入亮度值L w、对数平均值
Figure PCTCN2018110702-appb-000002
局部亮度输出值L l和亮度增益控制因子τ计算出每个像素点的亮度增益值；
S5：输入图像的三通道的像素值分别乘以亮度增益值，构造一个新的图像文件。
因此，采用该实施例的图像增强方法，不仅可以使输出图像的对比度和亮度得到增强，还能消除光晕伪影。本申请可以针对不同类型图像进行自适应增强，保证了图像增强的实时性。
下面分别对各个步骤进行详细介绍：
S1、输入图像数据，计算输入图像的输入亮度值L w，并根据如下公式计算对数平均值
Figure PCTCN2018110702-appb-000003
Figure PCTCN2018110702-appb-000004
其中，L w(x,y)为输入亮度值，
Figure PCTCN2018110702-appb-000005
为L w的对数平均值，N为所述输入图像中的像素点总数量，δ为第二预设控制值，L w(x,y)为像素点(x,y)的输入亮度值。
S2、全局自适应算法计算全局输出亮度值L g
全局自适应与早期的人类视觉系统类似，人类视觉系统感知亮度的近似为一个对数函数。因此视觉全局亮度转换公式：
Figure PCTCN2018110702-appb-000006
其中，L g(x,y)为像素点(x,y)的全局输出亮度值，L w(x,y)为像素点(x,y)的输入亮度值，L wmax为L w的最大值，
Figure PCTCN2018110702-appb-000007
为L w的对数平均值。
S3、局部自适应算法计算局部输出亮度值L l
在全局自适应处理之后，使用基于视觉理论的局部自适应算法。在本案例中，使用引导滤波器代替传统理论中的双边缘滤波器，两者都是具有保边特性的滤波器，而引导滤波器在靠近边缘处，有更好的性能。局部自适应公式为：
L l(x,y)＝log(L g(x,y))-log(H g(x,y))
其中，L g(x,y)为像素点(x,y)的全局输出亮度值，H g(x,y)引导滤波器的输出值，L l(x,y)为像素点(x,y)的局部输出亮度值。
根据如下公式计算引导滤波器的输出值：
Figure PCTCN2018110702-appb-000008
其中，ξ x,ξ y为邻域像素坐标，ω(x,y)为在像素点(x,y)处一个半径为r局部矩形窗口，|ω|为在ω(x,y)内像素点的数量，a(ξ x,ξ y)和b(ξ x,ξ y)分别为线性系数，且可分别根据如下两个公式计算得到：
Figure PCTCN2018110702-appb-000009
b(ξ x,ξ y)＝μ(ξ x,ξ y)-a(ξ x,ξ y)μ(ξ x,ξ y)
其中，μ(ξ x,ξ y)和σ 2(ξ x,ξ y)分别为L g在ω(ξ x,ξ y)内的均值和方差，μ 2(ξ x,ξ y)为
Figure PCTCN2018110702-appb-000010
在ω(ξ x,ξ y)内的均值，ε为正则化参数。
在滤波后，光晕效应会明显降低，但是全局的对比度较低。为了防止滤波器引起的扁平效应，提高算法的性能，引入两个重要因子。一个是对比度增强因子，公式如下：
Figure PCTCN2018110702-appb-000011
另一个因子是自适应非线性偏移因子，公式为：
Figure PCTCN2018110702-appb-000012
其中，L gmax为L g的最大值，λ为非线性控制参数，
Figure PCTCN2018110702-appb-000013
为L g的对数平均值。
通过以上个部分因子的计算，最终的局部自适应公式为：
Figure PCTCN2018110702-appb-000014
其中，L out(x,y)为像素点(x,y)的优化后的局部输出亮度值。
S4、确定亮度增益控制因子τ
对L out归一化，使它与输入亮度在同一个范围内。归一化方法可以采 用如下公式：
Figure PCTCN2018110702-appb-000015
其中，L’ out(x,y)为L out(x,y)的归一化亮度值，L outmax和L outmin分别为L out的最大值和最小值；
根据如下公式计算所述第一亮度增益值：
Figure PCTCN2018110702-appb-000016
其中，G(x,y)为像素点(x,y)的第一亮度增益值。
如果直接使用G(x,y)作为增益输出图像，不能有效地控制图像的输出亮度。对于较暗的图像，输出图像亮度过高，虽然看起来更加清晰，但已经与现实场景不相符合；特别是图像质量较差时，输出的图像伴有噪声，看上去效果变差。
为了解决上述问题，需要对输出亮度进行控制，使输出图像既能得到提升，又能与真实环境相符合。那么，我们希望在图像亮度变小时，G(x,y)也变小，且梯度变大。
由此，设计了一个亮度增益控制因子τ。τ是在是在G(x,y)的基础上做调整。
G′(x,y)＝(G(x,y)) τ
其中，τ为所述亮度增益控制因子，G(x,y)和G′(x,y)分别为像素点(x,y)的第一亮度增益值和第二亮度增益值。
根据如下公式计算得到所述亮度增益控制因子：
Figure PCTCN2018110702-appb-000017
其中，
Figure PCTCN2018110702-appb-000018
为所述输入亮度值的对数平均值，τ为所述亮度增益控制因子，
Figure PCTCN2018110702-appb-000019
为第一预设控制值，且
Figure PCTCN2018110702-appb-000020
在实际应用中，第一预设控制值可以根据经验值选择，例如选择为0.5，0.6等等，以获取最佳的亮度增益控制效果。另外，由于
Figure PCTCN2018110702-appb-000021
取值为0时，亮度增益控制因子为1，对第一亮度增益值无影响，因此，
Figure PCTCN2018110702-appb-000022
的值进一步优选在(0,1]的范围内选取。
由于τ为亮度增益控制因子是以输入亮度的对数平均值为输入，这样更能体现出原图像的亮度。
Figure PCTCN2018110702-appb-000023
是输入亮度值的对数平均值，且
Figure PCTCN2018110702-appb-000024
所以，τ∈[0,1]。
这样做的目的是当图像亮度较低时，第一亮度增益值会被较大的压缩，当图像亮度较大时，第一亮度增益值被较小的压缩，从而控制了输出亮度达到了有效控制的目的。
S5：根据如下公式增强输入图像中的各个像素点，得到输出图像：
RGB1(x,y)＝RGB0(x,y)*G′(x,y)
其中，G′(x,y)为像素点(x,y)的第二亮度增益值，RGB1(x,y)为输出图像中像素点(x,y)的RGB值，RGB0(x,y)为输入图像中像素点(x,y)的RGB值。
如图4所示，本申请实施例还提供一种图像增强系统，用于实现上述的图像增强方法，所述系统包括输入亮度获取模块100、亮度增益计算模块200、亮度控制模块300和图像增强模块400，其中：
所述亮度获取模块100获取输入图像中各个像素点的输入亮度值；所述亮度增益计算模块200计算所述输入亮度值的对数平均值；以及基于Retinex图像增强算法处理所述输入图像，得到各个像素点的第一亮度增益值；所述亮度控制模块300根据所述输入亮度值的对数平均值计算得到亮度增益控制因子，以及根据所述亮度增益控制因子计算得到各个像素点的第二亮度增益值；所述图像增强模块400根据所述第二亮度增益值增强所述输入图像中的各个像素点，得到输出图像。
该实施例中，采用Retinex图像增强算法对输入图像进行增强，并且通过采用亮度增益控制因子对亮度增益进行控制。亮度增益控制因子由输入亮度值的对数平均值来决定，可以很好地反映输入图像的整体亮度水平，根据输入图像的整体亮度水平对亮度增益进行控制，从而使输出图像的亮度既能得到提升，又能与真实环境相符合。
其中各个模块均可以实现算法的定点化处理，均可以在FPGA(Field－Programmable Gate Array，现场可编程门阵列)中实现，各个模块的具体操作步骤可以参见上述图像增强方法的实施例。具体地，如图5所示，亮度增益计算模块200还可以包括全局自适应单元201，采用全局自适应算法处理图像，得到全局输出亮度值；滤波单元202，采用引导滤波器对全局输出亮度值进行滤波；对比度控制单元203，计算对比度增强因子， 并且将对比度增强因子引入到局部自适应公式中；非线性偏移单元204，计算自适应非线性偏移因子，并且将自适应非线性偏移因子引入到局部自适应公式中。因此，采用该实施例的图像增强方法，不仅可以对输出图像的亮度进行有效控制，而且可以通过引导滤波器降低图像增强的光圈效应，通过对比度增强因子增强图像的全局对比度，通过自适应非线性偏移因子保持局部自适应算法的准确性。
本申请所提供的图像增强方法及系统具有下列优点：
本申请采用Retinex图像增强算法对输入图像进行增强，并且通过采用亮度增益控制因子，当图像亮度较低时，第一亮度增益值会被较大的压缩，当图像亮度较大时，第一亮度增益值被较小的压缩，这样就可以使暗的图像的亮度增益变小；从而实现了对输出图像的亮度进行有效控制，使输出图像的亮度既能得到提升，又能与真实环境相符合。
以上内容是结合具体的优选实施方式对本申请所作的进一步详细说明，不能认定本申请的具体实施只局限于这些说明。对于本申请所属技术领域的普通技术人员来说，在不脱离本申请构思的前提下，还可以做出若干简单推演或替换，都应当视为属于本申请的保护范围。

Claims (10)

  1. 一种图像增强方法，其特征在于，包括如下步骤：
    获取输入图像中各个像素点的输入亮度值，并计算所述输入亮度值的对数平均值；
    基于Retinex图像增强算法处理所述输入图像，得到各个像素点的第一亮度增益值；
    根据所述输入亮度值的对数平均值，计算得到亮度增益控制因子，所述亮度增益控制因子与所述输入亮度值的对数平均值正相关；
    根据所述亮度增益控制因子和所述第一亮度增益值，计算得到各个像素点的第二亮度增益值；
    根据所述第二亮度增益值增强所述输入图像中的各个像素点，得到输出图像。
  2. 根据权利要求1所述的图像增强方法，其特征在于，根据如下公式计算得到各个像素点的第二亮度增益值：
    G′(x,y)＝(G(x,y)) τ
    其中，τ为所述亮度增益控制因子，G(x,y)和G′(x,y)分别为像素点(x,y)的第一亮度增益值和第二亮度增益值。
  3. 根据权利要求2所述的图像增强方法，其特征在于，根据如下公式计算得到所述亮度增益控制因子：
    Figure PCTCN2018110702-appb-100001
    其中，
    Figure PCTCN2018110702-appb-100002
    为所述输入亮度值的对数平均值，τ为所述亮度增益控制因子，
    Figure PCTCN2018110702-appb-100003
    为第一预设控制值，且
    Figure PCTCN2018110702-appb-100004
  4. 根据权利要求1所述的图像增强方法，其特征在于，根据如下公式计算所述输入亮度值的对数平均值：
    Figure PCTCN2018110702-appb-100005
    其中，L w(x,y)为像素点(x,y)的输入亮度值，
    Figure PCTCN2018110702-appb-100006
    为L w的对数平均值，N为所述输入图像中的像素点总数量，δ为第二预设控制值，L w(x,y)为像 素点(x,y)的输入亮度值。
  5. 根据权利要求1所述的图像增强方法，其特征在于，基于Retinex图像增强算法处理所述输入图像，包括如下步骤：
    基于全局自适应算法处理所述输入图像，得到各个像素点的全局输出亮度值；
    基于局部自适应算法处理所述全局输出亮度值，得到各个像素点的输出亮度值；
    根据所述输出亮度值和输入亮度值，计算得到各个像素点的第一亮度增益值。
  6. 根据权利要求5所述的图像增强方法，其特征在于，所述基于全局自适应算法处理所述输入图像，包括如下步骤：
    根据如下公式计算各个像素点的全局输出亮度值：
    Figure PCTCN2018110702-appb-100007
    其中，L g(x,y)为像素点(x,y)的全局输出亮度值，L w(x,y)为像素点(x,y)的输入亮度值，L wmax为L w的最大值，
    Figure PCTCN2018110702-appb-100008
    为L w的对数平均值。
  7. 根据权利要求5所述的图像增强方法，其特征在于，所述基于局部自适应算法处理所述全局输出亮度值，包括如下步骤：
    根据各个像素点的全局输出亮度值，计算引导滤波器的输出值；
    根据如下公式计算各个像素点的局部输出亮度值：
    L l(x,y)＝log(L g(x,y))-log(H g(x,y))
    其中，L g(x,y)为像素点(x,y)的全局输出亮度值，H g(x,y)引导滤波器的输出值，L l(x,y)为像素点(x,y)的局部输出亮度值；
    将所述局部输出亮度值作为所述各个像素点的输出亮度值。
  8. 根据权利要求5所述的图像增强方法，其特征在于，所述基于局部自适应算法处理所述全局输出亮度值，包括如下步骤：
    根据各个像素点的全局输出亮度值，计算引导滤波器的输出值；
    根据如下公式计算各个像素点的局部输出亮度值：
    Figure PCTCN2018110702-appb-100009
    其中，L g(x,y)为像素点(x,y)的全局输出亮度值，H g(x,y)引导滤波器的输出值，L out(x,y)为像素点(x,y)的局部输出亮度值，α(x,y)为像素点(x,y)的对比度增强因子，β为自适应非线性偏移因子；且α(x,y)和β可分别由如下两个公式计算得到：
    Figure PCTCN2018110702-appb-100010
    Figure PCTCN2018110702-appb-100011
    其中，L gmax为L g的最大值，η为对比度控制参数，λ为非线性控制参数，
    Figure PCTCN2018110702-appb-100012
    为L g的对数平均值；
    将所述局部输出亮度值作为所述各个像素点的输出亮度值。
  9. 根据权利要求7或8所述的图像增强方法，其特征在于，根据如下公式计算引导滤波器的输出值：
    Figure PCTCN2018110702-appb-100013
    其中，ξ x,ξ y为邻域像素坐标，ω(x,y)为在像素点(x,y)处一个半径为r局部矩形窗口，|ω|为在ω(x,y)内像素点的数量，a(ξ x,ξ y)和b(ξ x,ξ y)分别为线性系数，且可分别根据如下两个公式计算得到：
    Figure PCTCN2018110702-appb-100014
    b(ξ x,ξ y)＝μ(ξ x,ξ y)-a(ξ x,ξ y)μ(ξ x,ξ y)
    其中，μ(ξ x,ξ y)和σ 2(ξ x,ξ y)分别为L g在ω(ξ x,ξ y)内的均值和方差，μ 2(ξ x,ξ y)为
    Figure PCTCN2018110702-appb-100015
    在ω(ξ x,ξ y)内的均值，ε为正则化参数。
  10. 一种图像增强系统，其特征在于，包括：
    输入亮度获取模块，用于获取输入图像中各个像素点的输入亮度值；
    亮度增益计算模块，用于：计算所述输入亮度值的对数平均值；基于Retinex图像增强算法处理所述输入图像，得到各个像素点的第一亮度增益值；
    亮度控制模块，用于：根据所述输入亮度值的对数平均值计算得到亮度增益控制因子，以及根据所述亮度增益控制因子计算得到各个像素点的 第二亮度增益值；
    图像增强模块，用于根据所述第二亮度增益值增强所述输入图像中的各个像素点，得到输出图像。
PCT/CN2018/110702 2017-11-13 2018-10-17 图像增强方法及系统 WO2019091270A1 (zh)

Priority Applications (3)

Application Number Priority Date Filing Date Title
GB2007219.5A GB2581907B (en) 2017-11-13 2018-10-17 Image enhancement method and system
DE112018005567.1T DE112018005567T5 (de) 2017-11-13 2018-10-17 Verfahren und System zur Bildverstärkung
US16/872,532 US11127122B2 (en) 2017-11-13 2020-05-12 Image enhancement method and system

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201711116454.5 2017-11-13
CN201711116454.5A CN107730475B (zh) 2017-11-13 2017-11-13 图像增强方法及系统

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US16/872,532 Continuation US11127122B2 (en) 2017-11-13 2020-05-12 Image enhancement method and system

Publications (1)

Publication Number Publication Date
WO2019091270A1 true WO2019091270A1 (zh) 2019-05-16

Family

ID=61215282

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/110702 WO2019091270A1 (zh) 2017-11-13 2018-10-17 图像增强方法及系统

Country Status (5)

Country Link
US (1) US11127122B2 (zh)
CN (1) CN107730475B (zh)
DE (1) DE112018005567T5 (zh)
GB (1) GB2581907B (zh)
WO (1) WO2019091270A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114511462A (zh) * 2022-02-11 2022-05-17 电子科技大学 一种视觉图像增强方法
CN116245771A (zh) * 2023-03-22 2023-06-09 新光维医疗科技(苏州)股份有限公司 一种基于图像二次增强的内窥镜图像亮度自动增强方法及系统

Families Citing this family (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107730475B (zh) * 2017-11-13 2019-08-23 苏州科达科技股份有限公司 图像增强方法及系统
CN109325922B (zh) * 2018-09-12 2022-03-11 深圳开阳电子股份有限公司 一种图像自适应增强方法、装置及图像处理设备
CN109360221A (zh) * 2018-10-25 2019-02-19 重庆市武隆区农业技术推广中心 一种适宜收获山区马铃薯的小型机
CN109712097B (zh) * 2019-01-04 2021-04-30 Oppo广东移动通信有限公司 图像处理方法、装置、存储介质及电子设备
US11263987B2 (en) 2019-04-04 2022-03-01 Hisense Visual Technology Co., Ltd. Method of enhancing contrast and a dual-cell display apparatus
CN111785222B (zh) * 2019-04-04 2021-05-28 海信视像科技股份有限公司 一种对比度提升算法及双面板显示装置
CN110473158A (zh) * 2019-08-14 2019-11-19 上海世茂物联网科技有限公司 一种车牌图像亮度的处理方法、装置及设备
CN110807748A (zh) * 2019-11-06 2020-02-18 广州市皓品信息科技有限公司 一种基于高动态范围的新色调映射图像增强方法
CN111833274B (zh) * 2020-07-17 2023-12-08 中国电子技术标准化研究院 一种图像增强方法及计算设备
CN112288652A (zh) * 2020-10-30 2021-01-29 西安科技大学 一种基于PSO寻优的引导滤波-Retinex低照度图像增强方法
CN112396563B (zh) * 2020-11-17 2023-07-25 北京嘉恒中自图像技术有限公司 一种基于局部亮度进行自适应增益调制的图像增强方法
CN112508809B (zh) * 2020-11-27 2023-04-21 湖南傲英创视信息科技有限公司 低照度图像/视频增强方法及系统
CN112581400A (zh) * 2020-12-22 2021-03-30 安徽圭目机器人有限公司 基于高斯标准差和对比度调谐图像增强方法
CN113947553B (zh) * 2021-12-20 2022-03-18 山东信通电子股份有限公司 一种图像亮度增强方法及设备
CN115512251A (zh) * 2022-11-04 2022-12-23 深圳市瓴鹰智能科技有限公司 基于双分支渐进式特征增强的无人机低照度目标跟踪方法
CN115841431B (zh) * 2023-02-06 2023-05-05 淄博市临淄区交通运输事业服务中心 基于视频监控的交通运输图像增强方法
CN117274113B (zh) * 2023-11-20 2024-02-02 山东九思新材料科技有限责任公司 基于图像增强的碎硅片清洗效果视觉检测方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090074317A1 (en) * 2007-09-19 2009-03-19 Samsung Electronics Co., Ltd. System and method for reducing halo effect in image enhancement
CN102509272A (zh) * 2011-11-21 2012-06-20 武汉大学 一种基于颜色恒常性的彩色图像增强方法
CN102789635A (zh) * 2012-07-18 2012-11-21 奇瑞汽车股份有限公司 一种图像增强方法和装置
CN107730475A (zh) * 2017-11-13 2018-02-23 苏州科达科技股份有限公司 图像增强方法及系统

Family Cites Families (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7342609B2 (en) * 2000-05-09 2008-03-11 Eastman Kodak Company Exposure adjustment in an imaging apparatus
US7881554B2 (en) * 2006-06-05 2011-02-01 Stmicroelectronics S.R.L. Method for correcting a digital image
US7602988B2 (en) * 2006-09-01 2009-10-13 Mediatek Inc. Method for adjusting saturation and contrast of an area of an image and apparatus thereof
US7856141B2 (en) * 2006-09-01 2010-12-21 Mediatek Inc. Method for adjusting saturation and contrast of an area of an image and apparatus thereof
KR100879536B1 (ko) * 2006-10-30 2009-01-22 삼성전자주식회사 영상의 화질 개선을 위한 방법 및 시스템
US9830691B2 (en) * 2007-08-03 2017-11-28 The University Of Akron Method for real-time implementable local tone mapping for high dynamic range images
TW201011691A (en) * 2008-09-05 2010-03-16 Primax Electronics Ltd Method for adjusting brightness of images
JP5235759B2 (ja) * 2009-04-07 2013-07-10 キヤノン株式会社 画像処理装置、画像処理方法、及びプログラム
US8781248B2 (en) * 2010-01-28 2014-07-15 Stmicroelectronics Asia Pacific Pte. Ltd. Image details preservation and enhancement
GB201002632D0 (en) * 2010-02-16 2010-03-31 Im Sense Ltd Method and system for generating enhanced images
WO2012017440A1 (en) * 2010-08-05 2012-02-09 Ramot At Tel-Aviv University Ltd. Method and system for processing an image featuring multiple scales
JP5165076B2 (ja) * 2011-01-31 2013-03-21 シャープ株式会社 映像表示装置
KR101987764B1 (ko) * 2012-09-12 2019-06-11 삼성전자주식회사 선명도 향상 장치 및 방법
US9412155B2 (en) * 2012-12-31 2016-08-09 Karl Storz Imaging, Inc. Video system with dynamic contrast and detail enhancement
US10074162B2 (en) * 2016-08-11 2018-09-11 Intel Corporation Brightness control for spatially adaptive tone mapping of high dynamic range (HDR) images

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090074317A1 (en) * 2007-09-19 2009-03-19 Samsung Electronics Co., Ltd. System and method for reducing halo effect in image enhancement
CN102509272A (zh) * 2011-11-21 2012-06-20 武汉大学 一种基于颜色恒常性的彩色图像增强方法
CN102789635A (zh) * 2012-07-18 2012-11-21 奇瑞汽车股份有限公司 一种图像增强方法和装置
CN107730475A (zh) * 2017-11-13 2018-02-23 苏州科达科技股份有限公司 图像增强方法及系统

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114511462A (zh) * 2022-02-11 2022-05-17 电子科技大学 一种视觉图像增强方法
CN114511462B (zh) * 2022-02-11 2023-04-18 电子科技大学 一种视觉图像增强方法
CN116245771A (zh) * 2023-03-22 2023-06-09 新光维医疗科技(苏州)股份有限公司 一种基于图像二次增强的内窥镜图像亮度自动增强方法及系统
CN116245771B (zh) * 2023-03-22 2024-04-02 新光维医疗科技(苏州)股份有限公司 一种基于图像二次增强的内窥镜图像亮度自动增强方法及系统

Also Published As

Publication number Publication date
DE112018005567T5 (de) 2020-06-25
GB2581907B (en) 2022-03-09
CN107730475A (zh) 2018-02-23
US20200273154A1 (en) 2020-08-27
CN107730475B (zh) 2019-08-23
GB2581907A (en) 2020-09-02
GB202007219D0 (en) 2020-07-01
US11127122B2 (en) 2021-09-21

Similar Documents

Publication Publication Date Title
WO2019091270A1 (zh) 图像增强方法及系统
Fu et al. Retinex-based perceptual contrast enhancement in images using luminance adaptation
Parihar et al. A study on Retinex based method for image enhancement
CN104240194B (zh) 一种基于抛物线函数的低照度图像增强算法
WO2016206087A1 (zh) 一种低照度图像处理方法和装置
CN110428379B (zh) 一种图像灰度增强方法及系统
WO2014169579A1 (zh) 一种色彩增强方法及装置
CN110852956A (zh) 一种高动态范围图像的增强方法
CN113989147A (zh) 基于直方图均衡和改进反锐化掩模的内窥镜图像增强方法
RU2622095C1 (ru) Способ улучшения цифровых цветных изображений
Gao et al. A new color contrast enhancement algorithm for robotic applications
CN111476732B (zh) 一种图像融合及去噪的方法及系统
Wen et al. Autonomous robot navigation using Retinex algorithm for multiscale image adaptability in low-light environment
CN110992287B (zh) 一种非均匀光照视频的清晰化方法
Singh et al. Image enhancement by adaptive power-law transformations
CN105405110A (zh) 非均匀光照补偿方法
Li et al. Saliency guided naturalness enhancement in color images
CN110298796B (zh) 基于改进Retinex与对数图像处理的低照度图像增强方法
Asari et al. Nonlinear enhancement of extremely high contrast images for visibility improvement
CN113160066A (zh) 一种低照度图像高效增强方法
CN110675332A (zh) 一种金属腐蚀图像质量的增强方法
CN116612050B (zh) 一种基于暗通道的先验去雾方法
Unaldi et al. Fast and robust wavelet-based dynamic range compression with local contrast enhancement
Liu et al. Low-light image enhancement based on Retinex theory for beam-splitting prism system
CN111223060B (zh) 基于自适应plip模型的图像处理方法

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: 18875813

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 202007219

Country of ref document: GB

Kind code of ref document: A

Free format text: PCT FILING DATE = 20181017

122 Ep: pct application non-entry in european phase

Ref document number: 18875813

Country of ref document: EP

Kind code of ref document: A1