CN102156966A - Medical image denoising - Google Patents

Abstract

The invention provides a method and a device for performing adaptive filtering on a medical image. According to one embodiment, a target image can be processed with iterative denoising treatment for many times based on image directionality, and a correlation degree between the resultant image after performing iteration each time and the target image is evaluated so as to judge whether the correlation degree is less than a predetermined threshold, thereby judging that the iterative treatment is converged.

Description

Medical Image Denoising

joint research

This application is jointly researched by North China University of Technology and Information Institute of Beijing Jiaotong University, and supported by the following funds: National Natural Science Foundation of China (No.60903066, No.60972085), Beijing Natural Science Foundation of China (No.4102049), New Teacher of the Ministry of Education Fund (No.20090009120006), and Beijing Municipal Higher Education Intensification Plan for Talents (PHR201008187).

technical field

The present invention relates to the field of medical image processing, and more specifically, to a medical image denoising method and device.

Background technique

The imaging principle of medical images is relatively complicated. In the process of imaging or transmission, due to the inherent characteristics of the equipment itself and the influence of the external environment, some noise will inevitably be introduced into the image, which has brought great impact on subsequent processing and medical diagnosis. Trouble. Therefore, denoising work in medical image preprocessing is very necessary, and finding new denoising methods is still a challenging problem in various fields.

In the field of image processing, generally speaking, traditional denoising methods include denoising based on spatial filtering and denoising based on frequency domain filtering. The denoising method based on spatial filtering uses a mask/kernel based on median filtering or mean filtering to perform neighborhood processing, so as to achieve the purpose of smooth denoising. The smoothing frequency domain filters that may be used in the denoising method based on frequency domain filtering may include ideal low-pass filters, Butterworth filters and high-pass filters. For example, spatial and temporal filters for smoothing are described in detail in "Digital Image Processing (Second Edition)" by Rafael C. Gonzalez and Richard E. Woods et al. (hereinafter referred to as "Reference 1"). Among them, the Gaussian smoothing filter, as a basic denoising method, is widely used to perform denoising processing on medical images, and its two-dimensional form can be given as follows:

$\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; D.</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

For a long time, a variety of denoising methods based on the Gaussian smoothing filter have been studied on the basis of the conventional Gaussian smoothing filter. However, the conventional Gaussian smoothing filter does not consider the directionality of the image, so the edge preservation effect of the image in the smoothing filter is poor, which will cause serious problems in medical image processing that requires precise edge positioning. Considering the directionality when filtering can fully preserve the edge, a denoising method that makes full use of the directionality information in the denoising process is needed.

In addition, in medical image processing, in order to obtain a satisfactory result image for a doctor's diagnosis, iterative denoising of the image may usually be required. Due to the massive nature of medical image data, the denoising preprocessing time increases with the number of iterations and can reach a point of impatience. However, it should be pointed out that the more iterations, the better the denoising effect will be. This is because in the denoising process, there will inevitably be a certain blur effect inside the image, and the increase in the number of iterations, the result will tend to converge. At present, the number of iterations can only be controlled by observing the denoising effect of the image with human eyes, which is highly subjective. Therefore, there is a need for an adaptive method that controls the number of iterations in the denoising process to achieve a good balance between processing time and denoising effect.

In addition, it is also expected that there will be a new adaptive convergence judgment method that can simulate the results of human eyes observing images as much as possible, so that the judgment of convergence can be more similar to the human visual experience.

Contents of the invention

Aiming at the above-mentioned problems in the prior art, the present invention proposes a method for adaptive iterative image denoising, comprising the following steps: (i) obtaining an input image; (ii) performing t-th times iterative denoising processing (t is a positive integer), and when t=1, the target image is the input image, and when t>1, the target image is the result of the t-1th iteration image; (iii) evaluate the correlation between the result image after the tth iteration and the target image to determine whether the correlation is less than a preset threshold; (iv) when the correlation is greater than When it is equal to the preset threshold, go to (ii) perform the next iterative denoising based on t; and (v) when the degree of correlation is less than the preset threshold, end the denoising process, and convert the The resulting image is output as a denoised image.

In another aspect, the present invention proposes a device for adaptive iterative image denoising, including: a module for obtaining an input image; a module for performing t-th iteration denoising processing on a target image based on image orientation module (t is a positive integer), and when t=1, the target image is the input image, and when t>1, the target image is the result image after the t-1 iteration; for A module for evaluating the correlation between the result image after the tth iteration and the target image to determine whether the correlation is less than a preset threshold; for when the correlation is greater than or equal to the preset When the threshold is set, then use the module for performing the t-th iterative denoising process on the target image based on the image directionality to perform the next iterative denoising based on t; and for when the correlation is less than the specified When the preset threshold is exceeded, the denoising process is ended, and the resulting image is output as a denoised image module.

In yet another aspect, the present invention proposes a medical image processing system, including: a medical image acquisition device for acquiring medical images; a communication circuit for facilitating wired or wireless transmission of the acquired medical images to filter circuit; a filter circuit configured to perform the following iterative smoothing filtering process: (i) receive the collected medical image as an input image, set the number of iterations t=1; (ii) perform the target image based on image directionality Perform the tth iterative denoising process (t is a positive integer), and when t=1, the target image is the input image, and when t>1, the target image is the t-1th iteration (iii) evaluate the degree of correlation between the result image after the tth iteration and the target image to determine whether the degree of correlation is less than a preset threshold; (iv) when the When the degree of correlation is greater than or equal to the preset threshold, go to (ii) perform the next iterative denoising based on t=t+1; and (v) when the degree of correlation is less than the preset threshold, end Denoising processing, outputting the resulting image as a denoised image; a display device, configured to receive the smoothed and filtered medical image from the filter circuit, and display the smoothed and filtered medical image; wherein, the The filter circuit is further configured to perform the nth iterative denoising process using the following formula,

$\left\{\begin{array}{c}\frac{<span\; class="notranslate">\; \&PartialD;</span>\mathrm{<span\; class="notranslate">\; I</span>}}{<span\; class="notranslate">\; \&PartialD;</span>\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; div</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; (</span>{<span\; class="notranslate">\; |</span><span\; class="notranslate">\; \&dtri;</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&dtri;</span>\mathrm{<span\; class="notranslate">\; I</span>}\\ \mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>\end{array}\right.<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

表示梯度，g(·)是方向性方程，t为迭代次数，x和y为I中的像素坐标，并且所述方向性方程g(·)为以下之一：

或

其中，k为预先设定的影响因子，并且其中，所述滤波器电路被进一步配置为，使用以下评估函数执行所述评估：Wherein, I is the target image, div is a divergence operator,

Represents the gradient, g( ) is a directional equation, t is the number of iterations, x and y are pixel coordinates in I, and the directional equation g( ) is one of the following:

or

where k is a preset influencing factor, and wherein the filter circuit is further configured to perform the evaluation using the following evaluation function:

E _Correlation = [l(x, y)] ^α [c(x, y)] ^β [s′(x, y)] ^γ (3)

或，并且其中，x和y分别是所述结果图像和所述目标图像，μ_x、μ_y、σ_x、σ_y、σ_xy分别为x和y的亮度均值，方差和协方差；C₁、C₂、C₃是为了避免分母为零设置的常数，并且σ′_x、σ′_y、σ′_xy分别为图像x，y的梯度图像的标准差和协方差，并且其中所述梯度图像采用Robert算子产生。in,

Or, and wherein, x and y are the result image and the target image respectively, μ _x , μ _y , σ _x , σ _y , σ _xy are the brightness mean, variance and covariance of x and y respectively; C ₁ , C ₂ , C ₃ are constants set to avoid the denominator being zero, and σ′ _x , σ′ _y , σ′ _xy are the standard deviation and covariance of the gradient image of image x, y respectively, and the gradient image Generated by Robert operator.

Various aspects and features disclosed herein are described in further detail below.

Description of drawings

Fig. 1 shows a kind of medical image processing system;

Figure 2 shows details of a denoising filter according to some embodiments;

Fig. 3 shows a flow chart of a method for smoothing and denoising filtering according to some embodiments of the present invention; and

Figures 4(a) and 4(b) show the variation curves of the number of iterations and the evaluation function for different noises.

Detailed ways

Various aspects are now described with reference to the figures. In the following description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of one or more aspects. It may be evident, however, that such aspects can be practiced without these specific details.

As used in this application, the terms "component", "module", "system" and the like are intended to refer to a computer-related entity such as, but not limited to, hardware, firmware, a combination of hardware and software, software, Or software in execution. For example, a component may be, but is not limited to being limited to, a process running on a processor, a processor, an object, an executable, a thread of execution, a program, and/or a computer. For example, both an application running on a computing device and the computing device can be components. One or more components can reside within a process and/or thread of execution and a component can be localized on one computer and/or distributed between two or more computers. In addition, these components can execute from various computer readable media having various data structures stored thereon. Components can communicate by means of local and/or remote processes, such as from signals having one or more data packets, for example, from interacting with another component in a local system, a distributed system, and/or with another component in a distributed system by means of a signal. Data of a component on a network such as the Internet that interacts with other systems by means of signals.

Fig. 1 shows a medical image processing system 100 according to an embodiment of the present invention. Apparatus 101 is a medical image acquisition device for acquiring images of a part of a patient's body (e.g. chest, head, etc.) according to any medical imaging technique known in the art, such known medical imaging techniques include ultrasound imaging, X-ray, CT, MRI, etc. The medical images collected by the medical image collection device 101 are transmitted to the image processing device 103 in a wired and/or wireless manner through the communication device 102, and the image processing device 103 performs smoothing processing on the received medical images to remove And the noise caused in the transmission process, such as Gaussian (Gaussian) noise, speckle (speckle) noise and Poisson (poisson) noise, etc., and provide the image after denoising processing to the display device 104 for display, for the doctor to diagnose for. However, it should be understood that the image processing device 103 may also perform various other processing on the input image, such as edge detection, image registration, pattern recognition and so on.

The image processing device 103 may use a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components Or any combination thereof designed to perform the functions described herein is implemented or performed. A general-purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing devices, eg, a DSP and a microprocessor, multiple microprocessors, one or more microprocessors with a DSP core, or any other such architecture. Additionally, at least one processor may comprise one or more modules operable to perform one or more of the steps and/or operations described above.

When the image processing device 103 is implemented with a hardware circuit such as ASIC, FPGA, it may include various circuit blocks configured to perform various functions. Those skilled in the art can design and implement these circuits in various ways according to various constraints imposed on the entire system, so as to realize various functions disclosed in the present invention. For example, the image processing device 103 implemented by hardware circuits such as ASIC and FPGA may include filter circuits and/or other circuit modules, which are used to perform denoising on input images according to various adaptive smoothing filtering schemes disclosed herein. Those skilled in the art should be able to understand and realize that the image processing device 103 described herein may optionally include any other available circuit modules other than the filter circuit, for example, configured to perform edge detection, image registration, mode Any circuit blocks identified. The functions realized by the filter circuit are described in detail below in conjunction with the flowchart of FIG. 3 .

The image storage device 105 can be coupled to the image acquisition device 101 and/or the image processing device 103 to store the raw data collected by the image acquisition device 101 and/or the output image processed by the image processing device 103 .

Figure 2 shows details of a denoising filter 200 according to some embodiments. Denoising filter 200 may include processing circuitry 210 and memory 220 . Wherein the processing circuit 210 may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components Or any combination thereof designed to perform the functions described herein. The processing circuit 210 may include various circuit modules for implementing various functions. In one embodiment, these circuit blocks may exist in the processing circuit 210 in the form of discrete components. In another embodiment, these circuit modules may be only functional modules in the electronic design diagram of the circuit, but do not exist in the actual circuit. For example, when using commercial electronic circuit design software to design circuit block diagrams and finally electronically write them into circuits, these circuit blocks may exist separately or collectively in one or more files supported by the electronic circuit design software , while merging into a single design at the final circuit writing stage.

In one embodiment, the processing circuit 210 may include: a circuit module 211 for obtaining an input image; a circuit module 212 for performing t-th iterative denoising processing on the target image based on image orientation; The correlation degree between the result image after the second iteration and the target image is evaluated to determine whether the correlation degree is less than a preset threshold circuit module 213; for when the correlation degree is greater than or equal to the preset threshold, then use the module for performing t-th iterative denoising processing on the target image based on the image orientation to perform the circuit module 214 for the next iterative denoising based on t; and for when the correlation is less than When the preset threshold is reached, the denoising process ends, and the circuit module 215 outputs the resulting image as a denoised image. In one embodiment, the memory 220 may be used to store input data and output data of the denoising filter 200 and intermediate data of each circuit module of the processing circuit 210 . For example, in one embodiment, the processing circuit 210 may store the result image and the target image of each iteration in the memory 220 for access during the next iteration. The memory 220 may be various random access memories (RAM), including but not limited to: RAM, DRAM, DDR RAM, and the like. The memory 220 is connected to the processing circuit 210 through a bus.

Fig. 3 shows a flowchart of a method 300 for performing smoothing and denoising processing on an input image according to multiple embodiments of the present invention.

In step S310 , an input image is obtained from the image acquisition device 101 or the image storage device 105 .

In steps S320-S340, an iterative denoising process is performed on the input image, wherein t can be used to represent the number of iterations, when the number of iterations t=1, the target image of the denoising process is the input image, and when the iterative When the number of times t>1, the target image of the denoising process is the result image after the t-1th iteration.

Specifically, in step S320, the t-th iterative denoising process is performed on the target image based on the image orientation. Specifically, the n-th iterative denoising process can be performed using the following formula (4) based on image orientation,

$\left\{\begin{array}{c}\frac{<span\; class="notranslate">\; \&PartialD;</span>\mathrm{<span\; class="notranslate">\; I</span>}}{<span\; class="notranslate">\; \&PartialD;</span>\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; div</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; (</span>{<span\; class="notranslate">\; |</span><span\; class="notranslate">\; \&dtri;</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&dtri;</span>\mathrm{<span\; class="notranslate">\; I</span>}\\ \mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; ;</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>\end{array}\right.<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$

表示梯度，g(·)是方向性方程，t为迭代次数，x和y为I中的像素坐标。在一个实施例中，方向性方程g(·)可以为以下之一：

或

其中，k为预先设定的影响因子。本领域技术人员可以理解和认识到，公式(4)是一个以I₀为初始条件的发展方程，该方程的解I(x，y；t)在离散情形下，对参数t选择合适的步长得到迭代方程，在得到满意的图像时可以停止迭代。In formula (4), I is the target image, and div is a divergence operator,

Indicates the gradient, g( ) is the directional equation, t is the number of iterations, and x and y are the pixel coordinates in I. In one embodiment, the directivity equation g(·) can be one of the following:

or

Among them, k is a preset impact factor. Those skilled in the art can understand and realize that the formula (4) is a development equation with I ₀ as the initial condition, and the solution I(x, y; t) of the equation is in a discrete situation, and an appropriate step is selected for the parameter t The iterative equation is obtained, and the iteration can be stopped when a satisfactory image is obtained.

In step S330, the degree of correlation between the result image after the t-th iteration in step S320 and the target image that has not undergone the t-th iteration denoising process in step S320 can be evaluated, and can be evaluated in step S340 In the process, it is judged whether the correlation degree is smaller than a preset threshold.

In one embodiment, the correlation degree may be evaluated by using the following evaluation function:

E _Correlation = [l(x, y)] ^α [c(x, y)] ^β [s(x, y)] ^γ (5)

并且x和y分别是所述结果图像和所述目标图像，μ_x、μ_y、σ_x、σ_y、σ_xy分别为x和y的亮度均值，方差和协方差；C₁、C₂、C₃是为了避免分母为零设置的常数。In formula (5),

And x and y are the result image and the target image respectively, μ _x , μ _y , σ _x , σ _y , σ _xy are the brightness mean, variance and covariance of x and y respectively; C ₁ , C ₂ , C ₃ is a constant set to avoid zero denominator.

Those skilled in the art can recognize and understand that, in fact, in formula (5), l(x, y), c(x, y), s(x, y) are the brightness correlations of two images x and y, respectively Degree, contrast correlation and structure correlation.

In a specific embodiment, in formula (5), it can be set that α=β=γ=1, C ₃ =C ₂ /2, so that formula (5) can be further changed into:

${\mathrm{<span\; class="notranslate">\; V</span>}}_{\mathrm{<span\; class="notranslate">\; Correlation</span>}}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; xy</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; the\; <figure-callout\; id="2"\; label="y"\; filenames="HSA00000469797900031.png,HSA00000469797900032.png"\; state="\{\{state\}\}">y</figure-callout></span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; the\; <figure-callout\; id="2"\; label="y"\; filenames="HSA00000469797900031.png,HSA00000469797900032.png"\; state="\{\{state\}\}">y</figure-callout></span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; )</span>$

替代公式(5)中的

其中，σ′_x、σ′_y、σ′_xy分别为图像x，y的梯度图像的标准差和协方差，从而公式(5)可以变为：In another alternative embodiment, the

instead of in formula (5)

Among them, σ′ _x , σ′ _y , and σ′ _xy are the standard deviation and covariance of the gradient images of images x and y, respectively, so that formula (5) can be changed to:

E _Correlation = [l(x, y)] ^α [c(x, y)] ^β [s′(x, y)] ^γ (7)

In a specific embodiment, the gradient image in formula (7) can be generated by using the Robert operator. Here, it should be understood that although the Sobel operator is also an operator known in the art to calculate image gradients, the Sobel itself has a weighted filtering function, which is not applicable to the adaptive filtering idea proposed in this paper. In the present invention, in order to highlight the difference in gradient content between two images with adjacent iterations, we use the characteristic of high positioning accuracy of the Robert operator to obtain the gradient image.

Due to the different types and sizes of image noise, the number of iterations required to achieve the best denoising effect is different. It should be pointed out that the higher the number of iterations, the better the noise removal effect. On the one hand, although the adaptive denoising filter described in the present invention can better protect the edges and details of the image, as the number of iterations increases, There will still be some blurring effect inside the image. On the other hand, the increase of the number of iterations will increase the amount of calculation and make the operation time longer. Therefore, finding the optimal number of iterations is the key to adaptive filtering.

Figures 4(a) and 4(b) respectively show the corresponding graphs of the number of iterations and the correlation degree E _Correlation for 10 iterations using formula (5) and formula (7). Observe the change trend of E _Correlation . As the number of iterations increases, E _Correlation tends to 1 in a relatively gentle way at a certain point. That is, the difference between two adjacent images has become smaller and smaller. It can be considered that at this time A basically satisfactory denoising effect is obtained. Due to the different types and sizes of noise, the optimal number of iterations required will be different. At the end of each iteration, the slope of the change curve is calculated, and when the slope of the point is less than the set threshold, the iteration is stopped.

Comparing Figure 4(a) and Figure 4(b), the curve Figure 4(b) obtained by formula (7) fluctuates significantly with the number of iterations, which is helpful for us to determine the difference between the effects of the two iterations before and after, so as to provide us with Best iterative steps to lay the groundwork.

In a specific embodiment, according to the experiment in Fig. 4(b), the obtained adaptive threshold is 0.0137.

Returning to FIG. 3 , further, when it is judged in step S340 that the correlation degree is greater than or equal to the preset threshold, it can be considered that the denoising process has not yet reached convergence, so the process of the method can return to step S320 to perform the next step based on t One iteration of denoising (of course, t becomes t=t+1).

When it is judged in step S340 that the degree of correlation is less than the preset threshold, it can be considered that the denoising process has reached convergence, and the method can end the denoising process and proceed to optional step S350, where the result The image is provided to the next processing stage as a denoised image, and the next processing stage in this optional step S350 may be: edge detection is performed on the denoised image, pattern recognition is performed on the denoised image, and / Or register the denoised image and so on. Those skilled in the art can design the next processing stage according to the design constraints of the actual medical image system. In an alternative embodiment, there may be no next processing stage at all and the denoised image directly displayed to the user at step S360. In FIG. 3, the box of step S360 is drawn as a dotted line to indicate its optionality.

According to the experimental results of the inventors, the adaptive denoising method proposed by the present invention has achieved relatively good results for images containing three types of noises, Gauss, Poisson and Speckled, after multiple tests on the "cameraman" image commonly used in the field of image processing. the result of.

While the foregoing disclosures discuss exemplary aspects and/or embodiments, it should be noted that many changes and/or changes may be made therein without departing from the scope of the described aspects and/or embodiments as defined by the claims. Revise. Also, although elements of the described aspects and/or embodiments are described or claimed in the singular, the plural is contemplated unless limitation to the singular is explicitly stated. In addition, all or part of any aspect and/or embodiment can be used in combination with all or part of any other aspect and/or embodiment, unless a difference is indicated.

Claims (9)

1. A method for adaptive iterative image denoising, comprising the steps of:

(i) obtaining an input image;

(ii) carrying out t-th iteration denoising processing on a target image based on image directionality (t is a positive integer), wherein when t is 1, the target image is the input image, and when t is more than 1, the target image is a result image after t-1 iteration;

(iii) evaluating the correlation degree between the result image after the t iteration and the target image to judge whether the correlation degree is smaller than a preset threshold value;

(iv) when the correlation degree is greater than or equal to the preset threshold value, turning to (ii) carrying out next iteration denoising based on t; and

(v) and when the correlation degree is smaller than the preset threshold value, ending the denoising processing, and outputting the result image as a denoised image.

2. The method of claim 1, step (ii) further comprising: the nth iteration denoising process is performed by using the following formula,

<math><mfenced open='{' close=''><mtable><mtr><mtd><mfrac><mrow><mo>&PartialD;</mo><mi>I</mi></mrow><mrow><mo>&PartialD;</mo><mi>t</mi></mrow></mfrac><mo>=</mo><mi>div</mi><mrow><mo>(</mo><mi>g</mi><mrow><mo>(</mo><msup><mrow><mo>|</mo><mo>&dtri;</mo><mi>I</mi><mo>|</mo></mrow><mn>2</mn></msup><mo>)</mo></mrow><mo>)</mo></mrow><mo>&dtri;</mo><mi>I</mi></mtd></mtr><mtr><mtd><mi>I</mi><mrow><mo>(</mo><mi>x</mi><mo>,</mo><mi>y</mi><mo>;</mo><mn>0</mn><mo>)</mo></mrow><mo>=</mo><msub><mi>I</mi><mn>0</mn></msub><mrow><mo>(</mo><mi>x</mi><mo>,</mo><mi>y</mi><mo>)</mo></mrow></mtd></mtr></mtable></mfenced></math>

wherein I is the target image, div is a divergence operator,

representing the gradient, g (-) is a directional equation, t is the number of iterations, and x and y are the pixel coordinates in I.

3. The method of claim 2, wherein the directional equation g (-) is one of:

wherein k is a preset influence factor.

4. The method of claim 1, step (iii) further comprising: the evaluation is performed using the following evaluation function:

E_Correlation＝[l(x，y)]^α[c(x，y)]^β[s(x，y)]^γ；

wherein, <math><mrow><mi>l</mi><mrow><mo>(</mo><mi>x</mi><mo>,</mo><mi>y</mi><mo>)</mo></mrow><mo>=</mo><mfrac><mrow><msub><mrow><mn>2</mn><mi>&mu;</mi></mrow><mi>x</mi></msub><msub><mi>&mu;</mi><mi>y</mi></msub><mo>+</mo><msub><mi>C</mi><mn>1</mn></msub></mrow><mrow><msubsup><mi>&mu;</mi><mi>x</mi><mn>2</mn></msubsup><mo>+</mo><msubsup><mi>&mu;</mi><mi>y</mi><mn>2</mn></msubsup><mo>+</mo><msub><mi>C</mi><mn>1</mn></msub></mrow></mfrac><mo>,</mo></mrow></math> <math><mrow><mi>c</mi><mrow><mo>(</mo><mi>x</mi><mo>,</mo><mi>y</mi><mo>)</mo></mrow><mo>=</mo><mfrac><mrow><msub><mrow><mn>2</mn><mi>&sigma;</mi></mrow><mi>x</mi></msub><msub><mi>&sigma;</mi><mi>y</mi></msub><mo>+</mo><msub><mi>C</mi><mn>2</mn></msub></mrow><mrow><msubsup><mi>&sigma;</mi><mi>x</mi><mn>2</mn></msubsup><mo>+</mo><msubsup><mi>&sigma;</mi><mi>y</mi><mn>2</mn></msubsup><mo>+</mo><msub><mi>C</mi><mn>2</mn></msub></mrow></mfrac><mo>,</mo></mrow></math> <math><mrow><mi>s</mi><mrow><mo>(</mo><mi>x</mi><mo>,</mo><mi>y</mi><mo>)</mo></mrow><mo>=</mo><mfrac><mrow><msub><mi>&sigma;</mi><mi>xy</mi></msub><mo>+</mo><msub><mi>C</mi><mn>3</mn></msub></mrow><mrow><msub><mi>&sigma;</mi><mi>x</mi></msub><msub><mi>&sigma;</mi><mi>y</mi></msub><mo>+</mo><msub><mi>C</mi><mn>3</mn></msub></mrow></mfrac><mo>,</mo></mrow></math>

and wherein x and y are the result image and the target image, respectively, μ_x、μ_y、σ_x、σ_y、σx_yMean, variance and covariance of the luminance for x and y, respectivelyA difference; c₁、C₂、C₃To avoid constants with denominators set to zero.

5. The method of claim 4, wherein α β γ 1, C₃＝C₂2, whereby the evaluation function becomes:

<math><mrow><msub><mi>V</mi><mi>Correlation</mi></msub><mo>=</mo><mfrac><mrow><mrow><mo>(</mo><msub><mrow><mn>2</mn><mi>&mu;</mi></mrow><mi>x</mi></msub><msub><mi>&mu;</mi><mi>y</mi></msub><mo>+</mo><msub><mi>C</mi><mn>1</mn></msub><mo>)</mo></mrow><mrow><mo>(</mo><mn>2</mn><msub><mtext>&sigma;</mtext><mi>xy</mi></msub><mo>+</mo><msub><mi>C</mi><mn>2</mn></msub><mo>)</mo></mrow></mrow><mrow><mrow><mo>(</mo><msubsup><mi>&mu;</mi><mi>x</mi><mn>2</mn></msubsup><mo>+</mo><msubsup><mi>&mu;</mi><mi>y</mi><mn>2</mn></msubsup><mo>+</mo><msub><mi>C</mi><mn>1</mn></msub><mo>)</mo></mrow><mrow><mo>(</mo><msubsup><mi>&sigma;</mi><mi>x</mi><mn>2</mn></msubsup><msubsup><mrow><mo>+</mo><mi>&sigma;</mi></mrow><mi>y</mi><mn>2</mn></msubsup><mo>+</mo><msub><mi>C</mi><mn>2</mn></msub><mo>)</mo></mrow></mrow></mfrac><mo>.</mo></mrow></math>

6. the method of claim 4, wherein

Substitution

Wherein, σ'_x、σ′_y、σ′_xyStandard deviation and covariance of the gradient images of images x, y, respectively, and wherein the gradient images are generated using the Robert operator.

7. The method of claim 6, wherein the preset threshold is 0.0137.

8. An apparatus for adaptive iterative image denoising, comprising:

means for obtaining an input image;

a module for performing a t-th iterative denoising process on a target image based on image directionality (t is a positive integer), wherein when t is 1, the target image is the input image, and when t > 1, the target image is a result image after the t-1 th iteration;

a module for evaluating the correlation between the result image after the t-th iteration and the target image to judge whether the correlation is smaller than a preset threshold value;

when the correlation degree is greater than or equal to the preset threshold value, performing t-based next iteration denoising by using the module for performing the t-th iteration denoising processing on the target image based on the image directivity; and

and the module is used for ending the denoising processing when the correlation degree is smaller than the preset threshold value and outputting the result image as a denoised image.

9. A medical image processing system, comprising:

a medical image acquisition device for acquiring a medical image;

a communication circuit for facilitating wired or wireless transmission of the acquired medical image to the filter circuit;

a filter circuit configured to perform the following iterative smoothing filtering process:

(i) receiving an acquired medical image as an input image, and setting the iteration number t to be 1;

(ii) carrying out t-th iteration denoising processing on a target image based on image directionality (t is a positive integer), wherein when t is 1, the target image is the input image, and when t is more than 1, the target image is a result image after t-1 iteration;

(iii) evaluating the correlation degree between the result image after the t iteration and the target image to judge whether the correlation degree is smaller than a preset threshold value;

(iv) when the correlation degree is greater than or equal to the preset threshold, turning to (ii) carrying out next iteration denoising based on t ═ t + 1; and

(v) when the correlation degree is smaller than the preset threshold value, ending denoising processing, and outputting the result image as a denoised image;

display means for receiving the smoothed filtered medical image from the filter circuit and displaying the smoothed filtered medical image; and

a storage device for storing the acquired medical image and/or the processed medical image;

wherein the filter circuit is further configured to perform an nth iteration denoising process using the following equation,

<math><mfenced open='{' close=''><mtable><mtr><mtd><mfrac><mrow><mo>&PartialD;</mo><mi>I</mi></mrow><mrow><mo>&PartialD;</mo><mi>t</mi></mrow></mfrac><mo>=</mo><mi>div</mi><mrow><mo>(</mo><mi>g</mi><mrow><mo>(</mo><msup><mrow><mo>|</mo><mo>&dtri;</mo><mi>I</mi><mo>|</mo></mrow><mn>2</mn></msup><mo>)</mo></mrow><mo>)</mo></mrow><mo>&dtri;</mo><mi>I</mi></mtd></mtr><mtr><mtd><mi>I</mi><mrow><mo>(</mo><mi>x</mi><mo>,</mo><mi>y</mi><mo>;</mo><mn>0</mn><mo>)</mo></mrow><mo>=</mo><msub><mi>I</mi><mn>0</mn></msub><mrow><mo>(</mo><mi>x</mi><mo>,</mo><mi>y</mi><mo>)</mo></mrow></mtd></mtr></mtable></mfenced></math>

wherein I is the target image, div is a divergence operator,representing a gradient, g (-) is a directional equation, t is the number of iterations, x and y are pixel coordinates in I, and the directional equation g (-) is one of:

or

Wherein k is a preset influence factor,

and wherein the filter circuit is further configured to perform the evaluation using the following evaluation function:

E_Correlation＝[l(x，y)]^α[c(x，y)]^β[s′(x，y)]^γ；

wherein, <math><mrow><mi>l</mi><mrow><mo>(</mo><mi>x</mi><mo>,</mo><mi>y</mi><mo>)</mo></mrow><mo>=</mo><mfrac><mrow><msub><mrow><mn>2</mn><mi>&mu;</mi></mrow><mi>x</mi></msub><msub><mi>&mu;</mi><mi>y</mi></msub><mo>+</mo><msub><mi>C</mi><mn>1</mn></msub></mrow><mrow><msubsup><mi>&mu;</mi><mi>x</mi><mn>2</mn></msubsup><mo>+</mo><msubsup><mi>&mu;</mi><mi>y</mi><mn>2</mn></msubsup><mo>+</mo><msub><mi>C</mi><mn>1</mn></msub></mrow></mfrac><mo>,</mo></mrow></math> <math><mrow><mi>c</mi><mrow><mo>(</mo><mi>x</mi><mo>,</mo><mi>y</mi><mo>)</mo></mrow><mo>=</mo><mfrac><mrow><msub><mrow><mn>2</mn><mi>&sigma;</mi></mrow><mi>x</mi></msub><msub><mi>&sigma;</mi><mi>y</mi></msub><mo>+</mo><msub><mi>C</mi><mn>2</mn></msub></mrow><mrow><msubsup><mi>&sigma;</mi><mi>x</mi><mn>2</mn></msubsup><mo>+</mo><msubsup><mi>&sigma;</mi><mi>y</mi><mn>2</mn></msubsup><mo>+</mo><msub><mi>C</mi><mn>2</mn></msub></mrow></mfrac><mo>,</mo></mrow></math> <math><mrow><msup><mi>s</mi><mo>&prime;</mo></msup><mrow><mo>(</mo><mi>x</mi><mo>,</mo><mi>y</mi><mo>)</mo></mrow><mo>=</mo><mfrac><mrow><msubsup><mi>&sigma;</mi><mi>xy</mi><mo>&prime;</mo></msubsup><mo>+</mo><msub><mi>C</mi><mn>3</mn></msub></mrow><mrow><msubsup><mi>&sigma;</mi><mi>x</mi><mo>&prime;</mo></msubsup><msubsup><mi>&sigma;</mi><mi>y</mi><mo>&prime;</mo></msubsup><mo>+</mo><msub><mi>C</mi><mn>3</mn></msub></mrow></mfrac></mrow></math> or,

and wherein x and y are the result image and the target image, respectively, μ_x、μ_y、σ_x、σ_y、σ_xyThe mean, variance and covariance of the luminance are x and y, respectively; c₁、C₂、C₃Is a constant set to avoid denominator being zero, and σ'_x、σ′_y、σ′_xyStandard deviation and covariance of the gradient images of images x, y, respectively, and wherein the gradient images are generated using the Robert operator.

Images