Abstract
Ultrasound imaging is an important form of medical imaging, in order to improve the clinical examination and disease diagnosis in the field of wide use. In this paper, by using the basic principle of the anisotropic diffusion model, mainly aimed at the ultrasonic image speckle noise removal algorithm is studied, the improvement schemes are put forward on the direction of diffusion control, and through the simulation of ultrasonic image, in ultrasonic image edge and noise filter both qualitative quantitative data comparison, to achieve the effectiveness of the algorithm. Experiments show that the algorithm can preserve the edge and texture details of the image while removing noise.The improved anisotropic diffusion based noise reduction algorithm for B ultrasonic image is proposed in this paper. The adaptive selection of diffusion threshold K reduces the step effect after treatment. By selecting the diffusion coefficient, each diffusion has the appropriate diffusion intensity in different regions.
Introduction
Ultrasound imaging technology, which USES high-frequency acoustic waves to obtain analytic images of soft tissues and organs, is a widely used form of medical imaging, which has very important applications in cardiovascular imaging, abdominal imaging, cardiology, urology and obstetrics and gynecology [1]. In addition, ultrasound imaging is often used as a guide for surgical treatment of many diseases. In many developing countries, ultrasound imaging is the most commonly used medical imaging diagnostic technique next to radiography [2]. In some developed countries, although some medical imaging methods adopting high-tech means (such as computed tomography and mri) are more easily favored by medical researchers, ultrasound is still a generally accepted and recognized medical diagnosis method [3]. The production and manufacturing of ultrasonic imaging system reflects a country’s level of scientific and technological development from the side. The United States, Japan and Germany are the main producers and exporters of ultrasonic equipment today. Developed in this paper, the main target customers (township-level family planning service station staff members at the grass-roots level in China) in three check to carry out family planning (check ring, pregnancy, illness), some problems in the work on the actual project the development of the digital ultrasound diagnostic system for family planning, primarily for ultrasound medical image noise removal algorithm research and simulation, and the design allow communications B ultrasonic data acquisition and image processing system [4].
Literature review
In 2012, Zhu and Chen, in the traditional anisotropic diffusion theory and adaptive filter, on the basis of establishing a new type of anisotropic diffusion model of SRAD (Speckle reducing anisotropic coursing together) compared with the previous diffusion model, SRAD in removing Speckle noise and keep the border made remarkable effect, at the same time for processing of ultrasonic image, especially suppress Speckle noise provides a new idea [5, 6]. Inspired by SRAD, many new models and improved algorithms have emerged. Fredj introduced SRAD into the spot noise removal application of 3d ultrasonic image [7]. Sahli with the Lee Kuan filter replaced the SRAD filter, this paper proposes a model of anisotropic diffusion based on details – DPAD (Detailpreservinganisotropicdiffusion) based on the diffusion coefficient related to accurately estimate the statistical parameters, and in the DPAD shows the iterative process more stable and more satisfactory filtering effect [8]; Nugroho proposed a self-directed anisotropic diffusion model-osrad (OrientedSRAD), which can generate different filtering behaviors along the image contour and the direction of main curvature [9]. Joshi introduces SUSAN edge detection operator in SRADc [10]. Lian introduced a Sigmoid function as the diffusion coefficient [11]. Other improved algorithms include SRAD based on adaptive window proposed by Liu, SRAD based on improved diffusion template proposed by Mittal, and speckle noise removal method based on the combination of anisotropic diffusion and total variation denoising proposed by Wang.
Ultrasonic image algorithm research
According to the principle of ultrasonic imaging technology, ultrasonic reflection wave will be attenuated and interfered to a certain extent in the process of returning, resulting in the appearance of speckle noise in the generated b-ultrasound image.
Traditional p-m diffusion model.
The main principle of the traditional median filtering algorithm is to arrange the N sampling values in a neighborhood according to the pixel size, and then take the median value to replace the effective value of the pixel in the current neighborhood center. Median filter has the advantage that can inhibit the volatility of the interference caused by accidental factors, have very good noise reduction effect of salt and pepper noise, compared with average filtering filter can better keep the edge information, such as the main disadvantages as filter size is difficult to choose, if choose larger multiple iterations affect time efficiency, whereas smaller noise reduction effect is poor.
Mean filtering
As shown in Fig. 1, the main principle of the algorithm average filtering method is to take the pixel value obtained by arithmetic average operation of N sampling values in the current neighborhood as the effective pixel value of the neighborhood center. The advantage of the mean filtering algorithm is that it has a stable range of fluctuating values for the random interfered signals, but its disadvantage is that it is difficult to choose the neighborhood range.
Results and analysis
Algorithm research and analysis
With the in-depth study of the traditional PM model, it is found that there are many deficiencies in the traditional classical model. In recent years, many literatures have also analyzed and studied the P-M model, and proposed a series of improved algorithms. Aiming at the problem that the traditional Perona-Malik (PM) anisotropic diffusion model cannot provide sufficient information through four-direction diffusion and the processing effect is not obvious, an improved eight-direction selective diffusion model is proposed in this paper. In view of the traditional algorithm’s failure to strong noise, the lack of empirical selection of diffusion threshold parameter K, and unclear iteration termination conditions, this paper proposes an improved diffusion coefficient calculation method, applies the adaptive selection of diffusion threshold parameter K, and adopts iteration termination criteria suitable for the improved algorithm. Through multiple sets of simulation experiments, it is proved that the algorithm in this paper can better control the diffusion process than the traditional model, and improve the robustness and efficiency of the algorithm. Combined with PSNR, FOM and other indicators, it shows that the algorithm has better noise reduction and edge retention effect than similar algorithms. Among them, Catté model has a good effect of noise reduction for ultrasonic images polluted by additive gaussian noise, and also solves the remaining mathematical problems in P-M model to some extent, so this algorithm is a relatively successful method in the improvement of P-M model. However, speckle noise of ultrasonic image does not have the symmetrical distribution structure of gaussian white noise, and the variance of gaussian function plays a decisive role in noise reduction effect. The uncertainty of variance is also the fatal defect of this method.
Analysis of founder test
From the performance indexes of PSNR and FOM in Table 1, it can be seen that the algorithm in this paper can achieve better denoising ability than other algorithms in Table 1, and at the same time better protect the image edge. It can be seen that, while removing noise, the algorithm can well maintain the edge details of the two images, and also retain rich texture details. The visual effect of the image is better than other algorithms. Considering the influence of Matlab on matrix operation optimization will lead to inaccurate evaluation of algorithm time, this paper only considers the cost of one iteration, and analyzes and compares the algorithm in this paper with SRAD algorithm. Iteration method a time of 0.382278 seconds, while SRAD about 0.346553 seconds, the iteration time SRAD time-consuming and time consuming method were similar, the number of iterations required to SRAD achieve PSNR peak is 1.5 times that of the algorithm in this paper, comprehensive iterations and each iteration time consuming factor, time needed for SRAD achieve optimal results can be interpreted as about 1.5 times that of the algorithm in this paper, thus this paper compared algorithm SRAD time performance is good. The algorithm occupies the space of an image storage space. Also follow the algorithm flow in the experiment:
Comparison of kidney image algorithms PSNR and FOM
Comparison of kidney image algorithms PSNR and FOM
First, for the contaminated image I, Canny edge detection algorithm was selected for edge detection to get the edge point M, and then cyclic iteration was started. According to the set M, judge whether the adjacent pixel point in the current direction of diffusion is an edge point; if so, set the value of the determinant factor n in this direction to 0, otherwise to 1; Using the method provided by Voci and Sugimoto et al. in the study, the diffusion threshold parameter K is adaptively calculated, and the obtained K is a monotone decreasing function, which can better adapt to the change of gradient threshold and has a good effect of edge preserving and denoising. Select diffusion coefficient diffusion in the texture detail area; In the noise area and the edge area, diffusion coefficient is selected for diffusion. Calculate the image pixel value after each iteration diffusion according to the improved diffusion model; Judge whether the iteration is completed according to the set increasing relative peak signal-to-noise ratio criterion; if so, terminate the iteration and output the image; otherwise, enter the next iteration.
The improved anisotropic diffusion based noise reduction algorithm for B ultrasonic image is proposed in this paper. The adaptive selection of diffusion threshold K reduces the step effect after treatment. By selecting the diffusion coefficient, each diffusion has the appropriate diffusion intensity in different regions. By analyzing the simulation results of different iteration times, the incremental PSNR iteration termination criterion is proposed, which can be stopped in time and has better effect, thus improving the efficiency and robustness of the algorithm. Experimental results show that the proposed algorithm is effective in noise removal and protection. Hold the image edge and texture details. The design of the system in this paper is finally integrated with the interface system, video management system and personnel management system. Therefore, when the actual project system integration is completed and the functions and safety and reliability are running normally, there is still room for optimization and improvement for the later practical maintenance. Ultrasound imaging technology, which USES high-frequency acoustic waves to obtain analytic images of soft tissues and organs, is a widely used form of medical imaging, which has very important applications in cardiovascular imaging, abdominal imaging, cardiology, urology and obstetrics and gynecology.