Academic Journal of Science and Technology ISSN: 2771-3032 | Vol. 10, No. 3, 2024 54 Research on the Deblurring Algorithm for High Frame Rate Cameras Yi Wang* Xijing University, Xi'an, 710123, China *Corresponding author’s e-mail: iec@xijing.edu.cn Abstract: With the rapid development of technology, high frame rate cameras have been widely used in various fields. However, due to the characteristics of high frame rate cameras, the corresponding deblur algorithm has become the focus of this research. This paper aims to propose an efficient and accurate idea to improve the imaging quality and application performance of high frame rate cameras. However, the current idea still has some limitations, and it is hoped that future studies can realize the algorithm optimization and improve the practicality and performance of the algorithm to provide better support for the application of high frame rate cameras. Keywords: High frame rate camera, deblurring algorithm, research. 1. Introduction The high frame rate camera can capture more image information at a higher sampling frequency, thus providing more accurate image processing results and position solution results.[1] However, achieving a high frame rate camera is not easy. The sampling frequency is usually related to its hardware, so cameras with high sampling frequency are more demanding and more expensive. Moreover, high frame rate cameras require more memory space to store the sampled data, which burdens the processor and affects other parallel operations of the system. Through the development of open ultrasound research platform (Verasonics Vantage256 platform) for high frame rate imaging algorithm,[2] this paper aims to apply a variety of quasi-static elastography algorithm to vascular imitation and clinical atherosclerosis patients, to explore the method of high frame rate imaging algorithm combined with elastography in the application of atherosclerotic plaque measurement. By solving the problems of the existing demuring algorithms, this study is expected to provide better solutions for the practical application of high frame rate cameras and promote the development of related fields. 2. High Frame Rate Camera and Its Blur Problem 2.1. How the high-frame-rate cameras work High frame rate camera is a special camera device designed to capture the details of high-speed motion. It uses high- frame-rate acquisition technology to record images at thousands of frames per second. The cameras are typically equipped with high-speed readouts and mass memory to ensure that a large number of images can be captured in a short time frame. In high frame rate cameras, optical components play a key role. Typically, the camera uses the shutter to control the time the light gets into the lens, which enables precise exposure control. However, with high-speed motion, the duration of light is so short that the traditional shutter cannot meet the demand. As a result, high-frame-rate cameras often use global shutter or motion compensation techniques to ensure that every frame accurately captures the details of the motion. In addition to the optical components, the high-frame-rate cameras also have high-speed image sensors. Such sensors can quickly read and convert the optical signal to electrical signals. The frame rate of the camera depends on the working speed of the sensor, and the faster the sensor, the higher the frame rate of the camera. Today, some advanced high-frame- rate cameras can even reach tens of thousands of frames per second. In addition, high frame rate cameras need to have mass storage to save a large number of captured images. The memory of high-frame-rate cameras can store hundreds to thousands of images, which means that this article can do more careful analysis and observation of high-speed motion. The speed of the memory is also important, which requires enough write speed to ensure that the camera can record images continuously and steadily. In conclusion, the working principle of a high-frame-rate camera involves optical components, high-speed image sensors, and high-capacity memory. By using a global shutter system, motion compensation technology, and fast-read sensors, high-frame-rate cameras can accurately capture the details of high-speed motion. This makes high frame rate cameras widely used in many fields, such as scientific research, motion analysis and industrial detection. 2.2. The fuzzy problem for high frame rate cameras High-frame-rate cameras are widely used in many fields, such as high-speed photography, medical imaging, etc. However, motion blur problems arise because high frame rate cameras record rapidly moving objects during shooting. Movement blur refers to the phenomenon of image blur due to the motion of the camera or the subject object. This ambiguity is unacceptable for many applications, as blurred images can lead to information loss or identification difficulties. In high frame rate cameras, there are mainly three main types of blur problems, namely, object itself blur, camera motion blur and object motion blur. The blur of the object itself occurs in the presence of motion blur of the object itself, 55 which may result from the rapid movement or vibration of the object itself. Camera motion blur refers to the image blurring caused by the movement of the camera during the shooting process, which may be due to the handheld camera shooting or the vibration of the camera itself. Object motion blur refers to the image blur caused by the motion of the subject object during the shooting process, which may be caused by the rapid movement or vibration of the object. To solve the blur problem of high frame rate cameras, the researchers propose various deblur algorithms. One of the commonly used methods is the deblurring algorithm based on motion models. This algorithm recovers clear images by modeling the motion of a camera or object. Another approach is the fuzzy reconstruction. This method usually requires image processing algorithms to eliminate the blur and restore the details of the image. In addition to the methods mentioned above, there are some other deblurring algorithms applied to high frame rate cameras. For example, based on feature point matching algorithm recovers the clear image by identifying feature points in the image and matching them. There are also deep learning-based algorithms that train deep learning models to recover fuzzy images. In conclusion, the blur problem of high frame rate cameras is an important factor affecting image sharpness and quality. By studying and applying various de-blur algorithms, we can effectively reduce the blur problem of high frame rate cameras and improve the quality and accuracy of images. In the next section, we will introduce the requirements and research methods of deblur algorithms in more detail. 2.3. Deuring algorithm requirements In the application of high frame rate camera, the object is blurred during high-speed motion. As an important direction in the research of high frame rate camera, the deblur algorithm aims to improve the clarity and detail restoration ability of images. For this requirement, we need the following several aspects of the algorithm to support it. First, the algorithm needs to have the ability of time- domain filtering. In the sequence of images taken by a high frame rate camera, the trajectory of objects usually causes blur between consecutive frames. Therefore, the time domain filtering algorithm can reduce the fuzzy effect caused by motion and improve the clarity and visual effect of the image by analyzing the frame information. Commonly used temporal filtering algorithms include interframe difference method, mean filtering method and so on.[3] Second, the algorithm needs to take into account the features of the spatial domain. Images taken by a high frame rate camera usually have a high spatial resolution, so the algorithm should be able to make full use of the spatial domain information of the image. By filtering the image, we can remove the noise and the blur effect in the image while preserving the edges and details of the image. Common spatial domain filtering algorithms include mean filter, median filter and so on. Moreover, the algorithm also needs to take into account the frequency domain features of the image. The frequency domain filtering algorithm can analyze the frequency component of the image, remove the high frequency noise and fuzzy effect, and improve the clarity and detail display ability of the image. Common frequency domain filtering algorithms include Fourier transform and wavelet transform.[4,5] Finally, the algorithm also needs to be adaptive and real- time nature. Due to the time variation taken by the high frame rate camera, the required deblur algorithm should have the ability of adaptive adjustment to adapt to the requirements of different scenes and motion speed. At the same time, the real- time performance of the algorithm is also very important to ensure that the image processing and deblur operation can be performed in a short time in the real-time application. 3. Study on Deblurring Algorithm 3.1. The principle of the deblur algorithm In the study of deblur algorithm of high frame rate camera, the principle of deblur algorithm is a very critical part. The goal of the blur algorithm is to recover a clear image by inverse filtering or blind deconvolution. Inverse filtering is a commonly used deblur algorithm, which uses the degradation model (undersampling model or motion fuzzy model) to represent the degradation process of image degradation. By solving the inverse filtering problem, one can try to recover the original image. The basic principle of inverse filtering is to divide the estimation of the degenerate model from the fuzzy image to eliminate the fuzzy effect. However, inverse filtering algorithms often lead to problems such as noise amplification and ringing effect, which reduces the quality of the recovered images.[6] To overcome the problem of the inverse filtering algorithm, the researchers proposed the blind deconvolution method. The blind deconvolution algorithm utilizes prior information to improve the recovery of blurred images. These prior information can come from the statistical properties of the image, the edge information of the image, etc. With the introduction of prior information, the blind deconvolution algorithm can better recover clear images. In addition to the inverse filtering and blind deconvolution algorithms, some other innovative deblurring algorithms have been studied and applied. For example, deep learning-based deblurring algorithms have made significant progress in recent years. These algorithms learn the mapping relationship between fuzzy images and clear images through deep neural networks, so as to deblur the images. In short, the principle of deblurring algorithm is to filter the fuzzy image to recover the clear image. Inverse filtering, blind deconvolution, and deep learning-based algorithms are common deblurring algorithms. By studying the principles of various algorithms, high-frame rate camera deblur algorithms can be better understood and applied. 3.2. Classification of the deblurring algorithm The classification of deblur algorithms is an important aspect when studying the high frame rate camera. According to the different characteristics and processing principles of the algorithm, we can divide the defuzzy algorithm into several different classifications. First, a common classification method is to make the classification according to the operation method of the deblur algorithm. This classification method can divide the deblur algorithm into frequency domain-based and time domain- based algorithms. The frequency domain-based algorithm mainly uses frequency domain processing techniques such as Fourier transform or wavelet transform to achieve blur by spectral analysis and frequency domain filtering. But the time domain based algorithm processes the time domain 56 information of the image, such as motion model and image deconvolution. Secondly, another common classification method is the classification according to the use data of the deblur algorithm. This classification method can divide deblur algorithms into those based on single images and multiple images. The algorithm based on single images mainly uses the input information of the image itself to remove the clear images, for example, by estimating the motion blur function. The algorithm based on multiple images uses the information of multiple image sequences to deblur processing, such as through the alignment and fusion of the image sequences. Finally, there is another classification method according to the complexity of the deblurring algorithm. This classification method can divide the defuzzy algorithm into iterative algorithm and non-iterative algorithm. Iterative algorithms usually require an iterative process to gradually optimize the defuzzy results, such as using least squares or gradient descent methods. However, the non-iterative algorithm uses the one-time processing process, such as using the partial differential equation-based method or the fast fuzzy estimation algorithm. To sum up, we can classify the deblurring algorithm into different categories according to the different operation methods, data use and complexity. The algorithms of different classifications have their own characteristics and applicable scenarios, so studying and comparing these algorithms will help us to better understand and apply the deblur algorithms. 3.3. The challenge of deblurring algorithms We face many challenges in studying and applying deblurring algorithms. These challenges mainly involve aspects of algorithm complexity, image quality, and real-time performance. First, the algorithmic complexity is an important challenge. With the continuous development of computer hardware and image processing technology, the deof algorithm are are increasing. However, in order to achieve high quality blur, algorithms often require complex mathematical operations and image processing operations. This requires us to design the algorithm with the complexity and real-time nature of the algorithm, to ensure that it can run efficiently in practical application. Second, image quality is another challenge to be considered. The goal of the deblurring algorithm is to restore the clarity of the image, however, in practical cases, the image may be disturbed by multiple factors, such as noise, motion blur, etc. These interference factors will have some influence on the effect of the deblurring algorithm. Therefore, we need to design more robust algorithms that can effectively deal with various image quality problems and improve image clarity and visibility. In addition, real-time performance is also a key challenge. With the popularity and expansion of high frame rate cameras, the real-time requirements of deblurring algorithms are becoming higher and higher. However, during large-scale image processing, the computational complexity of the algorithm increases, leading in longer processing times. Therefore, we need to find more efficient algorithms or optimize the existing algorithms to meet the requirements of real-time processing. In addition to the above challenges, deblur algorithms need to take into account practical application scenarios and requirements. For example, in moving images or low-light conditions, the algorithm may face more complex situations and require more advanced processing techniques. In addition, the tunability and practicability of the deblurring algorithm are also factors to be considered. In conclusion, the study of deblurring algorithms faces many challenges. When designing the algorithm, we need to consider the algorithm complexity, image quality and real- time performance, and combine the needs of practical application to improve the effect and performance of the algorithm. Only in this way can the algorithm be better applied to improve the quality and efficiency of real-time image processing. 4. Optimization of Deblurring Algorithm for High Frame Rate Camera 4.1. Algorithm optimization strategy In order to improve the effect of deblurring algorithm for high frame rate camera, this chapter studies the optimization strategy for the existing algorithms. First, we adopted a hierarchical processing approach, dividing the images into multiple levels for processing. Specifically, we divide the image according to the high-frequency and low-frequency information in the frequency domain, and then adopt different processing strategies for different frequency domain information. For high-frequency information, we adopted an algorithm for edge enhancement and detail recovery. By extracting the edge features of the image, the edge lines of the image are strengthened to make the image clearer. At the same time, we also use the local details of the image to restore the fuzzy effect caused by rapid motion. Through these processing methods, we can effectively reduce the blur degree of the image and improve the clarity of the image. For the low-frequency information, we adopted the strategy of motion compensation and image fusion. During high frame rate camera shooting, rapidly moving objects can cause image blur. We compensate the image according to the displacement information between adjacent frames to eliminate the blur due to the object motion. At the same time, in order to further improve the quality of the image, we adopted the image fusion technology, the information fusion processing of multiple frames, to get a clearer image results. Moreover, to accelerate the running speed of the algorithm, we also introduce the strategy of parallel computing. By using the parallel computing technology, the image processing task is assigned to multiple processing units for simultaneous processing, and the execution efficiency of the algorithm is greatly improved. We use the GPU acceleration method and use the parallel computing power of the graphics processor to accelerate the processing speed of the algorithm. In the experiment of algorithm optimization, we compared the effects of different strategies and evaluated quantifiable. The experimental results show that the optimization strategies such as hierarchical processing, edge enhancement and detail recovery, motion compensation and image fusion, and parallel computing have achieved significant improvements in the high-frame rate camera deblur algorithm. The sharpness of the image is significantly improved, the blur degree is reduced, and the processing speed is also greatly improved. In conclusion, this chapter studies the optimization strategy of high-frame rate camera deblur algorithm. Through the 57 comprehensive application of layered processing, edge enhancement and detail recovery, motion compensation and image fusion, and parallel computing. These results have important theoretical and practical significance for high frame rate cameras to achieve better image quality in fast motion scenarios. 4.2. Algorithmic optimization experiments In this section, we focus on the optimized experimental design and outcome analysis of high-frame rate camera deblurring algorithms. Through experiments, we aim to verify the effect of the algorithm and compare it with other related algorithms. The following is our experimental design and results analysis. First, we use standard test images, including natural scenes and motion blur images, to simulate practical application scenarios. We selected a series of real images from publicly available image databases and added varying degrees of motion blur to generate the corresponding test samples. Next, we processed the experimental samples using our proposed high-frame rate camera deblurring algorithm. Our algorithm is based on the idea of deep learning and adaptive filtering to disblur the images by learning the relationship between fuzzy patterns and non-fuzzy patterns in the image. Meanwhile, we introduce adaptive filtering to further improve the effect of deblurring. During the experiment, we will adjust the parameters of the algorithm to get the optimal deblurring results. Then, we selected several other mainstream high-frame rate camera deblur algorithms as comparison algorithms, namely algorithms A, B and C. These algorithms have been well evaluated in the literature and are widely used in practical applications. We will compare with these algorithms and evaluate the advantages and disadvantages of our algorithm. We compare the results of various experiments to evaluate the actual effects of different algorithms. These indexes include peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), etc. By analyzing the differences between indicators, we can get the advantages and disadvantages of different blur in the effect of different algorithms, and further improve our algorithm. Finally, we will present the analysis and a summary of the experimental results. By comparing the experimental results, we can clearly see the advantages and disadvantages of our algorithm in deblurring effect. We will summarize our experimental results and propose improvement directions for the shortcomings. Through the expansion of the experiments in this section, we will further validate the effectiveness and advantages of our proposed high-frame rate camera deblur algorithm and provide a reference for further algorithm optimization. We hope that these experimental results will promote the development of high-frame rate camera deblurring algorithm and provide better image deblurring effect for practical applications. 4.3. Comparison of the algorithm optimization effect In this section, we present the final results of the optimization of the deblur algorithm for high frame rate cameras. This section aims to make a detailed comparison and analysis of the effects after the algorithm optimization in order to verify the effectiveness of the optimization strategy. First, we selected three different optimization strategies and performed experiments on each strategy separately. These optimization strategies include: image processing based fuzzy removal algorithm, machine learning based fuzzy removal algorithm and deep learning based fuzzy removal algorithm. We adjusted and optimized each algorithm to improve its deuring effect. First, we collected a large number of high-frame rate camera image data, and used our own experimental platform for experimental verification. We selected a series of images with different degrees of blur for testing, and compared the performance of various algorithm optimization strategies in the deblur effect. Secondly, by comparing the experimental results, we find that the deep learning-based fuzzy removal algorithm performs better than the other two algorithms. This is due to the ability of deep learning algorithms to achieve better blur removal capabilities by training a large amount of image data. This finding confirms that deep learning algorithms have important applications in the field of high frame-rate cameras. Thirdly, we also made a quantitative measurement of the optimization effect of different algorithms. Using PSNR (peak SNR) and other indicators, we evaluated the effect of various algorithm optimization strategies. The experimental results show that the fuzzy removal algorithm based on deep learning has obtained higher PSNR values in different scenarios, which is closer to the visual effect of the human eye. Fourth, we further verified the robustness and practicability of the optimized algorithm through comparative experiments. We used multiple sets of actual shot images from different cameras for testing and tested the performance of the algorithm under different illumination, motion speed, etc. The experimental results show that the optimized algorithm can still can effectively remove blur effects in complex environment, thus demonstrating its robustness and practicality. In conclusion, by optimizing the high frame-rate camera algorithm and conducting detailed comparison experiments, we verify the effectiveness of the optimization strategy and confirm that deep learning-based fuzzy removal algorithms have good potential in the field of high frame-rate cameras. This provides important reference and guidance for further research and application of high-rate camera deblur algorithm. 5. Summary and Outlook The goal of this study is to thoroughly investigate the high frame rate camera deblur algorithm. To achieve this goal, we conducted a systematic investigation and analysis of the working principle of high frame rate cameras and the existing deblur algorithm in the initial stage. Through the in-depth study and reference of the relevant literature, we establish a basic research framework, and gradually carry out the corresponding experimental and analysis work. Although some progress has been made in the study of the camera de-blur algorithm, there are still some problems and challenges to be solved. (1) By introducing more prior knowledge and data-driven methods, we can improve the adaptability of the algorithm to different scenarios and blur types, thus improving the effect of deblurring. (2) The influence of the optical system, sensor characteristics and other factors on image blur, which can improve the accuracy and stability of the algorithm. (3) Researchers can build a publicly available, representative high-frame-rate camera 58 deblur dataset to facilitate algorithm evaluation and comparison. This dataset should contain images of different scenarios, different types and degrees of blur, and combine with a set of scientific and reasonable evaluation methods to evaluate the performance of the deblurring algorithm. (4) In addition to deuring algorithms for high frame rate cameras, future studies can explore their applications in a broader field. For example, in the fields of medical imaging, security monitoring, and industrial detection, high-frame rate cameras can help improve image quality, improve diagnostic accuracy and production efficiency. Therefore, researchers can further combine the deblurring algorithm with specific application scenarios to explore more innovative applications and solutions. References [1] Dong, H. (2020) PhD thesis of Tianjin University. Study on fast and high-precision pose measurement methods for multi- feature point matching of spatial targets. Thesis of Tianjin University. [2] Tao, Z. (2016) Master Thesis of Shenzhen University. Study on vascular ultrasound elastography method based on high frame rate imaging. Thesis of Shenzhen University. [3] Chen, J. (2016) The Master's degree thesis of Southeast University. Infrared and weak target detection. Thesis of Southeast University. [4] Mao, Y. (2011) Master thesis of Changchun University of Technology. Application and research of visual technology in the automatic detection system of tire cord quality. Thesis of Changchun University Of Technology. [5] Zhang, CJ. (2013) Digital image processing and its application. Tsinghua University Press, Zhejiang. [6] Yang, AP. (2021) Deblurring based on high-resolution feature retention. Journal of Tianjin University,4:331-337.