Reconstructing High Dynamic Range (HDR) video from image sequences captured with alternating exposures is challenging, especially in the presence of large camera or object motion. Existing methods typically align low ...
详细信息
ISBN:
(纸本)9798350353006
Reconstructing High Dynamic Range (HDR) video from image sequences captured with alternating exposures is challenging, especially in the presence of large camera or object motion. Existing methods typically align low dynamic range sequences using optical flow or attention mechanism for deghosting. However, they often struggle to handle large complex motions and are computationally expensive. To address these challenges, we propose a robust and efficient flow estimator tailored for real-time HDR video reconstruction, named HDRFlow. HDRFlow has three novel designs: an HDR-domain alignment loss (HALoss), an efficient flow network with a multi-size large kernel (MLK), and a new HDR flow training scheme. The HALoss supervises our flow network to learn an HDR-oriented flow for accurate alignment in saturated and dark regions. The MLK can effectively model large motions at a negligible cost. In addition, we incorporate synthetic data, Sintel, into our training dataset, utilizing both its provided forward flow and backward flow generated by us to supervise our flow network, enhancing our performance in large motion regions. Extensive experiments demonstrate that our HDRFlow outperforms previous methods on standard benchmarks. To the best of our knowledge, HDRFlow is the first real-time HDR video reconstruction method for video sequences captured with alternating exposures, capable of processing 720p resolution inputs at 25ms. Project website: https://***/HDRFlow/.
The proceedings contain 86 papers. The topics discussed include: robust real-time monitoring of complex human activities using multi modal video analytics;a robust approach for classifying laparoscopic video distortio...
ISBN:
(纸本)9798331506520
The proceedings contain 86 papers. The topics discussed include: robust real-time monitoring of complex human activities using multi modal video analytics;a robust approach for classifying laparoscopic video distortions using ResNet-50;enhancing x-ray image classification through neural architecture;revolutionary MRI imaging for Alzheimer’s: cutting-edge GANs and vision transformer solutions;advanced deep learning strategies for breast cancer image analysis;identifying surgical instruments in pedagogical cataract surgery videos through an optimized aggregation network;enhancing auxiliary cancer classification task for multi-task breast ultrasound diagnosis network;and bioinspired computer vision for effective extended reality applications.
As the scale of the power grid continues to expand, drone inspection operations are becoming increasingly popular. However, most of the existing inspection drones are for transmission line inspection in the open field...
详细信息
As the scale of the power grid continues to expand, drone inspection operations are becoming increasingly popular. However, most of the existing inspection drones are for transmission line inspection in the open field environment, with the characteristics of large size and high quality, which is difficult to be directly applied to transmission line inspection around the city area. To address the above issues, in this article, a small unmanned aerial vehicle (UAV) inspection system is designed with the aim of achieving autonomous inspection of overhead ground wires in urban peripheral areas, combined with imageprocessing technology, with a total weight of less than 400 g. Specifically, during the inspection of small UAV, Raspberry Pi uses traditional imageprocessing methods, such as Hough transform, to obtain the position information and pixel error of ground wire from real-timevideo stream;the flight control system uses the results of imageprocessing combined with data from millimeter-wave radar to achieve conversion from pixel error to actual distance error;finally, the ground wire is made to be in the center of the video as much as possible through the correction strategy, thus realizing the autonomous inspection task of the small UAV along the line. The experimental results show that the small UAV can stably identify the target transmission lines and achieve autonomous flight along the lines with horizontal deviation within plus or minus 0.3 m and height deviation within plus or minus 0.1 m, which is of great reference value for the application of small UAV in urban transmission line inspection.
Sketching is a uniquely human tool for expressing ideas and creativity. The animation of sketches infuses life into these static drawings, opening a new dimension for designers. Animating sketches is a time-consuming ...
详细信息
ISBN:
(纸本)9798331529543;9798331529550
Sketching is a uniquely human tool for expressing ideas and creativity. The animation of sketches infuses life into these static drawings, opening a new dimension for designers. Animating sketches is a time-consuming process that demands professional skills and extensive experience, often proving daunting for amateurs. In this paper, we propose a novel sketch animation model SketchAnimator, which enables adding creative motion to a given sketch, like "a jumping car". Namely, given an input sketch and a reference video, we divide the sketch animation into three stages: Appearance Learning, Motion Learning and video Prior Distillation. In stages 1 and 2, we utilize LoRA to integrate sketch appearance information and motion dynamics from the reference video into the pre-trained T2V model. In the third stage, we utilize Score Distillation Sampling (SDS) to update the parameters of the Bezier curves in each sketch frame according to the acquired motion information. Consequently, our model produces a sketch video that not only retains the original appearance of the sketch but also mirrors the dynamic movements of the reference video. We compare our method with alternative approaches and demonstrate that it generates the desired sketch video under the challenge of one-shot motion customization.
This paper presents the design and implementation of a client-server application for real-timevideo data transfer between two devices. A large number of real-timevideo data transfer techniques already exist. Movie a...
详细信息
The quality of images is crucial in image and video compression, especially for resource-constrained systems that prioritize simplicity. To achieve fast and low-energy compression, such systems aim to strike a balance...
详细信息
The quality of images is crucial in image and video compression, especially for resource-constrained systems that prioritize simplicity. To achieve fast and low-energy compression, such systems aim to strike a balance between image quality and computational complexity. While various Discrete Cosine Transform (DCT) approximations have been proposed, only two approximations with 14 additions are currently available. This paper presents a novel 8-point DCT approximation that improves image quality compared to the previous 14-addition transformations. Additionally, a pruned version is derived and shown to be efficient. The proposed approximation achieves an average quality gain of up to 1 dB while maintaining a similar computational structure to the previous transformations, resulting in comparable energy consumption. Therefore, this solution provides a compelling option for resource-constrained systems seeking efficient image compression while preserving high image quality.
The paper provided a brief analysis of video denoising characteristics, discussed and analyzed various existing video denoising methods, and proposed a new video denoising algorithm based on bidirectional time fusion ...
详细信息
The objective is to study the impacts of batting strength and angle of tennis players on batting results based on deep learning (DL) imageprocessing technology. A real-time evaluation algorithm of human motion is con...
详细信息
The objective is to study the impacts of batting strength and angle of tennis players on batting results based on deep learning (DL) imageprocessing technology. A real-time evaluation algorithm of human motion is constructed based on the camera videoimage and convolution neural network (CNN), and the selection of joint angles in volley training of tennis players is analyzed from the perspective of biomechanics. Gaussian Mixture Model (GMM), Visual Background Extractor (VIBE), and Optical Flow (OF) are introduced for simulation and comparison. Then, the proposed algorithm is applied to the volley experiments in areas A, B, and C of 6 tennis players (denoted by P1, P2, P3, P4, P5, and P6). The results show that the processing frame rate and batting and follow-up similarity score of the proposed algorithm based on the camera videoimage and CNN are significantly higher than those of GMM, VIBE, and OF. The return success rates of P1 in different areas are the highest, which are 75.46%, 75.62%, and 68.94%, respectively;while those of P6 are the lowest (19.55%, 17.46%, and 21.65%, respectively). The left ankle angle of P6 is much greater than that of P1, the angle of P1 is significantly lower than that of P3, P4, P5, and P6. The batting speed of P1 is significantly slower than that of P3, P4, P5, and P6, which is not much different from that of the left knee joint. The angles of the subjects' right forearm ring, left lower leg ring, and left thigh ring is obvious. Additionally, the displacement of the left foot of P1 and P6 in area A is 0.916m and 0.548m, respectively. Therefore, in the volley preparation stage, the left ankle angle (103-108 & DEG;) is greater than that of the right ankle (98-103 & DEG;);the tennis batting speed should be basically the same as that of the left knee joint to lower the gravity center of player. Thus, the proposed algorithm outperforms other algorithms in the volley experiment of tennis players.
When capturing videos with cameras, noise can occur due to variations in lighting conditions, movements of subjects or cameras, and the quality of camera sensors. The presence of noise complicates object detection and...
详细信息
processingimages using object detection, image restoration, and generative adversarial networks to directly convert real-world images into high-quality anime-style background images is one of today's research hot...
详细信息
暂无评论