11results about How to "Guaranteed visual quality" patented technology

The embodiment of the invention provides a video generation model training method, a video generation method, a medium and a program product, and the training method comprises the steps: carrying out the pre-training of a generation model based on a first training data set, determining a corresponding video generation model, and enabling the first training data set to comprise training video data divided into a plurality of video blocks; training data in a second training data set are input into the video generation model, corresponding video data are output, and the training data comprise audio data and reference pictures; and adjusting the video generation model based on a scoring model to obtain a trained video generation model. The block scheduling processing is realized through pre-training, the real-time performance and efficiency of the processing are improved, the knowledge of the scoring model is efficiently transferred to the video generation model, the problem of high model reasoning delay is solved, the visual quality can be maintained, the processing speed is remarkably improved, and the problem of long-term consistency of a long video is solved.

The invention provides a working face dynamic machine-following panoramic video stitching method and device. The method comprises the following steps: calling an original video stream of a target camera; wherein the target cameras are a plurality of cameras which are determined according to the position information of the coal mining machine and are within a preset range with the coal mining machine as the center; mapping the plurality of original video streams to a global coordinate system of the same working face according to the position information through a dynamic coordinate system conversion algorithm to obtain a machine-following panoramic video stream; performing local correction on the machine-following panoramic video stream through a local deformation correction algorithm to obtain a corrected video stream; fusing the overlapping areas of the adjacent cameras in the machine-following panoramic video stream through a multi-band fusion algorithm to obtain a fused video stream; the dynamic panoramic video which takes the coal mining machine as the center and continuously changes along with the movement of the coal mining machine is generated, and the remote control operation efficiency is remarkably improved. And the visual quality and naturalness of the dynamic panoramic video are ensured.

This invention discloses a method for generating missing modalities in brain MRI based on hypergraphs and attention mechanisms. This method addresses the issue of missing modalities in multimodal brain MRI sequences such as T1, T1ce, T2, and FLAIR in clinical scenarios by constructing a unified multi-input multi-output translation framework. The method introduces hypergraph convolution and region-level self-attention into the generative network, aggregating group-level features from different tumors and healthy tissues and modeling structural relationships between sub-regions, preserving the spatial tissue structure of the tumor. Combined with bidirectional Mamba sequence modeling, it enables the 2D network to efficiently capture long-range dependencies between layers of 3D volumetric data, ensuring voxel-level spatial continuity. Through a teacher-student knowledge distillation mechanism, the student network learns structurally perceptual features without a tumor mask, achieving high-fidelity generation. This method can improve the overall quality of generated images and the detail fidelity of tumor lesions, thereby enhancing the performance of downstream tumor segmentation tasks and showing promising application prospects.

The present application belongs to the technical field of medical equipment, and particularly relates to a progressive defocus hyperopia correction glasses and a design method thereof, which has a lens structure with three functional partitions, the lens structure comprising a base lens, a central correction area in a central part of the base lens, a ring-shaped transition stimulation area outside the central correction area, and a peripheral strengthening area outside the transition stimulation area. The design of the progressive hyperopia defocus amount of the present application conforms to the effect that the retinal morphology forms a stepwise increase of the hyperopia defocus amount when transitioning from the macular area to the peripheral retina, and has a good vision correction and hyperopia treatment effect: the central area forms a clear object image, without affecting the patient's amblyopia training; the central peripheral visual field is relatively clear, ensuring the visual quality; the peripheral retina forms hyperopia defocus, promoting the growth of the eye axis and the decrease of the hyperopia degree.

The utility model discloses a water stop pull rod for non-equal straight section wall construction, including inner rod, outer rod, water stop piece, cone nut, fender assembly and at least two fender nuts for fixing formwork, the middle part fixed water stop piece of inner rod, and the outer rod is connected through the cone nut between the inner rod to form the two -segment structure of detachable, and fender assembly includes the profiling pad piece for abutting bevel support and the plane gasket for abutting vertical support, and when two -segment structure is set in formwork, profiling pad piece movably sets up in the inner rod, and plane gasket movably sets up in the outer rod, and profiling pad piece is abutted to bevel support through the cooperation of one fender nut, and plane gasket is abutted to vertical support through the cooperation of another fender nut, so this can replace the three -segment water stop screw rod and silk screw rod of prior art, and guarantee the visual quality after wall forming.

This invention provides a wide-field-of-view 3D reconstruction method based on variable-space dual-core particles, comprising: acquiring multi-view RGB image data using underwater image acquisition equipment; initializing the acquired data using a motion recovery structure algorithm to generate point cloud data; establishing 3D Gaussian particles for the coordinate points on the initialized point cloud; and performing parameterization processing on the 3D Gaussian particles in Cartesian coordinate space to obtain parameter information such as their geometric center, 3D covariance matrix, normal vector, opacity, and color; establishing a Gaussian dual-core particle model; establishing a segmentation plane passing through the geometric center of the 3D Gaussian particles; establishing a Gaussian variable-space particle model; introducing a fourth-dimensional variable-space parameter for the Gaussian dual-core particles; and using the variable-space parameter to perform multi-parameter mapping of the Gaussian dual-core particles from Cartesian coordinate space to homogeneous coordinate space to complete the parameter enhancement and reconstruction of Gaussian particles in infinite space within the wide-field-of-view.

The application discloses a secret information embedding video method, a video steganography method and related devices, the application constructs a two-dimensional array of coding units, expands the original one-dimensional representation to a structured two-dimensional embedding space, significantly increases the number of candidate states available for each embedding operation, lays the foundation for high-capacity embedding, introduces polygon division, realizes region-based coding instead of point-based mapping, greatly improves the expression ability of the prediction unit division mode, so that more secret information can be embedded under limited structural modification, through the position relationship between the element S and the cover domain, the replacement set of the element S is screened out, the replacement element of the element S is determined based on the distance and is replaced to complete the secret information embedding, the distortion cost caused by different modification directions and structural disturbance can be more accurately simulated, the candidate replacement element is ensured to remain within an acceptable distortion range, so that the minimum distortion principle can be enforced, and the visual quality can be effectively maintained.

The invention relates to the technical field of 3D Gaussian splashing, in particular to a three-dimensional Gaussian compression method and system based on just perceptible distortion, and the method comprises the steps: collecting images, integrating and building a data set, and inputting the data set into a JND model to obtain a corresponding JND image; inputting the data set and the corresponding JND image into an anchor point-based 3DGS compression method model for training, sequentially obtaining the opacity and the JND score accumulation value of the anchor point, designing a loss function, and optimizing the attribute of the anchor point; judging whether the anchor point is pruned or not according to the opacity and the JND score accumulated value, and completing training of a 3DGS compression method model based on the anchor point; and the trained anchor point-based 3DGS compression method model outputs a rendered image. Extreme compression is achieved on the premise that human visual perception is not affected based on just perceptible differences, and an effective technical scheme is provided for storage and transmission of three-dimensional visual data.

This invention discloses an image data processing method based on heterogeneous multi-agent and frequency domain attention feedback, belonging to the fields of artificial intelligence security and digital image processing technology. The method acquires the image to be processed and real / generated category labels, constructs a heterogeneous agent detection model group containing a convolutional neural network and a visual Transformer; performs discrete cosine transform on the image to estimate frequency domain sensitivity and generate a frequency domain attention mask; adaptively constructs a frequency domain query subspace based on category prior, generates frequency domain integrated initialization data, and performs frequency domain decision boundary search on a target detection model that only returns hard labels; updates the attention mask and query subspace according to query feedback, and finally obtains a pixel-domain processed image that satisfies the perturbation budget through inverse discrete cosine transform. This invention can reduce the number of black-box queries, improve the processing success rate, and maintain image visual quality.

The invention belongs to the technical field of industrial collaborative augmented reality, and relates to an AR perception interaction system and method based on integration of computing and network, and the system comprises the following modules: a multi-modal perception collection module which collects a multi-modal original perception stream; the preliminary spatial relationship calculation module is used for generating a preliminary spatial relationship vector representing the local position of the equipment; the global collaborative feature construction module is used for constructing a global collaborative behavior feature matrix; the collaborative mode recognition and judgment module is used for generating a unique computing power and precision strategy instruction; the differentiated task distribution module is used for dynamically recombining the augmented reality content and generating and issuing differentiated collaborative rendering task packages; the self-adaptive virtual-real fusion display module is used for executing self-adaptive virtual-real fusion display; and the off-network self-sustaining collaboration module is used as a unique positioning reference and collaboration center for other off-network equipment in the field. According to the method, the problems that the high-precision cooperation requirement is difficult to meet, and the virtual-real fusion effect is poor due to position deviation accumulation among multiple devices are solved.

This invention proposes a method for converting between color and grayscale images. It can convert a color image to a grayscale image, significantly reducing storage requirements and concealing the true color image; conversely, it can convert the grayscale image back to a color image, restoring the true color image without visual distortion. By utilizing information hiding and data compression to store and transmit color images as grayscale images, the data volume can be reduced to one-third of the original. This ensures that the image holder can recover the original color image, while others can only obtain a grayscale image and cannot recover the original color image from it.