74 results about "Ghosting" patented technology

The invention relates to a video snapshot compression imaging reconstruction method and system. The method comprises the following steps: inputting a video frame sequence and a time-varying mask set thereof into a measurement model to obtain initial estimation; constructing a reconstruction network which comprises a feature extraction module, a gating residual network module and a video reconstruction module; the feature extraction module comprises two three-dimensional convolution layers, each three-dimensional convolution layer is connected with an activation function, and the feature extraction module extracts initial features from the initial estimation; inputting the initial features into a gating residual network module, and outputting reconstruction information features; and the video reconstruction module fuses the reconstruction information features, and performs up-sampling and detail refining to reconstruct a video sequence. According to the method, on the premise that parameters and computing power are hardly increased, ghosting and flickering are effectively restrained, the stability of long-time reconstruction is improved, and an effective scheme is provided for SCI reconstruction with the high compression ratio, the super-definition resolution ratio and the long sequence.

The invention discloses a dynamic scene-oriented multi-scale spatial-temporal feature fusion video super-resolution method, and belongs to the technical field of computer vision and image processing. The method comprises the following steps of: inputting an original video image into a trained video super-resolution network based on spatio-temporal feature fusion to obtain an optimized video image, and performing multi-scale extraction and fusion on video features by the video super-resolution network based on spatio-temporal feature fusion to obtain an optimized video image; according to the method, the problems of ghost shadow and optical flow caused by rapid change of an object are solved by utilizing a context residual learning network, and efficient feature extraction and fusion are realized by fully utilizing intra-frame space information and an inter-frame time relationship, so that the visual perception quality and the accuracy and robustness of super-resolution reconstruction are improved.

The application relates to a glass detection method based on deep learning and ghosting, and relates to the technical field of computer vision. The method comprises the following steps: performing glass detection based on a single original input image, the image being a single RGB image; extracting ghost features through a deep learning method of a backbone network based on the original input image to obtain a ghost area prediction map; connecting the ghost area prediction map with the original input image channel, extracting glass features based on the backbone network under the guidance of ghost clues, then performing glass feature decoding and glass area segmentation based on a convolutional neural network; and outputting a glass area prediction map. Compared with the prior art, the application performs glass detection based on a single image, and is more widely applicable. The backbone network can more accurately and efficiently extract ghost features and glass area features, and the ghosting phenomenon can be used to more accurately locate the glass area, obtain a high-quality glass area prediction map, and has good robustness.

The invention discloses a cross-domain HDR image dynamic ghosting removal method fusing statistical offset fuzzy membership, and the method comprises the steps: carrying out the ghosting removal of an LDR image sequence through employing a trained fuzzy statistical cross-domain ghosting removal model, and obtaining a ghosting-free HDR image; according to the model, after an input LDR sequence is subjected to preprocessing and feature extraction, a space context alignment module dynamically aligns non-reference frame and reference frame features through space-channel attention; meanwhile, the LDR image sequence is input into a statistical offset fuzzy membership module, and a membership graph is generated through sliding window statistical analysis and a fuzzy membership function; the aligned features are input into a multi-scale cross-domain cooperative processing module, multi-scale feature extraction and feature fusion are carried out, a frequency attention module is activated through membership graph gating, and global context modeling is introduced into a key region; and finally, outputting a high-quality ghosting-free HDR image. According to the method, fuzzy statistics and cross-domain processing are fused, and accurate ghosting removal and detail reservation in a dynamic scene are realized.

The application provides an image registration data processing method and system, relates to the technical field of image registration, and obtains a plurality of structural feature images in a plurality of monochrome laser images; according to the plurality of structural feature images, a plurality of ghosting areas with excessive eye movement are obtained, each of the ghosting areas includes a corresponding ghosting floating image and a ghosting reference image; a feature point corresponding relationship is obtained; a plurality of feature point pairs with accuracy greater than a preset threshold are obtained, feature point matching processing is performed, and a processed ghosting floating image is obtained; pixel registration processing is performed, and a region registration image is obtained; a plurality of region registration images are aligned and fused, a reference image and a floating image are combined, and a registration image is obtained. The application solves the technical problems of the prior art, such as inconsistent image offset in the image acquisition process, low whole-image registration accuracy, and the like, and achieves the technical effects of correcting the partial image offset caused by patient eye movement and improving real-time registration accuracy.

The invention discloses an image processing method, an electronic device, a vehicle and a computer readable storage medium, and the method comprises the steps: obtaining vehicle-mounted images of at least two adjacent cameras of the vehicle, and projecting foreground region images of the at least two adjacent cameras to a first plane when detecting that a foreground object exists in the vehicle-mounted images, and splicing the foreground area images projected to the first plane to obtain a spliced foreground image. Therefore, the foreground image can be determined on the basis of the projection of the foreground area image on the first plane, so that the situation of defects such as ghosting or dislocation in the foreground image obtained by splicing can be avoided to a certain extent, and the quality of a panoramic image obtained by the foreground image can be guaranteed; and the user experience of the panoramic monitoring image system can be guaranteed.

The invention relates to the technical field of mechanical conveying equipment, and discloses a machine vision-based electroplating defect real-time detection device, which comprises a conveyor belt, a following adjustment mechanism is arranged on one side of the conveyor belt, a protective shell is arranged on the outer side of the following adjustment mechanism, a camera device is arranged in the following adjustment mechanism, and the camera device is connected with the conveyor belt. A light supplementing device is further arranged in the following adjusting mechanism, an inductive probe is arranged on one side of the following adjusting mechanism, and a receiver is further arranged on one side of the following adjusting mechanism. According to the electroplating defect real-time detection device based on machine vision, through the arrangement of the following adjusting mechanism, in the movement process of the transmission belt, synchronous movement of an electroplated part and the camera device is achieved, the situation that the production rhythm is interrupted by the traditional process of stopping, detecting and re-conveying is avoided, the overall production efficiency is improved, and the production cost is reduced. Moreover, the problems of blurring and ghosting caused by relative movement are reduced, and the conditions of false detection and missing detection caused by poor image quality are reduced.

The invention discloses an electronic ink screen local refreshing method and system based on space-time prediction and waveform optimization, and relates to the technical field of electronic ink screen local refreshing, and the method comprises the steps: collecting user behavior characteristics and screen state data, inputting the data into a user behavior prediction module, and outputting a dirty region, a potential modification region and prediction confidence thereof; the pre-rendering candidate region is combined with the prediction confidence degree and then converted into a standardized data structure to be issued; the intelligent waveform optimization module outputs a target waveform parameter based on the extracted key features of the dirty region and the prediction confidence; the ghosting elimination module generates a reverse compensation waveform according to the target waveform parameter and the information of the dirty area, is used for eliminating ghosting, and finally indicates the display controller to refresh a screen and update display content; according to the refreshing method, the characteristics of no backlight and no stroboflash of the electronic ink screen are fully exerted, and a healthy programming learning environment is provided for young children; and asthenopia and visual impairment risks caused by long-time use are reduced.

Systems and methods for generating autocomplete text using a language model are disclosed. An image and text-prefix may be entered at an input field of a search application. The image is processed to generate an image description. The image description and the text-prefix signals may be used as input at a language model to generate an autocomplete text by the language model. A contextual history may also be included as input to the language model. The autocomplete text is an output by the language model based on the input at the language model. The auto-complete text may be a next-word ghosting.

The invention discloses a PS-OCT polarization artifact suppression method based on a cross coupling coefficient. Comprising two parts of pre-measurement based on matrix calculation and PS-OCT data processing based on a cross coupling coefficient, according to the pre-measurement, a dual-channel interference signal is obtained by scanning a wave plate through a PS-OCT system, and then a ghosting artifact displacement set and the cross coupling coefficient are obtained; in the data processing step, a PS-OCT data processing method based on a cross coupling coefficient is used for detecting a sample and processing a dual-channel interference signal, and finally a PS-OCT image without polarization artifacts is obtained. Through the cross coupling coefficient, the cross coupling effect generated by various polarization elements can be quantified, and accurate suppression of polarization artifacts is achieved; the cross coupling coefficient is obtained in advance through the detection wave plate, in actual imaging, the cross coupling coefficient does not need to be measured any more, and the speed of suppressing polarization artifacts is increased; on the premise that the original configuration of the system is not changed, the polarization artifacts are comprehensively suppressed.

The application discloses a panoramic depth high dynamic range imaging system and method based on compound eye structure, relates to the technical field of computational imaging and image processing, and synchronously collects and multi-scale motion estimation ensures the time sequence consistency between sub-images, effectively reduces the splicing dislocation and ghosting caused by the rapid movement or partial occlusion of an object, and improves the stability of a dynamic scene; an adaptive HDR fusion mechanism based on exposure difference and local features preferentially retains high-contrast unsaturated region details through intelligent weight distribution, overcomes the overexposure and underexposure problems under backlight or light mutation, and makes the output image naturally smooth in the bright-dark transition area; in addition, multi-scale depth synthesis and edge perception filtering work cooperatively, enhances the depth estimation accuracy of a low-texture area, avoids depth jumping and edge blurring, and guarantees the overall definition of the panoramic depth image; meanwhile, rolling shutter correction and space-time interpolation compensation further reduce dynamic ghosting and noise interference.

The invention relates to a millimeter-wave-based local amplification display device and a control method, and the method comprises the steps: obtaining three-dimensional point cloud data of a human head through the scanning of a millimeter-wave radar, extracting the contour features of the human head, calculating and generating a circular frame parameter matched with the contour of the human head, and carrying out the overlapping display of a circular frame on a display module according to the circular frame parameter. The circular frame is used for defining the boundary of the local amplification area; performing real-time image processing on the picture in the selected area of the circular frame through the FPGA chip; by applying the method, the aiming precision and the game winning rate can be improved, the global view field can be kept, the'tunnel view field 'effect can be avoided, meanwhile, self-adaptive personalized adaptation is carried out according to the shape of the head of each person, the comfort can be effectively improved, operation is convenient and fast, and fine adjustment is supported; in addition, the scheme is based on hardware-level real-time processing, low delay and high smoothness are achieved, and jamming or smear caused by processing delay is avoided; and the game fairness can be maintained through the non-intrusive design.

The embodiment of the invention relates to a video interpolation frame generation method and device, display equipment and a storage medium, and the method comprises the steps: segmenting a current frame of a target video into a current frame main body layer and a current frame background layer, and segmenting a reference frame into a reference frame main body layer and a reference frame background layer; determining the matching degree between each background information block in the background layer of the current frame and each background information block in the background layer of the reference frame, dividing the background information blocks into repeated background information blocks and newly added background information blocks according to the matching degree, and splicing the processed background information blocks to generate a background layer interpolation frame after fusion processing; generating a main body layer interpolation frame for the current frame main body layer and the reference frame main body layer; and generating a complete interpolation frame according to the background layer interpolation frame and the main body layer interpolation frame. Therefore, the phenomena of dislocation, ghosting and void of the main body and the background in the video picture are remarkably reduced, and the visual coherence and the motion stability of the interpolation frame and the utilization efficiency of computing resources are improved.

The invention discloses an NPU-based expressway monitoring video high dynamic range fusion enhancement method and system, and relates to the technical field of image enhancement processing, and the method comprises the steps: obtaining RAW data frames outputted by an image sensor supporting the LOFIC technology in a single physical exposure period; extracting a bright frame, a middle frame and a dark frame at the same exposure moment to form three RAW subframes; respectively carrying out ISP preprocessing to obtain three frames of RGB; splicing the three frames of RGB into 9 channels for input according to channels, inputting the neural network model operated by the NPU for fusion and enhancement, and outputting 3-channel high-dynamic range RGB; and encoding the output image and outputting to a storage or transmission interface. Through the technical scheme of the invention, the loss of details such as ghosting, overexposure and underexposure is reduced, and the monitoring imaging stability of tunnel entrances, backlighting and night vehicle lamp scenes is improved.

The application discloses a method and device for removing ghosting, an electronic device and a computer readable storage medium. The method comprises: determining a ghosting area in a to-be-processed image through a ghosting detection model; determining target ghosting pixels in each ghosting sub-area according to a reference image, wherein the reference image and the to-be-processed image come from the same video stream, the ghosting sub-area is obtained by dividing the ghosting area, and the number of the ghosting sub-areas is more than two; screening a target ghosting sub-area according to the number of the target ghosting pixels in each ghosting sub-area; and replacing each target ghosting sub-area in the to-be-processed image with a corresponding area of the reference image to obtain a first target image. Through the application, the damage to the original image background during ghosting removal can be reduced, and the difference between the image after ghosting removal and the original image can be avoided.

The invention relates to the technical field of computer vision, and discloses a multi-camera video stitching method based on regionalized symmetric weighting. The method comprises the following steps: realizing multi-camera frame-level synchronous acquisition (time sequence jitter is less than or equal to 10ms) through an event-driven multi-thread mechanism; obtaining distortion parameters (k1, k2 and k3) by adopting a Zhang Zhengyou calibration method, and correcting the image in real time according to a nonlinear model; dynamically constructing a fusion region: firstly expanding pixels of an overlapping region We belonging to [100, 200] to compensate a registration error, then generating an optimal suture line through a graph cutting method, forming a suture line region through morphological expansion, and finally dividing the optimal suture line region into a suture line region and a transition region; during partition fusion, an image source with the minimum pixel difference is directly selected in a suture line area to eliminate ghosts, and a dynamic boundary weight strategy is adopted in a transition area to achieve seamless transition. Finally, two fusion problems of ghosting elimination and splicing seam elimination are realized based on the two, image transition becomes smoother and more natural, and the method is suitable for panoramic security and protection monitoring and the like in industrial complex scenes such as wind power and the like.

The information processing apparatus includes a motion compensation unit, an occlusion region detection unit, a ghosting removal unit, and a DNN. A motion compensation unit performs motion compensation on past frame estimation results and generates a corrected history frame image. The occlusion region detection unit detects an occlusion region in which ghosting may occur in the correction history frame image on the basis of a motion vector representing the motion of the subject between the frames. The ghosting removal unit performs ghosting removal processing with respect to the occluded region. The DNN estimates the current frame based on the corrected historical frame image in which the ghost removal process has been performed and the current frame image to be processed.

Techniques are disclosed for denoising videos. In some embodiments, video frames are denoised using a denoising model that includes an encoder-decoder architecture and attention modules. During training of the denoising model, the attention modules learn weightings to upweight certain dimensions of input features to help pixel registration, remove ghosting artifacts, and improve temporal consistency when the frames of a video are being denoised. The denoising model can also be used to train a student denoising model that has a same architecture as, but is smaller and faster than, the denoising model. After training, noisy video frames can be input into the denoising model and / or the student denoising model to generate corresponding denoised video frames.

Provided is a lens device, an imaging device, and an optical filter unit that can suppress the generation of ghost images and light spots. A lens device (100) includes, in order from an object side in an optical path, a first optical filter (134A, 134B) and a second optical filter (136A, 136B). The first optical filter (134A, 134B) is composed of an optical filter (e.g., a band-pass filter) having a light transmission band in a specific wavelength region. The second optical filter (136A, 136B) is composed of an optical filter (a band-stop filter) having a light absorption band in a wavelength region different from the light transmission band of the first optical filter.

The embodiment of the invention provides a multi-exposure image fusion method and device, electronic equipment and a computer readable storage medium, and relates to the technical field of image processing. According to the method, distinguishing and modeling of a moving object and a static background in a dynamic scene are achieved by generating a moving area weight map, a median threshold bitmap, a superpixel area map and a basic weight map. The median threshold bitmap and the superpixel area map are used for correcting the motion area weight map, the space consistency and the boundary fitting degree of motion detection are improved, motion interference in the basic weight map is eliminated through the final motion weight map, and an accurate fusion weight map is obtained. On this basis, multi-exposure fusion is completed in combination with an original image sequence, ghosting artifacts caused by object movement can be effectively suppressed while highlight and shadow details are reserved, and the quality and authenticity of the fused image are improved.

The application discloses a multi-path video real-time splicing processing method based on an embedded platform and particularly relates to the technical field of image processing, which comprises the following steps: obtaining multiple images with consistent time stamps through hardware frame synchronization, calculating the light flow module length of the overlapping area after pre-alignment through cylindrical projection to generate a binary mask, establishing a control grid in the overlapping area to perform sparse disparity estimation and correction to obtain a sparse disparity map, up-sampling the disparity map to a dense disparity map, constructing a composite energy map containing a spatial color gradient cost and a time penalty cost, modulating the spatial cost by using the binary mask and smoothing the time penalty map, obtaining a dynamic avoidance seam through shortest path search, and forming an adaptive feathering band with the seam as the center. The application effectively eliminates static disparity ghosting and motion object half-cut artifacts under limited computing resources and realizes high-quality real-time panoramic splicing.

The present application proposes a video frame deletion forensics method based on image copy-move feature, which uses the similarity between ghost features in video frame deletion tampering residual error map and image copy-move tampering features to conduct video frame deletion forensics. First, a Multi-PyRes model is designed, and the model is pre-trained on an image copy-move tampering dataset to learn the features of copy-move tampering. Then, the pre-trained model parameters are migrated to the frame deletion detection task. The method extracts copy-move features from the saliency residual feature sequence map of the frame deletion video by introducing a constraint convolution layer that can learn copy-move features, and uses a residual pyramid convolution layer structure combined with a CBAM attention mechanism to extract copy-move tampering features at multiple scales, thereby realizing video frame deletion tampering detection. The present application extracts copy-move tampering features similar to ghost features of frame deletion tampering at multiple scales for image classification, which can realize video frame deletion tampering detection.

The invention relates to an image ghosting processing method and device, equipment and a storage medium. The method comprises the following steps: before shooting an image by using an image acquisition device, obtaining a preview data stream of the image acquisition device for a current shooting scene, and obtaining continuous multi-frame first images from the preview data stream; determining whether the current photographing scene is a highlight motion scene according to the multiple frames of first images; under the condition of determining that the current photographing scene is the highlight motion scene, controlling the image acquisition equipment to acquire multiple frames of second images in an exposure mode; and performing ghosting removal processing on the multiple frames of second images to generate a target image. According to the method, judgment is carried out through a highlight motion scene, it is ensured that image data needing to be processed are within a dynamic range, basic data of ghosting existence are provided, meanwhile, multiple frames of second images are collected under the same exposure condition, the ghosting situation in the multiple frames of second images is more prominent, the ghosting removal accuracy is improved, and the ghosting removal efficiency is improved. And higher information quality of the target image is ensured.

The invention provides a video frame rate adjusting method and system based on an LED display device, a medium, a program and a terminal, different frame compensation strategies are selected according to a second motion intensity value, and when low motion is needed and only one frame needs to be inserted, frame repetition, zero motion estimation and zero additional delay are adopted to ensure that the rhythm is natural and the real-time performance is met; and switching to motion compensation frame insertion when the system is in medium-high motion or needs multi-frame insertion, reconstructing an intermediate state according to displacement, continuously restoring a motion track, and inhibiting jitter and trailing. Moreover, when the motion compensation frame insertion strategy is executed, afterglow compensation is carried out on the insertion frame, so that the phenomena of smear, ghosting, local brightness overexposure and the like caused by the inherent afterglow effect of LED pixel points in the LED display device when a high-speed motion picture is played are avoided.

The application provides a multi-person real-time action recognition method based on a millimeter wave radar, multi-person action recognition is performed by receiving reflected millimeter wave signals in a recognition environment and processing real-time collected point cloud data and Doppler data, the number of detected objects between two consecutive frames is detected to eliminate transient ghosting, and the appearance of a new object is effectively perceived, then the ghosting mirror characteristic is used to eliminate the ghosting of the newly appeared object for a plurality of consecutive frames, so that consecutive ghosting is eliminated, the Doppler data of a plurality of detected objects is segmented based on a continuous sequence of actions, local segmentation makes the data field more focused on the action, and the position change of the action with time is more intuitively reflected, so as to be beneficial to action recognition, and the real speed of the detected object is calculated according to the speed index value with the largest intensity, so that the speed dimension in the Doppler data is compressed, the number of parameters of the neural network is reduced, and the calculation complexity is reduced.

The invention provides laminated glass, a head-up display system and a vehicle. The laminated glass is provided with a projection display area, and the projection display area is provided with a wedge-shaped structure; a first relation curve exists between a vertical ghosting value observed in the upper eye box area corresponding to the farthest virtual image surface and a wedge angle of the wedge-shaped structure; a second relation curve exists between a vertical ghosting value observed in the lower eye box area corresponding to the nearest virtual image surface and the wedge angle of the wedge-shaped structure; the wedge angle of the wedge-shaped structure is a wedge angle corresponding to the intersection point of the first relation curve and the second relation curve. The wedge angle at the intersection point of the first relation curve of the upper eye box area associated with the farthest virtual image surface and the second relation curve of the lower eye box area associated with the nearest virtual image surface is defined as the wedge angle of the wedge-shaped structure, so that the problem of ghosting eliminating conflict caused by relatively large imaging distance difference of far and near images is solved, and ghosting tolerance is reduced; the image definition is improved, and the effect that double images and main images of far and near images are overlapped by adopting the laminated glass with a single wedge angle is achieved.

A VR headset, belonging to the field of head-mounted display technology, addresses the issue that in gaming and entertainment scenarios, the 3D movement of the user's head can easily cause the VR headset's headband to slip. Combined with device play and inertial jitter, this can lead to overall device misalignment and loosening, resulting in optical focusing distance misalignment, blurred images, ghosting, and dizziness. Existing lens adjustment solutions cannot solve the device shifting problem, and long-term use can easily induce latent strabismus. Furthermore, significant offline dizziness can occur when switching between virtual and real scenes. This invention includes a display device with a fixed connecting base on its top. This invention can adaptively follow the 3D movement of the head, solving the headband slippage and overall device misalignment problems, maintaining stable optical focusing, avoiding blurring, ghosting, and dizziness, reducing the risk of strabismus, alleviating offline dizziness, improving wearing comfort, reducing component wear, and adapting to high-motion scenarios such as gaming and entertainment.

The invention provides a ghosting detection method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring a to-be-detected image; performing defect region extraction on the to-be-detected image to obtain a candidate defect region set; dividing the to-be-detected image to obtain a plurality of image partitions; for a candidate defect region pair formed by any two candidate defect regions in the candidate defect region set, determining a relative azimuth angle of the candidate defect region pair; determining an image partition to which the candidate defect region pair belongs, and obtaining a ghosting feature condition corresponding to the image partition to which the candidate defect region pair belongs; according to the ghosting feature condition and the relative azimuth angle, a ghosting detection result is determined, and the ghosting detection result is whether the candidate defect area contains ghosting or not. According to the scheme, efficient and accurate ghosting detection is realized.