81 results about "Kernel (image processing)" patented technology

This invention provides a method for selecting strong scattering points in space-based ISAL systems by incorporating multi-dimensional features, relating to the field of ISAL image processing. This invention employs a multi-stage fusion processing strategy to form a complete link from image preprocessing to the final output of strong scattering points: pixel replication and extension are used to process convolutional boundary regions, significantly reducing boundary amplitude distortion caused by traditional zero-padding or periodic extension, and improving the signal-to-noise ratio of edge regions in the filtered amplitude map; the dynamic threshold generation process works synergistically with the preceding Gaussian kernel convolution preprocessing to suppress speckle noise while preserving the true amplitude characteristics of weakly scattering targets; amplitude-weighted averaging is used to calculate the centroid coordinates within the connected domain, overcoming the energy diffusion problem caused by the ISAL system's point spread function; transient interference points are eliminated based on the peak-to-mean ratio of multi-frame amplitude sequences, and the spatial distribution characteristics of the centroid position's spatial amplitude value and density constraints are combined to ensure that the output set of preferred scattering points simultaneously satisfies temporal stability, spatial saliency, and uniform distribution.

This invention provides a novel raw domain denoising method and system, belonging to the field of digital image processing technology. The method includes: obtaining normalized raw image data and corresponding noise intensity parameters; determining the filter kernel size parameters corresponding to each pixel position, and simultaneously performing multi-directional structural analysis on the normalized raw image data to generate direction parameters and direction confidence parameters; constructing a direction-adaptive filter kernel, and performing filtering processing on the normalized raw image data to generate low-frequency denoised image data; calculating high-frequency residual information and determining high-frequency compensation parameters; performing high-frequency component compensation on the low-frequency denoised image data to generate denoised raw image data, and performing inverse normalization to output the raw domain denoising result. This invention achieves adaptive adjustment of the noise suppression process by combining noise intensity modeling and multi-directional structural perception within the raw domain, maintaining the stability of image structural information while ensuring effective noise suppression.

The application discloses a marine organism image processing method and device, electronic equipment and a readable storage medium, and is applied to the technical field of digital image processing. The method comprises the following steps: performing step-by-step down-sampling on a to-be-processed optical image by using a convolution module with different convolution kernels, so as to obtain a plurality of initial images with different scales. Global feature extraction is performed on each initial image under different scales, so as to obtain a plurality of multi-level feature maps; the multi-level feature maps are input into a Transformer encoder, and multi-level detail feature extraction is performed on the multi-level feature maps by adopting a multi-level feature deepening extraction fusion mode. The output features of the Transformer decoder and the multi-level detail features are up-sampled, so as to obtain an enhanced optical image. The application can solve the problem that the underwater image degradation phenomenon is serious or the details are blurred, and effectively improve the image enhancement effect of the marine organism image.

The application discloses a kind of municipal drainage pipeline image recognition and analysis system, it is related to image recognition technical field, the system includes: data acquisition module gathers pipeline inner wall image, water quality, flow and equipment speed data;Scene variable coupling module is constructed model by dynamic weight calculation and multi-peak membership analysis, and outputs comprehensive scene identification;Image enhancement collaborative module is sequentially self-adapting enhanced image according to filtering, standardization, correction, sharpening;Defect identification analysis module extracts feature, classifies defect and generates structured report;Adaptive iteration module compares identification rate threshold, and optimizes and updates parameter library;The application constructs multivariate coupling model, combines sewage noise adaptive filter kernel formula and dynamic distortion correction strategy, solves the problem that traditional system scene is not adapted, image processing is not good, and identification is not accurate, adapts diversified pipeline detection demand.

The application belongs to the technical field of image processing, and relates to an extruded pipe surface defect detection method and system based on machine vision. The method acquires an original gray image of the extruded pipe surface, extracts texture energy response by using a multi-direction filter set, acquires local maximum, minimum and global average texture energy response values; constructs a ratio operation according to the response values to calculate a local spatial anisotropy index; utilizes the nonlinear attenuation characteristics of an exponential function to dynamically map a basic scale constant based on the local spatial anisotropy index, and calculates a dynamic smoothing scale of each pixel point; uses a Gaussian kernel function corresponding to the dynamic smoothing scale to perform adaptive smoothing processing on the image and constructs a Hessian matrix, solves eigenvalues, and combines a reference background curvature constant to calculate a confidence enhancement response, so as to extract real crack defects. The application realizes noise suppression and weak crack reservation, and improves the accuracy of pipe surface defect detection.

The application relates to an image processing method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring initial RAW image data and corresponding gyroscope data; converting the initial RAW image data into initial YUV image data; determining a shaking track of an image sensor when collecting the initial RAW image data according to the gyroscope data based on a positional relationship between a gyroscope and the image sensor; determining a blur kernel corresponding to the image sensor according to the shaking track of the image sensor; and performing deblurring processing on Y channel data in the initial YUV image data according to the blur kernel to obtain a target YUV image. The method can reduce calculation redundancy, improve operation speed, and obtain a target YUV image with higher definition.

This invention discloses an electronic image sharpening method based on image processing, comprising: mapping the image to an initial continuous signal function continuously distributed in coordinates; constructing a liquid time constant prediction branch and generating liquid response parameters based on local gradient and contrast features; reconstructing the discrete state transition matrix into a dynamic evolution kernel, constructing a liquid state space model architecture and performing bidirectional scanning to obtain a latent space state feature sequence; calculating an edge gain adjustment factor and coupling it to the input weight matrix to generate an evolutionary feature tensor; performing spatiotemporal smoothing correction in conjunction with sensor motion parameters; and outputting the electronic image through spatial dimension recovery and implicit function resampling. This invention achieves enhanced edge details while suppressing background noise, eliminating displacement deviations and artifacts, and significantly improving the sharpness and structural fidelity of electronic images.

The application discloses a dark underwater image processing method based on color space conversion, which comprises the following steps: firstly, pre-acquired underwater images are down-sampled to reduce the image size and copied into two images; secondly, color space conversion is performed on the two parts of underwater images respectively; thirdly, an underwater optical imaging image model is established, and the first converted underwater image is processed by using an image deblurring algorithm to obtain a clear image with high contrast; fourthly, an improved white balance algorithm is used to process the deblurred image; fifthly, a nonlinear channel prior is introduced into a traditional image deblurring model to minimize the item to eliminate the motion blur problem caused by water flow fluctuation; sixthly, the obtained accurate blur kernel and potential clear image are iterated with the initial restored image to obtain a final restored image; and finally, the two images are fused, re-sampled and outputted to obtain a final clear underwater image. The application can better realize image restoration, improve the brightness and contrast of the image and improve the recognition degree.

This invention belongs to the field of medical image processing technology, specifically relating to a brain tumor segmentation method based on a dynamically enhanced diffusion model. The core design incorporates a dynamically convolutional kernel-driven multi-scale feature adaptive fusion mechanism implemented in a lightweight dynamic attention denoising UNet network (LDA-DU). Simultaneously, a lightweight global enhancement feature encoder LG-FE is designed to provide a high-quality multi-scale feature base for this core mechanism, and a step-size uncertainty fusion module adaptively fuses the multi-step outputs of this core mechanism. Through this end-to-end architecture of "core mechanism + supporting modules," accurate 3D brain tumor segmentation (WT / TC / ET) is achieved. This simultaneously addresses the imbalance between lightweight design and accuracy, and the low efficiency of multi-step prediction, improving the model's robustness to non-ideal clinical data and providing technical support for rapid and accurate clinical diagnosis of brain tumors.

The image processing apparatus comprises at least a first arithmetic unit and a second arithmetic unit. The first arithmetic unit and the second arithmetic unit are cascaded by means of a serial transceiver, and are configured to execute the image processing method in parallel. The image processing method comprises: acquiring sub-images obtained by segmenting an original image by an external processor, and using the sub-images as current-level result images; extracting current-level image elements in the current-level result images according to a specified convolution kernel and the segmentation mode in which the sub-images are obtained; sending the current-level image elements to a front-level arithmetic unit, and receiving a back-level image elements sent by a back-level arithmetic unit; synthesizing the current-level result images and the back-level image elements into current-level images to be processed; and performing convolution operation on said current-level images to obtain the current-level result images.

The application belongs to the technical field of image processing, and particularly relates to an underwater image restoration method based on an adaptive multi-scale large kernel attention module. An obtained underwater image is decomposed into a low-frequency component and three directional high-frequency components through Haar discrete wavelet transform; a mixed domain attention module is introduced at a highest resolution stage of the underwater image decomposed through the Haar discrete wavelet transform; a composite shape convolution module is added at a bottleneck position of a high-frequency branch; a PolyKernel main body is used as a large kernel encoder-decoder in a low-frequency path, and an adaptive multi-scale large kernel attention module is introduced at a bottleneck stage; after independent processing of each frequency branch, the network re-fuses high-frequency and low-frequency features into a spatial domain through inverse wavelet transform; and a light-weight unified output refining-adjusting head is used to perform structural refining and tone normalization on the fusion result.

The present application relates to the field of computer image processing, in particular to a feature enhancement and self-adaptive regression tea leaf picking point positioning method and system. The basic principle of the method is: firstly, the collected tea bud image is preprocessed for clarity; then, a two-stage model architecture of detection first and then positioning is adopted, an improved YOLOv5 network is used in the detection stage to robustly detect multi-scale buds and suppress background interference; then, the target suitable for picking is screened out and cut into a single bud image; in the positioning stage, an improved YOLOv11-Pose network integrated with an adaptive convolution kernel module is used to accurately regress the picking point coordinates; finally, the coordinates are mapped back to the original image and output. The core technical effect of the present application is: through the synergistic optimization of the two-stage process and the targeted improvement of the model components, the problem of inaccurate picking point positioning and poor robustness caused by image degradation, multi-scale targets, complex background and variable bud morphology in the natural environment is effectively solved, providing a high-precision solution for tea leaf automatic picking.

This application discloses an image processing method and related apparatus. The method acquires a corresponding undetermined pixel image, which is obtained by mapping an original depth image into pixel space, thus forming the original pixel image. Then, it identifies hole regions in the undetermined pixel image and determines a target filling kernel based on the attribute characteristics of these hole regions. Finally, the target filling kernel is used to fill the hole regions, resulting in the target pixel image. Since the attribute characteristics of the hole regions reflect their features, the target filling kernel determined accordingly is the most suitable for them, effectively repairing the hole regions in the undetermined pixel image after hole filling. Therefore, the target pixel image has improved image quality compared to the original undetermined pixel image, and can be directly used to identify the object to be recognized, resulting in higher recognition efficiency.

The present application relates to the technical field of image processing, and particularly relates to an image processing method, an electronic device and a readable storage medium, the image processing method comprises the following steps: fusing and blocking a current frame and a reference frame to obtain a plurality of to-be-encoded spatial image blocks, and combining the to-be-encoded spatial image blocks into a plurality of macroblocks, and configuring a corresponding encoding quantization table and a decoding quantization table for each macroblock; performing discrete cosine transform by using a first integer DCT transform kernel; performing quantization by using the encoding quantization table; performing entropy encoding and storing into an off-chip memory; reading a compressed code stream from the off-chip memory, performing entropy decoding, and performing inverse quantization by using the decoding quantization table; performing inverse discrete cosine transform by using a second integer DCT transform kernel; performing block recombination to obtain a reconstructed reference frame, and taking the reconstructed reference frame as a new reference frame, taking a next frame input as a new current frame, and recycling. The present application can effectively solve the technical problem of distortion accumulation in the multi-round iteration encoding and decoding process based on JPEG.

This invention discloses a method for restoring Cauchy noise images using convolutional sparse coding and low-rank constraints, belonging to the field of digital image processing technology. This invention integrates local convolutional sparse coding and low-rank regularization of structure groups to construct a joint optimization restoration model for images contaminated by Cauchy noise. First, the preprocessed noisy image undergoes convolutional sparse decomposition and dictionary learning to obtain the global sparse features of the image. Simultaneously, the image is divided into blocks, and image blocks similar to reference image blocks are extracted within a search window to construct structure groups. The kernel norm minus the Frobenius norm is used as a regularization term to impose low-rank constraints on the structure groups, suppressing noise while preserving image structural information. This invention employs the alternating direction multiplier method to efficiently solve the joint restoration model, significantly suppressing Cauchy noise, effectively restoring the edge contours and texture details of the image, and improving the visual quality and recognizability of the image. Therefore, it can be used for the restoration of Cauchy noise images.

This invention discloses a crop disease diagnosis method that integrates spatial multi-scale perception with environmental decoupling, belonging to the field of agricultural remote sensing image processing technology. It utilizes large-kernel deep convolution to perform wide-area spatial perception on remote sensing images, generating a spatial weight map encoding global environmental background information. Subsequently, the spatial weight map is reshaped into a dynamic convolution kernel, driving small-kernel group convolution to perform context-adaptive fine-grained disease texture extraction within local neighborhoods, achieving cross-scale feature collaborative modeling. Deep features are decomposed into disease features and environmental features. Adversarial training using gradient inversion layers forces disease features to be free from environmental interference, and mutual information minimization and counterfactual reinforcement learning eliminate environmental spurious correlations, ultimately obtaining robust disease diagnosis results to environmental changes. This invention effectively solves the problem of insufficient generalization performance of existing technologies under varying environmental conditions and has significant application value in multi-temporal and multi-plot agricultural remote sensing monitoring scenarios.

This invention relates to the field of image processing technology and discloses an adaptive smoothing and enhancement method for animation capture trajectories based on temporal feature fusion. The method includes the following steps: S1, acquisition and preprocessing of raw trajectory data; S2, extraction of multi-dimensional temporal features; S3, noise and signal identification based on gated feature fusion; S4, generation of adaptive smoothing kernel and trajectory repair; S5, trajectory enhancement and output: post-processing based on physical constraints is performed on the smoothed trajectory to enhance the dynamic rationality of the action, and a smoothed and feature-enhanced animation capture trajectory is output. This invention intelligently distinguishes between real high-frequency motion and noise by fusing kinematic features, spatial correlation features, temporal semantic features, and attention weights, avoiding the over-smoothing distortion phenomenon of traditional methods. Through a multi-scale feature fusion method optimized by mean drift, key motion regions are automatically identified and assigned higher detail preservation weights, achieving differentiated smoothing enhancement.

The application provides a lung nodule CT image processing method and system, computer equipment and a storage medium, and belongs to the field of image processing. The method comprises the following steps: collecting a lung nodule CT image; segmenting the lung nodule CT image to obtain a nodule segmentation mask; extracting nodule structure semantic features from the nodule segmentation mask; extracting lung nodule texture features from the nodule segmentation mask based on a gray level co-occurrence matrix; extracting lung nodule shape features from the nodule segmentation mask by a Fourier descriptor; splicing the lung nodule texture features and the lung nodule shape features to obtain hand-crafted features; fusing the nodule structure semantic features and the hand-crafted features based on kernel canonical analysis to obtain fused features; and determining the nodule structure complexity corresponding to the lung nodule CT image according to the fused features. The method retains the powerful pattern recognition capability of the nodule structure semantic features, and also integrates the stable discrimination information of the hand-crafted features in a small sample scene, thereby improving the accuracy and stability of classification.

The present application relates to the technical field of image processing, in particular to a fisheye lens image pixel-level edge enhancement and denoising method, comprising: a data acquisition step: acquiring original fisheye image data and preset fisheye lens optical distortion parameters; a weight generation step: determining the distortion stretching rate of a pixel point according to the optical distortion parameters; and generating a spatial density weight map; a threshold determination and denoising step: determining a target denoising threshold according to the spatial density weight map; denoising the original fisheye image data to generate an intermediate denoising image; a convolution kernel generation step: generating an adaptive curved surface convolution kernel according to the optical distortion parameters; an edge enhancement step: using the adaptive curved surface convolution kernel to perform edge enhancement processing on the intermediate denoising image; and generating an enhanced image; the present application effectively avoids the risk of irreversible erasure of key semantic features in the edge field, and ensures high-reliability feature preservation in the extreme field.

The application provides a ship target detection method, and relates to the fields of target detection and remote sensing image processing. The method comprises the following steps: acquiring a remote sensing image containing a ship target, and pre-processing the remote sensing image; inputting the pre-processed remote sensing image into a trained target detection model to obtain the category and position information of the ship target; wherein the target detection model is improved based on a YOLOv5 network and comprises a feature extraction module, a feature fusion module and a feature detection module; in the feature extraction module, a two-dimensional convolution layer of the YOLOv5 network is replaced by a deep separable large kernel convolution; in the feature fusion module, an additional small target extraction layer is added, and a channel attention module is added before each fusion for the fusion of multiple feature maps after up-sampling and down-sampling; and in the feature detection module, an anchor-free target detection head is configured for each input feature map. The application can improve the accuracy and precision of lightweight remote sensing image target detection.

This invention discloses a hyperspectral image classification method, medium, device, and product, relating to the field of image processing. The method includes: performing standardized preprocessing and superpixel segmentation on a hyperspectral image, and constructing a superpixel graph structure; constructing a hyperspectral image classification model including three parallel branches; each branch includes a spectral transform sub-network, a superpixel-level graph sub-network, and a pixel-level convolutional sub-network, wherein the spectral transform sub-network is a convolutional network with different kernel sizes, used to extract features from the hyperspectral image; based on the hyperspectral image features and graph structure, the parallel superpixel-level graph and pixel-level convolutional sub-networks extract superpixel-level and pixel-level features and perform feature concatenation and fusion, and obtain a classification probability distribution based on the fused features; the three parallel branches of the classification model are trained independently, and the trained model is used to classify hyperspectral images; the classification results of the three branches are determined by a majority voting method to determine the final classification result.

This invention provides a deep learning-based target detection method for imbalanced image data, relating to the fields of image processing and target detection technology. First, the dataset is augmented by transferring various meteorological styles to minority class samples, expanding the number of minority class samples in the training set to assist the subsequent detection model and mitigate the adverse effects of class imbalance on target detection. Second, a progressive multi-scale feature enhancement module (CSP-MFEA) is designed, which enhances the feature representation ability of difficult samples through a stepped convolutional kernel and a feature splitting and reuse mechanism. A dynamic Tanh polarization self-attention module (DYT-PSA) is designed, introducing a learnable Tanh activation function to suppress easy samples and enhance difficult samples, optimizing the gradient propagation process. Finally, a YOLO-MFEA-DYT detection model is constructed to improve the accuracy, robustness, and generalization ability of the detection model under imbalanced data distribution.

The present application relates to the field of image processing, in particular to a kind of unmanned aerial vehicle dynamic fuzzy target identification method and system, comprising: obtaining final area and contrast edge in final area;The chain code of different contrast edge in each final area is compared, and the fuzzy performance of each final area is obtained;According to the gradient amplitude of each pixel point on each contrast edge in each final area and the distribution of gradient amplitude, and the fuzzy performance of each final area, the final blurring degree of each final area is obtained;According to the final blurring degree of each final area, the blur kernel of each final area is obtained, and the deblurring processing of image is completed.The present application aims to solve the problem that when the current image is deblurred, the same degree of deblurring intensity is used for different parts of the image, so that the deblurring effect of the image is relatively poor.

This invention relates to the field of image processing technology, specifically to an image inpainting method based on wavelet-enhanced Mamba and multi-domain feature learning. In this invention, Discrete Wavelet Transform (DWT) is used to decompose image features into low-frequency and high-frequency sub-bands, achieving collaborative restoration of structure and texture in different frequency domains. An LFSC-Mamba module is designed for the low-frequency sub-band, combining Mamba's linear sequence modeling capability with a channel-space dual attention mechanism to achieve global structure reconstruction while maintaining linear computational complexity. An HFDP module is designed for the high-frequency sub-band, employing a dynamic filtering kernel generation strategy in the Fourier domain and combining it with a multi-head attention mechanism to achieve fine-grained synthesis of directional and pattern textures, significantly improving the realism and diversity of texture generation.

The application provides a visible light and infrared target detection method and system based on progressive prior guided fusion, and belongs to the technical field of computer vision and image processing, and comprises the following steps: inputting a visible light image and an infrared image into a double-flow backbone network to extract single-mode visible light features and infrared features in multiple stages; inputting single-mode features extracted by each key stage of the backbone network into a progressive prior guided fusion module PGPF; the PGPF comprises a mode prior generation module MPG and a prior guided dynamic convolution module PGDC, the MPG performs initial fusion on the single-mode features to generate mode quality priors comprising mode quality information, the PGDC enhances the mode quality priors and generates dynamic convolution kernels specific to the modes, and the single-mode features and the enhanced mode quality priors are corrected and fused to obtain final fusion features; and the final fusion features of each key stage are input into an encoder, a decoder and a detection head to output a detection result. The application can effectively perform target detection.

This invention discloses a curvature-guided distillation method and system for synthetic image datasets, belonging to the field of image processing technology. To address the technical problems of existing technologies, such as large variance in higher-order statistical estimation under subsampling conditions, uncontrollable computational overhead, and insufficient robustness of synthetic samples, this invention preprocesses real samples and extracts their features and kernel functions along with those of the synthetic samples to be optimized; calculates the sampling weights of the real samples using a sampler; calculates the first-order loss based on the weights and features, and extracts Hessian second-order information to obtain the curvature loss term; synchronously updates the parameters of the synthetic samples and sampler using the first-order loss and the curvature loss term to generate a parent set; finally, the parent set is cropped and corrected. This invention can stabilize higher-order statistical estimation and reduce gradient variance, improve the structural fidelity and generalization ability of synthetic samples, and achieve single distillation adaptability to high-quality datasets of various sizes.

This invention provides a heterogeneous computing real-time video processing method and system based on dmabuf, belonging to the field of real-time video processing and display technology. The method includes: storing the original image in a target dmabuf of the video memory; when the platform has OpenGL hardware acceleration, creating a first OpenGL texture from the target dmabuf, performing image processing on the first OpenGL texture using the image processing functions of the hardware acceleration part, and creating a first OpenCL cache from the first OpenGL texture; creating a blank second OpenGL texture, and creating a second OpenCL cache from the second OpenGL texture; using OpenCL kernel functions to perform calculations on the data in the first OpenCL cache to obtain the finished image, and saving the finished image to the second OpenCL cache. This invention can achieve load balancing, solve the problem of excessive differences in CPU and GPU usage, avoid data transmission from GPU to CPU and then back to GPU, reduce the latency from image fetching to display, and simultaneously reduce overall power consumption.