16 results about "Color image processing" patented technology

The embodiment of the invention provides a pseudo-color image processing method, system and device and a storage medium, and relates to the technical field of image processing, and the method comprises the steps: receiving a pseudo-color image and corresponding annotation information, and determining an execution enhancement processing module based on the annotation information; calling an enhancement processing module to extract a local image patch from the pseudo-color image, and performing enhancement processing on the local image patch to obtain an enhanced patch; and fusing the enhanced patch with a corresponding area in the pseudo-color image according to the fusion coefficient, and outputting a processed image. By receiving the marked pseudo-color image and selecting different enhancement processing modules based on the marked information, differential enhancement processing of a foreground target key area and a background environment in the image is realized. Through the operations of extraction, enhancement and re-fusion, on the premise of keeping the authenticity of the overall structure of the image, the characterization diversity of the key area is pertinently improved.

This invention discloses an AI-based automatic analysis color diagnosis system for fundus lesions, relating to the field of ophthalmic disease diagnosis technology. It includes a fundus image acquisition module and an AI color analysis module. This AI-based automatic analysis color diagnosis system for fundus lesions, through its fundus image acquisition module, AI color analysis module, and analysis model construction module, inputs acquired fundus color images into a fundus color image processing and analysis model to obtain fundus color feature data. It then identifies the patient's lesions, diagnoses the patient's ophthalmic disease based on the identified lesions, and determines the location of the lesions in the fundus images using the fundus color feature data. This constructs a fundus image color evaluation system, improving the accuracy and reliability of ophthalmic diagnosis and assisting in the diagnosis and treatment of fundus diseases.

The application discloses a kind of structural color image processing methods with dynamic tool movement trajectory, comprising the following steps: original image is divided into multiple pixel units;According to the hue and brightness of each pixel unit, obtain the tool movement trajectory under all pixel units corresponding;Make tool according to the tool movement trajectory under current pixel unit corresponding to carry out cutting processing, after processing current pixel unit, the movement trajectory of tool is changed from the tool movement trajectory under current pixel unit corresponding to the tool movement trajectory under next pixel unit corresponding;This method can dynamically adjust the movement trajectory of tool in the process of cutting processing, flexibly adjust the frequency and shape of tool elliptical vibration, and then control the hue and brightness of structural color, so as to enrich the color space in the field of structural color processing.In the process of cutting processing, the nominal cutting speed of tool can be kept unchanged, so as to avoid the adverse vibration caused by frequent sudden movement, improve the machining precision.

The invention discloses a wave interference removing method and system based on a pseudo-color mapping improved generative adversarial network, and relates to the field of navigation and positioning. The method solves the problems of how to effectively remove noise and recover damaged polarization information caused by underwater condition interference for underwater polarization images collected in a complex underwater scene in the prior art, and the like, and comprises the following steps: constructing a mathematical model of polarized light transmission in a fluctuating water surface environment, and constructing a corresponding data set; performing pseudo-color image processing on the polarization image in the complex underwater scene by adopting an improved generative adversarial network to obtain a simulation data set of the wave interference scene subjected to pseudo-color processing; training a deep neural network, and outputting a non-interference wave polarization image; cooperatively training and improving the generated adversarial network by adopting multiple loss functions; wave polarization image recovery is realized through collaborative optimization of the generator and the discriminator, detail features of the wave polarization image are reserved, and wave interference removal of the pseudo-color mapping improved generative adversarial network is completed.

The application belongs to the technical field of image processing, and provides a mixed color correction model construction method, a color deviation prevention image processing method and device. The mixed color correction model construction method comprises the following steps: obtaining a first spectral reflectance characteristic curve according to the color information and corresponding reflection spectrum data of a printing material; obtaining a second spectral reflectance characteristic curve according to the color gamut range and corresponding reflection spectrum data of a target ink on an unengraved printing material; obtaining a third spectral reflectance characteristic curve through the color value and corresponding reflection spectrum data of a multi-level gray background produced by a laser engraving machine; superimposing the target color ink and the multi-level gray background to obtain mixed color values and mixed spectrum data, and then constructing a mixed color correction model. The method and device realize accurate reverse mapping from the target color to the laser parameters and ink parameters, and avoid the occurrence of significant and nonlinear color deviation of the superimposed color ink.

Embodiments of the present application disclose a full-color image processing method and apparatus based on low-light night vision, an electronic device, and a readable medium. A specific implementation of the method comprises: collecting a target image information group; on the basis of target color image information, generating visual color image information; performing feature extraction on infrared image information to obtain infrared image feature information; performing brightness channel extraction on the visual color image information to obtain a brightness channel; performing feature fusion on the infrared image feature information and the brightness channel to obtain fused feature information; inputting the fused feature information into a pre-trained brightness enhancement processing model to obtain target brightness information; changing the visual color image information; and converting the changed visual color image information. The implementation can generate color image information and obtain more color details, thereby improving the quality of a generated color image, and further enhancing the effect of image processing.

The present application belongs to the technical field of three-dimensional human posture estimation, and in particular to a three-dimensional human posture estimation method fusing color images and depth images. The method comprises the following steps: step one, splitting an RGB-D dataset into a color image dataset and a depth image dataset, which are respectively taken as two inputs of a depth image processing module and a color image processing module; step two, performing image preprocessing and depth image voxelization processing on the input depth image; step three, sending the input color image into a two-dimensional human key point detection module, and extracting low-level features from the color image through a convolutional neural network. The present method detects human key points in the color image based on a Transformer network and outputs a key point heat map, and the network is improved by introducing relative position coding, thereby improving the performance of the posture estimation network and providing accurate data input for three-dimensional human posture estimation.

The invention relates to a method for imaging a living biological tissue, comprising: obtaining (S1) a plurality of successive colour images of said biological tissue, the processing method comprising a calibration phase (S2) in which at least one zonal time-dependent vector representative of the time evolution of spatial averages of intensity values of each of the colours in an area of the images is determined (S21), a bandpass filter centred on a heartbeat frequency is applied (S24), then a projection basis is determined (S26) by principal component analysis of the filtered zonal time-dependent vector, the method comprising deriving (S33) colour data representative of the time evolution of luminous intensities of the living biological tissue in the tissue images, and projecting (S35) colour data onto the projection basis.

Apparatuses, systems, and techniques to process image data. In at least one embodiment, a neural network is trained to perform demosacing of two-dimensional image data obtained from an image sensor.

The invention provides a cigarette ash packing quality detection method and equipment based on machine vision, and the method comprises the steps: obtaining a color image of a target region through machine vision, processing the color image into a gray image, carrying out the data analysis of the gray image when the gray image meets a set condition, obtaining a plurality of ash packing performance indexes, and carrying out the detection of the ash packing quality. Outputting ash coating performance indexes after the detection is finished; the ash coating performance indexes comprise at least one of the combustion rate, the carbon line, the inclination angle, the flash, the crack and the ash holding time. According to the invention, through a set program, the cigarette is ignited to start real-time dynamic detection of the ash pack, the equipment and the program are combined to realize detection of performance indexes such as ash pack combustion time, gray scale, cracks, flash, ash holding time and carbon lines and output the performance indexes to a user, and meanwhile, the quality of the cigarette ash pack can be judged according to the multiple indexes, so that the detection efficiency is improved. And multiple indexes in the industry are integrated, so that the performance of the ash coating can be comprehensively evaluated.

The invention belongs to the technical field of image processing, and provides a mixed color correction model construction method and a color cast prevention image processing method and device.The mixed color correction model construction method comprises the steps that a first spectrum reflection characteristic curve is obtained according to chromaticity information of a printing material and corresponding reflection spectrum data; obtaining a second spectral reflection characteristic curve according to the color gamut range of the target ink on the non-carved printing material and the corresponding reflection spectrum data; obtaining a third spectral reflection characteristic curve according to the chromatic value of the multi-level gray ground color generated by the laser engraving machine and the corresponding reflection spectrum data; the method comprises the following steps: overprinting ink of a target color and a multi-level gray background color to obtain a mixed chromatic value and mixed spectrum data, and further constructing a mixed color correction model; according to the method and the device, accurate reverse mapping from the target color to the laser parameters and the printing ink parameters is realized, and remarkable and nonlinear color cast of overprinted color printing ink is avoided.

The application discloses a pose estimation method and a related model training method, device, equipment and medium. The pose estimation model training method comprises the following steps: obtaining a sample image containing an object to be positioned, wherein the sample image comprises a sample color image and a sample depth image corresponding to the sample color image; processing the sample color image by using a pose estimation model to obtain a sample initial pose of the object to be positioned; optimizing the sample initial pose based on depth information of the object to be positioned in the sample depth image to obtain an optimized pose of the object to be positioned; and adjusting network parameters in the pose estimation model based on a difference between the optimized pose and the sample initial pose. The above scheme can improve the efficiency of model training.

This disclosure relates to a head-mounted device, image processing method, apparatus, storage medium, and program product, which can enhance the display format of the head-mounted device's perspective function and improve the user experience. The head-mounted device includes: two single-channel cameras for capturing two single-channel images; a multi-channel camera for capturing color images; a processor for acquiring depth information of the external scene based on the two single-channel images and fusing the depth information with the color images to obtain a color image with depth information; and a display screen for displaying the color image with depth information.

This application relates to the field of optical device technology, providing a metasurface structure, optical processing system, and method for dual-layer optical convolution. The metasurface structure includes: a polarization-multiplexing metasurface unit performing dual-layer convolution operations on incident light carrying color image information; the first layer of convolution utilizes the inherent properties of optical vortices to achieve edge enhancement and feature extraction of the image; the second layer of convolution achieves weighted combination of feature images through the spatial superposition of multi-channel spot matrices on the detector plane; the incident light includes four different polarization states, used to generate different optical vortex arrays and zero-order spot arrays constituting residual channels corresponding to the modulation of different colors of light; a dynamic amplitude modulation unit and a control unit cooperate to provide brightness weights for the second-layer convolution operation. This application enables dynamic dual-layer convolution calculation on a single optical device, supports color image processing, and possesses a hardware-level residual connection mechanism to achieve visual optimization.

The invention discloses a high-quality true color image processing method and system and a medium, and the method comprises the steps: moving an area array sensor, and collecting the image data after each movement, so as to obtain the actual measurement data of all pixel positions under different channels, the distance between the central points of any two adjacent pixel positions in the transverse direction and the longitudinal direction is 1 / N pixel; sub-pixels are divided, so that a sub-pixel region is 1 / N2 of an original pixel region of the area array sensor; constructing convolution models under different channels, so that a square matrix formed by any N2 sub-pixels is subjected to convolution to form actual measurement data of corresponding pixel positions; and calculating the pixel value of each sub-pixel under different channels through deconvolution so as to obtain a super-resolution true color image. According to the invention, the pixel displacement technology and the deconvolution algorithm are combined, the limitation of the physical pixel of the sensor on the resolution is broken, the super-resolution true color effect is realized, and the image quality is obviously improved.