32results about How to "Good image detail" patented technology

The invention provides a target image generation method, a human body posture transformation method and device, a server and a storage medium, and belongs to the technical field of machine vision. According to the scheme provided by the embodiment of the invention, the method comprises: predicting the first transformation parameter through the first image generation model, and adjusting the firstanalytic image based on the first transformation parameter, so that the attitude of the first analytic image is aligned with that of the generated target analytic image; predicting the second transformation parameter through the second image generation model according to the target analysis image obtained by the first image generation model, and adjusting the combined image to be processed and thefirst analysis image based on the second transformation parameter, so that pixels between the image to be processed and the generated target image are aligned, and better image details are achieved.The parsed images before and after transformation are kept in posture alignment, so that the images before and after transformation can be kept in space alignment, and the generated target image not only guarantees space alignment, but also guarantees clear details.

The invention discloses a color digital image enhancing and denoising method under random illumination. The method is characterized by comprising the following steps of: 1, image denoising processing, namely converting an original image from an RGB space to a YCbCr space, removing a Gaussian noise by using a Gaussian filter and removing a salt and pepper noise by performing median filter; 2, image brightness / contract stretch processing, converting the image from the RGB space to an HSI space, decomposing the image by using a two-sided filter, processing by an improved Retinex model algorithm,and obtaining a new image saturation by performing saturation compensation; and 3, performing fusion display on the image acquired after denoising the YCbCr space and the image acquired by HSI space processing. By the method, the color constancy of the image can be kept, the dynamic range of the image can be improved well, and simultaneously noises of the image can be inhibited and removed with less texture and detail information loss.

The invention discloses a residual instance regression super-resolution reconstruction method based on multi-level dictionary learning. For the feature vector of the image block, use K-SVD to learn a dictionary with strong representation ability as the anchor point; use the learned dictionary to perform least squares regression on the low-resolution and high-resolution blocks in the block to obtain a linear mapping relationship; Estimate the high-resolution features, calculate the reconstruction error, and map the estimated high-resolution features with the reconstruction error while doing further dictionary learning; after the L layer, a set of residual regressions is obtained; use the input image and the obtained The residual regressor performs reconstruction, and the obtained high-resolution features are used for the reconstruction of the next layer; all estimated high-resolution image blocks are summed and calculated to synthesize a high-resolution image. The invention has stronger super-resolution capability and can be used for enlarging low-resolution natural images.

The invention provides an ultra-high-definition video processing method and system. The ultra-high-definition video processing method comprises the following steps: receiving each standard-definitionvideo frame, and obtaining a first enhanced image ISGB after enhancement processing; obtaining a relative brightness value of each frame of image after enhancement processing; obtaining an enhanced adjustment coefficient according to the relative brightness value; obtaining a second enhanced image IFINAL of each pixel point according to the first enhanced image ISGB and the enhancement adjustmentcoefficient; and performing two-stage or multi-stage conversion processing on the second enhanced image IFINAL to obtain an amplified ultra-high-definition video signal. According to the finally obtained ultra-high-definition video image, the video definition is far higher than the initial video definition, and the color is brighter, and the image details are more obvious, and extremely high visual experience is obtained.

The invention discloses an underwater image enhancement method based on a foreground model. The method comprises the following steps that: improving a background light estimation method so as to effectively avoid the influences of underwater image overexposure, artificial light sources and the like; combining with the cognition of people for the underwater image, and utilizing a dark channel priori algorithm to remove background scattering and extract the foreground model; and combining with a white balance algorithm to put forward a color correction method suitable for the underwater image, and utilizing the attenuation characteristic of light in water to correct channel gains according to a relationship between channel attenuation coefficients so as to compensate color distortion caused by attenuation; and utilizing the channel gains to regulate a fogless image, and finally, obtaining the enhanced underwater image. By use of the method, the enhancement effect of an object part is clear and distinct, and a visual effect is better; image blurring is effectively removed, so that the definition of the enhanced underwater image is greatly improved, image details are better, the image enhancement of the background part is not affected on the basis of the color correction of the foreground model, the enhanced underwater image has more natural integral colors, and image brightness is within an acceptable range.

The invention discloses an ancient font classification method based on a convolutional neural network. According to the method, firstly, an ancient font category image data set is crawled by using a crawler technology; through data expansion, training set samples tend to be balanced; graying processing is carried out on the balanced training set sample and setting an image size to a target image size; histogram equalization processing is performed on the sample set, isolated noise points are removed in the image through an N8 connected noise reduction algorithm, and finally binarization processing is performed on the image based on a fuzzy set theory and by using a Shannon entropy function, so that detail features of the image are well reserved; based on the objective function of the classification task. The center loss function and the traditional cross entropy loss function are matched for use. The inter-class distance is increased. The intra-class distance is reduced. The distinguishing capability of features is improved to a certain extent, preprocessed images are trained through a pre-defined network model, and the accuracy of a classification result is evaluated through a confusion matrix. According to the method. The preprocessing effect on the degraded ancient font image is remarkable, and a more accurate ancient font classification effect is achieved by optimizing parameter setting and utilizing appropriate training skills to train the convolutional neural network model.

The invention discloses an X-ray image multi-scale detail enhancement method in integrated circuit packaging, which comprises the steps of 1) acquiring X-ray images oriented to an integrated circuit packaging process; 2) performing Laplacian pyramid decomposition to the X-ray images to obtain X-ray pyramid detail images at all scales; 3) adjusting the brightness and the contrast of the images at the bottom layer of a pyramid by adopting a logarithm enhancement method; 4) adjusting the brightness and the contrast of the detail images at the top layer of the pyramid by adopting a histogram equalization enhancement method; 5) reconstructing the detail images at all scales to obtain detail enhanced images. The method disclosed by the invention has the advantages that the method is simple and fast, the dynamic range compression of images can be realized aiming at the X-ray images with low signal-to-noise ratio and poor defect contrast in the integrated circuit packaging, the useful information of the images is expanded, the image details are enabled to be more outstanding, the contrast of the images is improved and the accuracy is higher.