32results about How to "Avoid estimated impact" patented technology

The invention provides electronic image stabilization method and device of the vehicle-mounted camera and readable storage medium. The method comprises determining a moving object recognition area ofa first reference image collected at the last moment and a plurality of feature points in an area outside the moving object recognition area of the first reference image; obtaining reference points corresponding to the feature points from a second reference image collected at the current moment; determining a first affine transformation matrix according to each feature point and each correspondingreference point; determining a compensation parameter according to a difference value between an observed quantity and a state quantity in the first affine transformation matrix; determining a secondaffine transformation matrix according to the compensation parameter, transforming the second reference image through the second affine transformation matrix to obtain the stable image at the current moment, and therefore the feature points in the moving object recognition area where the moving object in the image is located are removed, the influence of the moving object on moving parameter estimation is avoided, and the image stabilization accuracy is improved.

The present invention provides a magnetic resonance diffusion tensor denoising method and system. The method is based on the magnetic resonance diffusion weighted imaging model and the Gaussian distribution property of the sampling noise, utilizes the sparseness of the relative anisotropy of the diffusion, and adopts the maximum a posteriori probability estimation The method of K-space directly obtains the denoised diffusion tensor matrix corresponding to each spatial position. The invention can avoid the influence of the image denoising error on the estimation of the diffusion tensor, suppress the noise in the diffusion tensor more effectively, and improve the estimation precision of the diffusion tensor.

The invention discloses a rigid target three-dimensional motion estimation method in a multi-view environment, and is specifically implemented according to the steps of: 1. building a multi-view target image acquisition environment, and obtaining a plurality of target images; 2. rebuilding a target three-dimensional shape S(t) according to the plurality of target images that are obtained in the Step 1, and initializing motion parameters of the target three-dimensional shape S(t); 3. building an energy function E which reflects distance between a multi-view two-dimensional projection of the target three-dimensional shape and the target images, judging whether the energy function E converges, if not, turning to Step 4, and if yes, outputting the motion parameters of the target three-dimensional shape S(t), i.e., the motion parameters of a rigid target three-dimensional shape S(t); and 4. updating the motion parameters of the target three-dimensional shape S(t), and turning to the Step 3. Estimation of target three-dimensional motion by the method is not influenced by illumination variation and chromatic aberration of a camera.