Large convolution kernel real-time approximate fitting method based on bilinear filtering image hierarchy

Abstract

The invention discloses a large convolution verification real-time approximate fitting method based on a bilinear filtering image hierarchy. The method comprises the steps: 1) carrying out the downward sampling of an input image through bilinear filtering, and obtaining an image pyramid MIP; gradually sampling the MIP upwards from the highest level to obtain a convolution calculation approximate fitting image; wherein in the up-sampling stage process, an image, subjected to convolution approximate filtering, of the Lth layer in the MIP is generated through linear interpolation between a pixelsample p (L + 1) and a pixel sample pdown (L), and an interpolation mixed parameter depends on a target convolution kernel function: p (L) = (1-alpha (L)) p (L + 1) + alpha (L) pdown (L); wherein alpha (L) represents the interpolation mixing parameter of the Lth layer, p (L) represents the pixel in the output image of the Lth layer in the up-sampling stage, p (L + 1) represents the pixel sample output by up-sampling of the (L + 1) th layer, and pdown (L) represents the pixel sample output by down-sampling of the Lth layer.

Description

technical field [0001] The invention belongs to the technical field of computer graphics, and relates to a real-time approximate fitting method for a large convolution kernel used for illumination sampling, in particular to a real-time approximate fitting method for a large convolution kernel based on a bilinear filter image hierarchy. Background technique [0002] In real-time rendering, convolution filtering is often used, and the most common effects include image bloom (bloom), depth of field (depth of field, DOF) and other post-processing special effects. In addition, when using some popular real-time physics-based rendering (PBR) and image-based lighting ( When using image-based lighting (IBL) technology, Gaussian blur is also commonly used to weaken high-frequency noise and improve the stability of time domain and space domain at one time. However, the computational complexity of traditional convolution is usually related to the size of the convolution kernel, resulti...

AI Technical Summary

Problems solved by technology

[0009] In order to realize the approximation of high-quality large-kernel convolution calculation at a higher speed, especially for the current large amount of calculation in the required application, the present invention combines the MIP map of the bilinear interpolation box filter to propose a new approximate calculation convolution method

Images