Unlock instant, AI-driven research and patent intelligence for your innovation.

A Relighting Method Based on Image and Ensemble Learning

An integrated learning and re-illumination technology, applied in neural learning methods, 3D image processing, image data processing and other directions, can solve problems such as poor universality and large collection volume

Active Publication Date: 2022-04-12
HOHAI UNIV
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Generally speaking, the more base images are collected, the more accurate the reconstructed scene lighting information is, but dense sampling is often required, the amount of collection is very large, and the universality is not strong

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A Relighting Method Based on Image and Ensemble Learning
  • A Relighting Method Based on Image and Ensemble Learning
  • A Relighting Method Based on Image and Ensemble Learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0036] Below in conjunction with accompanying drawing, technical scheme of the present invention is described in further detail:

[0037] The present invention is a re-illumination method based on image and integrated learning, such as figure 1 shown, including the following steps:

[0038] (1) Collect 3D scene data, including the obtained image set ImageSet and the corresponding light source position, one light source position corresponds to one sampling image. For simplicity, a 2D plane light source is used, that is, the available coordinates of the light source position [L x , Ly ] means that the RGB values ​​of the pixels in all image sets are normalized to the [0, 1] range;

[0039] Table 1 scene data

[0040] Scenes image size Light field size type Mitsuba 800×600 21×21 virtual dragon 800×600 31×31 virtual Waldorf 696×464 32×32 reality bull 696×464 32×32 reality

[0041] (2) An image consists of H×W pixels, and ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a re-illumination method based on image and integrated learning. The method divides pixels into two categories of "easy training" and "difficult training" in the image space, that is, Pixel Easy and Pixel Hard , and then divide and conquer. For Pixel Hard , using the AdaBoost method of integrated learning and neural network for learning, for the pixel Pixel Easy , which is further divided into two categories, Pixel Easy I and Pixel Easy II. to Pixel Easy Class I pixels, using the Bagging method of integrated learning and neural network for learning; for Pixel Easy II, use the characteristics of the human eye to assign values. The image-based re-illumination of the method refers to recovering or reconstructing the illumination information of the scene by using sampled images under new illumination conditions, and the method of collecting base images and illumination reconstruction is the key of the present invention. Based on the above method, four virtual and real scenes are tested. Under the control of the same relative reconstruction error accuracy, the number of image samples required for re-lighting is less and the effect is better.

Description

technical field [0001] The invention belongs to the field of computer graphics, in particular to a re-illumination method based on image and integrated learning. Background technique [0002] The image-based relighting (Image-based Relighting, IBR) technology regards the continuous light transmission space as a discrete light transmission space, approximates the light transmission matrix through the captured image, and then uses the obtained light transmission matrix to reconstruct it in a new Scene under lighting conditions. Its biggest advantage is that it does not require the geometric information of the scene model, is not affected by the complexity of the model, and can reconstruct special lighting effects such as highlights, scattering, caustics, and hard shadows. Therefore, since IBR was put forward, it has been a research hotspot in the field of computer graphics. [0003] In IBR technology, the acquisition of the base image and the method of illumination reconstru...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06T15/20G06N3/04G06N3/08
CPCG06T15/205G06N3/045
Inventor 韦伟刘惠义钱苏斌陈霜霜
Owner HOHAI UNIV