Representation, Analysis of Bayesian Refinement Approximation Network: A Survey

03/27/2021
by   Ningbo Zhu, et al.
0

After an artificial model background subtraction, the pixels have been labelled as foreground and background. Previous approaches to secondary processing the output for denoising usually use traditional methods such as the Bayesian refinement method. In this paper, we focus on using a modified U-Net model to approximate the result of the Bayesian refinement method and improve the result. In our modified U-Net model, the result of background subtraction from other models will be combined with the source image as input for learning the statistical distribution. Thus, the losing information caused by the background subtraction model can be restored from the source image. Moreover, since the part of the input image is already the output of the other background subtraction model, the feature extraction should be convenient, it only needs to change the labels of the noise pixels. Compare with traditional methods, using deep learning methods superiority in keeping details.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
11/04/2019

Superpixel-Based Background Recovery from Multiple Images

In this paper, we propose an intuitive method to recover background from...
research
04/16/2021

Arithmetic Distribution Neural Network for Background Subtraction

We propose a new Arithmetic Distribution Neural Network (ADNN) for learn...
research
07/02/2019

Seismic data denoising and deblending using deep learning

An important step of seismic data processing is removing noise, includin...
research
02/11/2020

Background Matting

The current state of the art alpha matting methods mainly rely on the tr...
research
03/20/2022

NeuralReshaper: Single-image Human-body Retouching with Deep Neural Networks

In this paper, we present NeuralReshaper, a novel method for semantic re...
research
01/18/2019

DeepOtsu: Document Enhancement and Binarization using Iterative Deep Learning

This paper presents a novel iterative deep learning framework and apply ...
research
09/13/2016

Image Decomposition Using a Robust Regression Approach

This paper considers how to separate text and/or graphics from smooth ba...

Please sign up or login with your details

Forgot password? Click here to reset