Mobile phone shot image de-noising method based on dual convolutional networks

A technology for shooting images and convolution, which is applied in the field of image denoising, which can solve the problems of inability to effectively estimate the noise level of images taken by mobile phones, low peak signal-to-noise ratio of denoised images, etc., to solve the problem of greatly reducing the denoising effect and reducing the number of channels , good structural effect

CN112634175AInactive Publication Date: 2021-04-09SOUTHWEST PETROLEUM UNIV
6 Cites 3 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Publication Date
2021-04-09
Estimated Expiration
Not applicable · inactive patent

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

The invention discloses a mobile phone shot image de-noising method based on dual convolutional networks, which improves the defects of FFDNet, adds a noise estimation sub-network, constructs a redundant link block structure, better saves the structure of an original image, and solves the problem that the de-noising effect is greatly reduced due to the fact that the subjective estimation noise level of FFDNet is wrong.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] The invention relates to the field of image denoising, in particular to images captured by mobile phones, and in particular to a method for denoising images captured by mobile phones based on double convolution networks. Background technique

[0002] Today, smartphones are recognized as the most important technology product that affects the daily life of ordinary people. Among the various functions of smartphones, taking pictures is a very important part. Although people are becoming more and more accustomed to using light smartphones to take pictures instead of bulky digital SLR cameras, due to the limitation of the internal space of mobile phones, there is still a gap between the sensor size, aperture size and other hardware configurations and digital SLR cameras. The advantage of smartphones over digital SLRs is that they have powerful chips and neural network processors. Smartphones can use the powerful computing capabilities of their chips to solv...

Examples

Embodiment Construction

[0036] The main realization principles, specific implementation methods, etc. of the technical solution of the present invention will be described in detail below.

[0037] 1. Construct a double convolution denoising network model;

[0038] 1) Construct the noise estimation subnetwork:

[0039] Construct three network layers, input layer, hidden layer, and output layer; the input layer is the input of the image taken by the mobile phone; the output layer is the output of the estimated noise; the hidden layer is divided into 5 parts: the first part of the hidden layer is the convolutional layer , the convolution layer contains 64 convolution kernels, the size is 3x3, the step size is 1, the padding is 1, and the activation function is a linear rectification function; the second part of the hidden layer is 3 redundant link blocks, each redundant The co-link block contains two batch normalizations and two convolutional layers. The convolutional kernel size of the first convoluti...