A fringe image repair method based on convolutional neural network denoising and regularization

Abstract

The invention discloses a fringe image repair method based on denoising and regularization of convolutional neural network, and relates to the technical field of image processing. bright and short exposure fringe pattern I dark ; S2, for short exposure fringe pattern I dark The modulation chart is binarized to determine the normal exposure fringe pattern I bright The highlight area in ; S3, fringe pattern I according to normal exposure bright and short exposure fringe pattern I dark , the method of grayscale adjustment and image area replacement is used to obtain the initial value I of the iterative repair algorithm of CNN denoising and regularization fused ; S4, according to the normal exposure fringe pattern I bright and the initial value I of the iterative repair algorithm fused , and use the iterative repair algorithm of CNN denoising and regularization to calculate the final fringe image and complete the repair of the fringe image. The invention has the advantages of simple operation, good repair result, short processing time, no need to introduce additional hardware facilities, and low operation cost.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular, to a fringe image restoration method based on denoising and regularization of a convolutional neural network. Background technique [0002] Fringe projection profilometry (FPP) is a widely used 3D reconstruction method. When there are objects with high dynamic range reflectivity on the surface measured by fringe projection profilometry, the intensity-saturated areas in the acquired fringe images will cause errors or lack of phase calculation in the corresponding areas, and ultimately affect the recovery of 3D topography. [0003] At present, there are three kinds of fringe image repair methods, namely, the multi-exposure fusion repair method, the method of directly repairing the missing area of ​​the fringe, and the method of adding a polarizer by using the properties of light. The defects of these three methods are as follows: [0004] 1) Using the method of mult...

AI Technical Summary

Problems solved by technology

[0004] 1) Using the method of multi-exposure fusion, it is necessary to perform multiple exposures on the measured object, generally speaking dozens of times, and the operation is very troublesome;

[0005] 2) The method of directly repairing the missing stripe area is not ideal for a large range of highlighted areas, and the repaired result does not introduce the true height distribution of the object;

[0006] 3) The method of adding a polarizer using the properties of light requires additional hardware, which increases the cost of the measurement system

Images