A method of image deconstruction based on machine learning

Abstract

The invention relates to an image deconstruction method based on machine learning. The image deconstruction method comprises the following steps that 1) four layers of convolutional neural networks are established to act as relative reflectivity classifiers; 2) multiple sets of pixel pair information of a first image is acquired from a data set; 3) the pixel pair information and a second image obtained after size adjustment of the first image act as the input of the four layers of convolutional neural networks so that the classification result is acquired; 4) a hinge loss optimization problem is generated according to the classification result; and 5) CFR solving is performed on the hinge loss optimization problem so that the reconstruction result of the first image is acquired. Compared with the methods in the prior art, the image deconstruction method based on machine learning has the advantages of being more delicate and smoother in the image reconstruction result.

Description

technical field [0001] The invention relates to an image deconstruction method, in particular to an image deconstruction method based on machine learning. Background technique [0002] Image analysis is a very popular field in computer graphics in recent years. After 2010, many papers have been researched in this direction. Many neural networks such as HSC and CNN have been used to realize this function. The most influential and earliest one should be bell The team's analysis based on the IIW dataset. [0003] Most of the original image deconstruction methods are based on comparing images, and the results obtained by this method are rough and not fine enough. For example: S.bell et al. proposed in the paper "Intrinsic images in the wild" (picture essence based on IIW) published in "ACM Transactions on Graphics" to extract the information of point pairs in the picture from the data of IIW, and solve the CRF (Conditional Random Fields, conditional random field) method to pro...

AI Technical Summary

Problems solved by technology

For example, J.T.Barron et al. proposed the same convolution in the paper "Intrinsic scene properties from a single RGB-Dimage" published in "IEEE Conference on Computer Vision and Pattern Recognition, 2013" (2013 Computer Graphics and Pattern Recognition Conference Collection) Neural network, and use it to improve the accuracy of comparison. These methods can get relatively finer deconstruction maps, but the problem is that they require RGB and depth inputs, which greatly reduces the scope of application.

[0005] Finally, when CNN and other neural networks are also used for local-to-global estimation, the weight calculation methods used by most methods are too arbitrary, resulting in unrefined images, and the final processing of the matrix often requires a large amount of calculation. Therefore, there is a need for a smoother estimation method and an estimation method that can get good enough results and greatly reduce the amount of calculation

Images