Fast artistic style study method supporting diverse images

Abstract

The invention provides an artistic style study method supporting diverse images. The input is one group or one artistic master image, a control image and a target image B, the output is an artistic master image B' with the similar style to the sample image, and the method comprises the steps that a user selects a sample pattern style block Ti (i=1, 2, to nexem) capable of best representing the style features from the artistic master image, and the nexem is the number of the sample image style blocks. The Ti is converted into a YIQ space, and a Y passage value is selected to be used as the feature; the edge tangential flow field of the image B (direction B) is calculated; the size of the image B (size B) is used for generating the Ti Gaussian pyramid T=(T0, to Tx, to T<L>) (wherein x is in a range of [0, L]) and generating the coordinate pyramid S=(S0, to Sx, to S<L>) (wherein x is in a range of [0, L], and L=log<2>(sizeB)) of the B', the Tx represent s the x-th layer image of the T Gaussian pyramid of the sample image style block, and the Sx represents the x-th layer image of the coordinate pyramid; the control image is read in ,and the initial layer sinitl is set; the coordinate pyramid of the B' is cyclically subjected to rotating matrix calculation, super sampling and correction steps from the rough layer Sinitl to each layer S1 in the finest layer, and the steps are stopped until the finest layer is reached; the brightness remapping is carried out, and the B' is obtained through calculation.

Description

technical field [0001] The invention relates to computer image processing technology. Background technique [0002] The learning-based artistic style drawing technology can process the target image through the learning algorithm according to the artistic style reference image and a target image provided by the user, and generate a result image with a style similar to the reference image. This type of technology is relatively simple for users. Users do not need to have an in-depth understanding of painting styles. They only need to provide art reference pictures to get desired works with a certain artistic style. This type of technology was first proposed in the image analogy framework (Hertzmann, A., Jacobs, C.E., Oliver, N., Curless, B., Salesin, D.H.: Image analogies. In: Proceedings of the 28th Annual Conference on Computer Graphics and Interactive Techniques , 2001, pp.327–340), the user gives a pair of relationship samples A and A', there is a certain style conversion ...

AI Technical Summary

Problems solved by technology

Most of the above methods learn the style of the entire sample image, and cannot draw different brushes according to the characteristics of the object area in the target image, nor can they transfer the artistic style of multiple sample images to different areas

Images