Research On Stylized Drawing Based On Preserving Colors

Image stylized drawing technology is the use of computer technology to give the painting art style for ordinary natural images,it does not pursue the scenes and details of the image,it is more inclined to make the generated images have artistic sense and visual beauty.The stylized drawing represented by non-photorealistic drawing technology began research and exploration of computer art drawing more than a decade ago.In recent years,the use of convolutional neural networks to extract art styles automatically,and thus generate a stylized transitional rendering method with artistic style images has provided a new direction for the development of stylized drawing.At present,neural stylized migration rendering techniques have many problems such as unstable iterative results,poor semantic perception of network models,lack of visual rationality,and loss of visual aesthetics.Due to the exploration of the stylized rendering of the retained color image,as well as the solution to the problems such as the color confusion of the generated image and the loss of visual aesthetics in the existing methods,this paper proposes a stylized rendering of the retained image color based on a convolutional neural network.Methods include the following two types:(1)Based on the color migration method,a color migration operation is performed on the style image and the content image,so that the generated stylized image not only has an artistic style but also retains the color matching of the original image.(2)Based on the luminance stylization method,brightness and color are separated from the style image and the content image,and the luminance channel is stylized.Finally,the color and brightness stylization results are fused to obtain the final generated image.From the experimental results,the method of this paper can improve the color confusion,the lack of art style and artistic aesthetics in the benchmark method,and obtain a more visual artistic impression.