| Non-photorealistic rendering of images is an important research direction in the field of computer graphics,which refers to the use of computers to simulate the styles of various visual arts and draw images with artistic styles.The non-photorealistic rendering pursues more visual effects and emphasizes the audience’s emotional perception.Stylized shadow rendering has been a popular research topic for non-photorealistic rendering.Stylized shadows are fundamentally different from physically real shadows in that stylized shadows are more artistically oriented.Shadows are drawn by the artist,depicting the mood of the character and expressing the artist’s emotions,and are more focused on freedom from the constraints of the structure of physically real shadows.In some cartoon movies,animation and games,stylized shadows are used to enhance the visual effect of the scene.Drawing shadows for simple stroke drawings and line drawings is the basis for the further creation of various artworks.Artists draw shadows that constantly adjust to the structure of the image and the direction of light,which takes a lot of time and effort.In order to improve the efficiency of drawing stylized shadows,this thesis proposes a stroke-based shadow drawing method to generate stylized soft shadow effects for simple stroke drawings and line drawings.Meanwhile,this thesis proposes a deep learning-based shadow drawing method to generate stylized hard-shadow effects for line drawings.The main work and innovations of this thesis are as follows.(1)In order to improve the quality of stylized shadows generated for simple stroke drawings and line drawings based on the stroke calculation method,this thesis proposes a method for calculating stroke density for simple stroke drawings and line drawings,which can automatically add shadows to simple stroke drawings and line drawings by stroke density calculation,creating a new stylized soft-shadow effect.In this thesis,the original image and the processed image are calculated separately.The disturbance of image RGB value is proposed for the characteristics of single color and small stroke density in simple stroke drawings and line drawings.At the same time,a new stroke density calculation method is proposed to effectively calculate the stroke density of simple stroke drawings and line drawings,which solves the problem of difficult to estimate the stroke density due to the small stroke density in simple stroke drawings and line drawings.The image RGB value disturbance and stroke density calculation generate a stroke density map for the subsequent lighting calculation.After generating a lighting effect map,it is combined with the original image to generate stylized shadows and lighting effects for the image.For the convenience of users,this thesis creates a user interface based on the method of adding shadows to simple stroke drawings and line drawings by stroke density calculation.Users can move the mouse,change the direction of the light,and view the shadows generated by the image in real time.Meanwhile,different styles of shadows can be generated for the users.(2)This thesis presents a deep learning-based method for generating shadows for line drawings.Based on Star GAN,a shadow generation adversarial network(Shadow GAN)is designed,which can automate the creation of stylized shadows with different light directions.This method defines eight light directions.Users can select one of the eight light directions around the 2D image to specify the light source according to the encoding of the light direction,and generate the shadow corresponding to the light direction.Input image and target shadow state labels,the network is trained to accurately identify the direction of shadow and light.Due to the small size of the line drawing shadow dataset,the format of the existing data does not match the shadow network.In this thesis,a new dataset is created containing line drawings with shadows and label information corresponding to the light direction of the shadows. |
