Graphics processing unit‐accelerated high‐quality watercolor painting image generation

Abstract

AbstractStroke‐based rendering is a rendering method that mimics the actual painting technique by drawing a stroke by stroke on a blank canvas image. In this paper, we propose a watercolor image generation method using stroke‐based rendering. The proposed method generates an image that is a good approximation of the input image as well as having the characteristics of a watercolor painting by repeatedly painting strokes while referring to the input image. To generate a high‐quality image, that is, an image that closely resembles an actual watercolor painting, various techniques are employed: modeling of watercolor paper, detailed physical simulation of the movement of water and pigment, strokes using a brush model, among others. The proposed method generates a large number of strokes and performs computationally intensive watercolor simulations for each stroke. Therefore, this paper also presents its parallel algorithm using a Graphics Processing Unit (GPU). We implemented this parallel algorithm on an NVIDIA A100 GPU. The experimental results show that the CPU implementations with sequential and parallel executions take 34,651 and 867 s to generate a 4K‐watercolor image of size , respectively. In contrast, the GPU implementation with parallel execution succeeded in reducing the time to 44 s.