Research On Non-uniform Image Cartoon Stylization Technique

The development of computer technology promotes the developmentof image rendering. As a branch of non-photorealistic rendering, imagecartoon stylization becomes a hot spot of the research. One of the primarytasks of the image stylization is to enhance the visual characteristics ofspecific objects while removing some unimportant information.Nowadays, the technique of generation of cartoon-style images has beenapplied to many fields, such as movie, television and advertisement.This paper introduces the related research background on the imagestylization and analyses some automatic processing techniques. In orderto further improve the effect of the cartoon stylization, this paperproposes a non-uniform image stylization approach based on thesegmentation of visual saliency.The general algorithm of this article is as follows. First of all, weobtain the saliency map based on the frequency-tuned technique and detect the whole foreground regions with the Mean-Shift segmentation.Then, the bilateral filtering technique is employed to reduce the visualdetails of the input image and smooth the whole image. Next, thecoherent line drawing technique is used to enhance the structuralboundaries in the image. The color quantization function can effectivelyproduces the color blocks which are common in cartoon images. Finally,the foreground regions after the saliency segmentation are used asinterpolation factors for achieving the task of the automatic non-uniformimage cartoon stylization. Experimental results show that the proposedapproach can produce foreground-enhancing cartoon-like images in anautomatic manner.