STRUCTURE PRESERVING IMAGE ABSTRACTION AND ARTISTIC STYLIZATION FROM COMPLEX BACKGROUND AND LOW ILLUMINATED IMAGES
Pavan Kumar, Poornima B H, H. S. Nagendraswamy, Manjunath C, Bidarekere Eshwarappa Rangaswamy · ICTACT Journal on Image and Video Processing · 2020
The paper deals with NPR filtering and image processing techniques to produce the structure preserving image abstraction and artistic stylization effect from complex background and low illuminated images. Structure preserving image abstraction and stylization is most useful in the animation process, film industry, and artistic illustration and education sectors for innovative teaching. Abstraction concept reduces the image complexity and Stylization produces good visual effect to human’s sense. The work involves combining different NPR filtering techniques to create an effective NPR artistic illustration. The proposed technique consists of adoptive structure tensor flow, difference of Gaussian filter, 2D modified coherence shock filter, order dithering and Mean Curvature Flow (MCF). The work involves applying all these techniques in a series and the proposed scheme is found to give a good rendering effect on images with complex background and low luminance images. Moreover the proposed method does not require any kind of post processing techniques for abstraction and artistic stylization. The applied method produces the best abstraction effect and avoids halo effect. Implementation of proposed work is carried out in the Matlab environment. Efficiency of proposal work has been corroborated by conducting different experiments on various types of images and the results are compared with contemporary works. This approach is found to be computationally efficient in rendering effective structure preserving abstraction and stylization to the Human Visual System (HVS) and this approach opens up new research paths towards image and video stylization.