Style Transfer Between Images

This thesis focuses on three important problems in image processing.First,it focuses on color transfer between images;secondly digital makeup transfer between images and thirdly photographic painting style transfer between images.In short,it aims to study a problem of transferring color and style between images of different perspectives with different circum-stances.We would like to provide the details of our research which has been done on these directions as follows:Color/style transfer between multiple images:We present a novel approach of local color transfer between images based on the simple statistics and locally linear embedding.A sketching interface is proposed for quickly and easily specifying the color correspondences between target and source image.The user can specify the correspondences of local region using scribes,which more accurately transfers the target color to the source image while smoothly preserving the boundaries,and exhibits more natural output results.The distinctive salient features of our proposed method are described as follows:·It is not restricted to one-to-one image color transfer and can take multiple target images;·It does not require to choose the same color style and image size between source and target images;·Computational load is significantly reduced by proposing the sub-sampling;·It preserves the boundary and color details in the target images more effectively.Digital makeup transfer between images:We present a novel approach of creating face makeup upon a face image with another images as the style examples.Our approach is analogous to physical makeup,as we modify the color and skin details while preserving the face structure.The main advantages of our method over existing techniques are as follows:·It does not need face matching,as one could use more than one example images;·It is not restricted to head shot images as we can also change the makeup style in the wild;·It does not require to choose the same pose and image size between subject and example images;·The advantage of reducing limitations of previous techniques benefits us with results of exquisite nature with better visual effects.Photographic painting style transfer between images:In this research direction,we have proposed an automatic photographic painting style which takes an example image as an input and efficiently transfer the painting style to a source image.The main advantage to our technique is that our technique is not restricted to the headshot images and produces good results with better visual affects of changing the painting style in the wild.The distinctive features of our technique are as follows:·It performs the labeling and style transfer simultaneously which considerably reduce the algorithmic complexity;·It is not restricted to head shot images as we can also change the photographic painting style in the wild;In short,we have proposed novel techniques in all the directions mentioned above.Our techniques considerably reduce the limitations of previous techniques of similar nature.They significantly broaden the diversity of both direction of research within the wide spectrum of image processing.