Photosketcher series#
This has been shown to work well for filling holes in given images or stitching together a single image from a series of photographs with similar content. The idea is that it is more likely that each part of the desired scene is part of a picture in the collection than finding a single, perfectly matching image. Instead of searching for a single image that perfectly fits the requirements, the image is interactively composed out of parts of images. Our approach, named Photosketcher, combines two main ideas: 1. We propose a system that combines aspects of all of the solutions above: it is based on photographs but lets the user interactively search and compose them, allowing anyone to create complex pictures without any particular artistic skill. Both solutions require talent, skill, and experience. It seems that for a more complete rendition of the mental image it is necessary to create it with the help of a computer, either by directly drawing it, or by creating a 3D scene that can be rendered into an image. Yet, it is to be expected that any existing photograph is only an approximation of your imagination. With the advent of large open image collections it is possible that a similar photograph has already been taken and made available. How would you create a digital representation of this image? If the scene exists in reality it could be photographed, however, this is expensive and time-consuming.
1 Introduction Suppose you are thinking of an image depicting a real scene. We demonstrate that the resulting system allows interactive generation of complex images. Optionally, Photosketcher lets users blend the composite image in the gradient domain to further reduce visible seams. The compositing step again is based on user scribbles: from the scribbles we predict the desired part using Gaussian Mixture Models and compute an optimal seam using Graphcut. The search is based on a bag-offeatures approach using local descriptors for translation invariant part retrieval. Users sketch the rough shape of a desired image part and we automatically search a large collection of images for images containing that part. Compared to existing approaches for synthesising images from parts of other images, Photosketcher works on the image content exclusively, no keywords or other metadata associated with the images is required. Abstract We introduce Photosketcher, an interactive system for progressively synthesizing novel images using only sparse user sketches as the input. Photosketcher: interactive sketch-based image synthesis Mathias Eitz 1, Ronald Richter 1, Kristian Hildebrand 1, Tamy Boubekeur 2 and Marc Alexa 1 1 TU Berlin 2 Telecom ParisTech/CNRS Figure 1: Photosketcher: realistic images from user sketches using large image collection queries and interactive image compositing.
0 kommentar(er)
