PicWords: Generator and viewer

PicWords is an automatic non-photorealistic rendering (NPR) packing system which takes a source picture and some keywords introducing the information about the picture as input, and generates the source picture composed of the related keywords as output. Given a source picture, we first convert it into its silhouette form, which is a binary image that contains a Yang part and a Yin part. Yang part is considered to be significant and can be used for placing keywords while the Yin part can be removed manually. Next, the Yang part is segmented into smaller patches which serve as containers of the keywords, each containing a keyword. To ensure that more significant keywords are put into larger image patches, we rank the patches as well as the keywords. The latter is ranked manually. Thus a correspondence is constructed between the patch list and the keyword list. Then, mean value coordinates method is employed for the keyword-patch warping using Open GL. PicWord generator is implemented using a set of core Java classes and uses C++ image processing library through Java Native Interface (JNI). Finally, certain post-processing techniques are adopted to improve the aesthetics of PicWords. The PicWord Viewer User Interface will allow user to select PicWord image and passes clicks to a window with the keywords and the URL related to the picture.