A-DEWS 2016 - Innovation of Life in Asia - Asian Design Engineering Workshop

Year: 2016
Editor: Umeda, Y.; Nomaguchi, Y.; Koga, T.; Kishita, Y.
Author: Umezu, Nobuyuki
Institution: Ibaraki University
Section: CAD/CAM and Computer Graphics
Page(s): 142-148


Automatic generation of photomosaic is generally performed by dividing a target image into uniform grid regions and replacing each grid with a small material image with the color distribution similar to the grid in the target. Their results, however, have a disadvantage in that the arrangement and size of the material images are limited, and in that one can hardly recognize each of material images because they are extremely small. A new method is proposed to generate photomosaic that maintains boundaries of both the target image and the material images. This can be achieved by considering salient features at boundaries in images and by varying size of material images by using several levels of resolution. The output photomosaic generated by the presented method has material images of larger size and finer boundaries. Experiments using human subjects were also conducted to evaluate the results with human visual system. Extracting features from a large number of material images requires tens of seconds, for example, 83.7 seconds for 2,000 images with 640 x 480 pixels on a PC with Core i5 2.7GHz.

Keywords: Image quality assessment, Multiresolution, Visibility, Boundary patterns

Please sign in to your account

This site uses cookies and other tracking technologies to assist with navigation and your ability to provide feedback, analyse your use of our products and services, assist with our promotional and marketing efforts, and provide content from third parties. Privacy Policy.