An object extraction method for image synthesis
Seiki Inoue · Systems and Computers in Japan · 1992
Abstract In image synthesis, extraction methods to extract meaningful regions (simple meanings) such as people or flowers from ordinary natural images are important. Although the chroma‐key method is widely used to extract these meaningful regions, it requires a blue background, a special photographic environment, and cannot be applied to natural images. Video‐matte, which extracts objects assisted by the user, has been developed and put to practical use as a method applicable to natural images. In this method, however, the region's boundary must be accurately specified in units of picture elements (pixels) and demands a great deal of effort and skill. In this paper, we propose a new method that simplifies the task of specifying the region in video‐matte. The user only has to specify the object to be extracted, which is easily done; then the actual region extraction is performed automatically by image processing techniques. In addition, a new thinning method for obtaining precise boundaries is also proposed. This method has many practical features: applicability to any natural image, simple specification of the boundary, a one‐pixel wide boundary with no branches or breaks, and no parameters.