Annual Progress Report, Office of Naval Research Grant No. N00014-95-1-1163 (AASERT).

摘要: Abstract : The problems of image sequence compression and new view synthesis have both received a lot attention recently. In the former case, it is desired to compactly represent original set by exploiting redundancy correlation. This issue particularly important in applications storage transmission. contrast, goal generate arbitrary viewpoints given scene primarily for visualization purposes. Notice that there exists trade-off between representation size quality synthesized images: As more views are added representation, increases as does size. Hence, an interesting problem consider at once; is, construct compact which reconstructs images synthesizes views. We propose two depth-based representations address these problems. first approach involves several so called reference frames depth intensity information defined. New generated warping data manner similar interpolation techniques. second proposed integrates all available with respect single frame akin layered representations. then consists multivalued array values overcomes occlusions 4. These assume sequences arise from static 3-D captured moving camera restricted x-y plane. Note exact motion unknown alpha priori will be estimated.