Simulation scheme of dusk scene using piece-wise multiple regression based on time-series color-block images

摘要

Abstract. Dusk and dawn are usually the most beautiful moments of thenday, and are almost always too short for busy people nowadays to wit-nness their coming. In this work, an efficient strategy for simulating a dusknscene of an outdoor scene image taken at other times before the sunsetnis presented. The strategy is a hybrid approach combining the piece-nwise multiple regression u0001PMRu0002 of data, discrete cosine transformationnu0001DCTu0002, and a look-up table algorithm. The process begins using a seriesnof color-block images taken in the afternoon of a day. The best fittingnfunctions of PMR for these color block images exist on separate planesnu0001red, green, and blueu0002 in the DCT domain individually. The referencendatabases of the DCT coefficients varying with respect to time are thennestablished according to the best fitting functions of PMR. Finally, thendusk scene of an outdoor scene taken in the afternoon is synthesized bynquerying the reference database. The experiment results show that thenpresented algorithm can precisely simulate the desired dusk scene fromna scene image taken in the afternoon. © 2010 Society of Photo-Optical Instrumen-ntation Engineers. u0003DOI: 10.1117/1.3487744