A Hybrid Model for Sprite Generation

• Main Authors: Cheng-Jan Chi, 吉正然
• Other Authors: Chin-Hsing Chen
• Format: Others
• Language: en_US
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/pkej6s

碩士 === 國立成功大學 === 電機工程學系碩博士班 === 90 === A novel sprite generator, base on Yan Lu’s work in MPEG-4 Optimized Model (MPEG-4 OM), is presented in the thesis. Three improvements are included in the generator: fast image warping, hybrid model global motion estimation (GME) and adaptive switching. To accelerate the speed of image warping, a simpler interpolation kernel is used. The bilinear interpolation kernel is replaced by nearest neighborhood (NN) kernel. Hybrid model GME is used to enhance the image quality and estimation accuracy. With hybrid model GME, the sprite generated with NN kernel has equal or even better quality compared with those generated with Yan Lu’s method. The adaptive switching chooses the proper GME model frame by frame. It saves the computation time and the transmission data rate. Four test sequences are used to test the new generator. Two of the sequences are reversed frame by frame. The experimental results show that the image quality is improved more if the camera operations are more complex and faster. The computation speed is 1.7 times faster or equal to the OM generator.