The Study of Image Zooming Techniques

• Main Authors: Jun-Rong Zeng, 曾俊榕
• Other Authors: Yu-Chen Hu
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/30498076442857500465

碩士 === 靜宜大學 === 資訊管理學系 === 99 === In image zooming issues, it can be separate into two aspects, adaptive and non-adaptive zooming. The difference between adaptive and non-adaptive is that non-adaptive technique use one algorithm to reconstruct whole pixels in target image, without considering the edges or pixel’s correlation. And adaptive technique usually uses two or more algorithms to deal with, considering the edges, pixel’s correlation, or even the type of target image, this technique can be mutative by different person. In this thesis we proposed two adaptive zooming techniques. The first one combines bilinear interpolation and vector quantization algorithm and we set a threshold to separate the type of needy reconstruct pixels into smooth or complex. When the pixel is smooth, we use bilinear interpolation to rebuild. After all smooth pixels have been rebuilt well, we divide whole image into the same blocks, and step by step check all blocks have been rebuilt or not, if a block haven’t rebuilt yet, we use vector quantization algorithm to deal it and till all blocks have been rebuilt. In the second method, we keep down the gradient from original image, and set a threshold to classify the type of each needy reconstruct pixels into smooth or complex. And smooth pixels use bilinear to rebuild it. Then based on the gradient we can find the edge directions, and rebuilt different edges. The experimental results show that the proposed methods can improve the ambiguous in edges, and the image quality is also better than others.