Adaptive Image Interpolation and Scaling

• Main Authors: Che-Cheng Hu, 胡哲誠
• Other Authors: Jiann-Der Lee
• Format: Others
• Language: zh-TW
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/99688875446654665098

碩士 === 長庚大學 === 電機工程研究所 === 95 === Cubic convolution is a popular method for image interpolation. In the kernel of cubic convolution, there is an adjustable parameter whose value is -0.5. This value is not optimal for all images. In this thesis, we propose a method that selects the parameter adaptively and apply cubic interpolation to image scaling. The original image is divided into non-overlapped blocks. The next step is to find the parameter suitable for each block. The final step scales original image with the optimal scaling parameter which corresponding divided blocks to obtain scaled image. Besides the separable kernels, we also use nonseparable kernels that have two adjustable parameters. Experiments indicate that the scaled images have better quality than those by fixed-parameter cubic convolution if optimal parameter is found. Using suboptimal parameters, the scaled images have compatible quality.