パラメータ可変型2次補間カーネルによるディジタル画億の解像度向上

摘要

ディジタル画像は，取得するデバイスによってそのサrnイズが決定してしまい，その画像をモニタなどの表示デrnバイスに出力する際，画像のサイズを合わせるための解rn像度変換，すなわち拡大・縮小処理が必要となる．画像rnの拡大処理は2次元信号の補間が必要であるが，1次元rnで定義された補間カーネルを，画像を構成する縦横の1rn次元信号に対して縦・横方向にそれぞれ施すことで達成rnする方法があり，これまでに多くの1次元補間カーネルrnが提案されている［1-7］．%This paper presents a quadratic kernel with four variable parameters for image enlargement. The proposed kernel can generate peak and valley information of the signals by considering the characteristics of the image and by changing these parameters. Moreover, as the proposed kernel is the extension of the linear interpolation and the nearest-neighbor interpolation, the forms of the proposed kernel are also used selectively depending on the characteristics of signal points which are used for parameter estimation. The various standard images are enlarged using the proposed kernel, and the performance of the proposed kernel is evaluated with MSE of the enlarged images. It has been shown that the performance of the proposed kernel is better than the other methods, such as the cubic interpolations that are often used in image processing.