ウェーブレット変換を用いた画像復元理論

概要

論文の詳細を見る
In this paper a new theory for image restoration using wavelet transform is proposed. The multi-resolution approximation using orthogonal wavelet gives a method for decomposition of a given signal into a low-pass singal and a high-pass signal with less resolution. This process is reversible, that is, the original signal can be reconstructed from the two less-resolution signals. This multi-resolution approximation can generalized for digital images. Let a given image be degraded by a Gaussian averaging and additive noise. A new idea for image restoration is as follows. The degraded image is decomposed into 4 less-resolution images which are one low-pass inage and three high-pass ones. If the three high-pass images are intensified by mutiplying a given constant and an image with original resolution is reconstructed using the four images, then the reconstructed image will be close to the original image. From this idea, a new theory for image restoration is formulated. Results are compared with other restoration methods, which supports high efficiency of the proposed method.
東海大学の論文