| Summary: | 碩士 === 國立中興大學 === 資訊網路多媒體研究所 === 98 === In this thesis, we propose a new exploitation for the multi-resolution method by suitable decomposition scale, then extracts and fuses the feature for regions of images.
On brain medical image, reasearchers usually fusing PET color images and MR gray image to become a new fusion image. Our fusion method is divided two stages, the first stage is IHS transformation for PET images, combine the hue of PET images and the activity information of brain cells to separate the high and low activity regions for input images. The second stage is log-Gabor wavelet transformation with suitable decomposition scale and orientation for the separated regions, extracting the high frequency parts of MRI to keep structure information and the low frequency parts of intensity of PET images to keep color information, the purpose is simultaneously keep structure and color information for the fusion image. The choice frequency parts inverse log-Gabor wavelet transformation, then add these regions and inverse IHS transformation to produce a new fused image. The experiment results appear our method reduce the color discrepancy and fuse the better MRI structure, it is a good fusion performance for comparing other fusion methods.
|
|---|
