主成分分析による線状陰影の検出法とリアルタイム透視画像処理への応用 An Image Processing Method for Fluoroscopy using a Linear Shadow Detection
A noise reduction method with preserving the contrast of linear shadows on x-ray fluorography, such as guide wires and catheters in IVR procedure, is proposed. In the method, the direction of the linear shadow (θ) and the strength of its directivity (ζ) are calculated locally, then a space-variant filter is applied. The value of θ and ζ on a local area are obtained by the calculation of two principal components of the density distribution on an x-ray image. When an obvious linear shadow is running on an area, the strong value of ζ is obtained by calculating the difference between two eigen values of principal components, and the θ is the same direction of the eigen vector. In this case, the kernel of the spatial filter is also directive, which averages the pixels in the direction of θ. Inversely, when no linear shadow is running, zero or weak ζ is obtained and the isotropic low-pass filter is applied to reduce noise. This method is tested on a fluoroscopic image with the incident dose of 5μR. The result shows significant reduction of the background noise with keeping wire contrast, the SNR of a wire (0.25mm in diameter) improved by around 120%.
- 医用画像情報学会雑誌 = Japanese journal of imaging and information sciences in medicine
医用画像情報学会雑誌 = Japanese journal of imaging and information sciences in medicine 21(3), 239-251, 2004-09-01