Conference proceeding
Scale-based method for correcting background intensity variation in acquired images
Proceedings of SPIE, Vol.4684(1), pp.1103-1111
Medical Imaging 2002: Image Processing
05/15/2002
DOI: 10.1117/12.467067
Abstract
An automatic, acquisition-protocol-independent, entirely image-based strategy for correcting background intensity variation in medical images has been developed. Local scale - a fundamental image property that is derivable entirely from the image and that does not require any prior knowledge about the imaging protocol or object material property distributions - is used to obtain a set of homogeneous regions, no matter what each region is, and to fit a 2nd degree polynomial to the intensity variation within them. This polynomial is used to correct the intensity variation. The above procedure is repeated for the corrected image until the size of segmented homogeneous regions does not change significantly from that in the previous iteration. Intensity scale standardization is effected to make sure that the corrected images are not biased by the fitting strategy. The method has been tested on 1000 3D mathematical phantoms, which include 5 levels each of blurring and noise and 4 types of background variation - additive and multiplicative Gaussian and ramp. It has also been tested on 10 clinical MRI data sets of the brain. These tests, and a comparison with the method of homomorphic filtering, indicate the effectiveness of the method.
Details
- Title: Subtitle
- Scale-based method for correcting background intensity variation in acquired images
- Creators
- Ying Zhuge - University of PennsylvaniaJayaram K Udupa - University of PennsylvaniaJiamin Liu - University of PennsylvaniaPunam K Saha - University of PennsylvaniaTad Iwanage - University of Pennsylvania
- Resource Type
- Conference proceeding
- Publication Details
- Proceedings of SPIE, Vol.4684(1), pp.1103-1111
- Conference
- Medical Imaging 2002: Image Processing
- DOI
- 10.1117/12.467067
- ISSN
- 0277-786X
- Language
- English
- Date published
- 05/15/2002
- Academic Unit
- Electrical and Computer Engineering; Radiology
- Record Identifier
- 9984197359902771
Metrics
3 Record Views