Image analysis – Image enhancement or restoration – Image filter
Reexamination Certificate
1999-11-22
2004-03-02
Johnson, Timothy M. (Department: 2625)
Image analysis
Image enhancement or restoration
Image filter
C382S194000, C382S131000, C382S264000, C382S274000, C382S298000
Reexamination Certificate
active
06701025
ABSTRACT:
BACKGROUND OF THE INVENTION
The field of the invention is the enhancement of digital images acquired using various imaging modalities. More particularly, the invention relates to producing higher-quality digital images using a combination of noise-reduction and non-uniformity correction techniques.
Various techniques have been developed for acquiring and processing discrete pixel image data. Discrete pixel images are composed of an array or matrix of pixels having varying properties, such as intensity and color. The data defining each pixel may be acquired in various manners, depending upon the imaging modality employed. Modalities in medical imaging, for example, include magnetic resonance imaging (MRI) techniques, X-ray techniques, and ultrasonic techniques. In general, each pixel is represented by a signal, typically a digitized value representative of a sensed parameter, such as an emission from material excited within each pixel region or radiation received within each pixel region.
To facilitate interpretation of the image, the pixel values must be filtered and processed to enhance definition of features of interest to an observer. Ultimately, the processed image is reconstituted for displaying or printing. In many medical applications, an attending physician or radiologist will consult the image for identification of internal features within a subject, where those features are defined by edges, textural regions and contrasted regions.
Unless further processing is applied to a digital image, the image is likely to have a poor signal to noise ratio (SNR), resulting in blurred or ambiguous feature edges and non-uniformities in spatial intensity. Structures, textures, contrasts, and other image features may be difficult to visualize and compare both within single images and between a set of images. As a result, attending physicians or radiologists presented with the images may experience difficulties in interpreting the relevant structures.
With respect to non-uniformity correction, in many areas of imaging including MRI and computed tomography, acquired images are corrupted by slowly varying multiplicative inhomogeneities or non-uniformities in spatial intensity. Such non-uniformities can hinder visualization of the entire image at a given time, and can also hinder automated image analysis. Such inhomogeneity is a particular concern in MRI, when single or multiple surface coils are used to acquire imaging data. The acquired images generally contain intensity variations resulting from the inhomogeneous sensitivity profiles of the surface coil or coils. In general, tissue next to the surface coil appears much brighter than tissue far from the coil. Therefore, in order to optimally display and film the entire image, the signal variation due to the inhomogeneous sensitivity profile of the surface coil needs to be corrected.
Several prior art methods either enhance features or correct for non-uniformities, but not both. For example, existing techniques for enhancing features may require operator intervention in defining salient structures, sometimes requiring processing of raw data several times based on operator adjustments before arriving at an acceptable final image. This iterative process is inefficient and requires a substantial amount of human intervention. Other prior art methods have been developed for enhancing features of the image while suppressing noise. For example, in one known method, pixel data is filtered through progressive low pass filtering steps. The original image data is thus decomposed into a sequence of images having known frequency bands. Gain values are applied to the resulting decomposed images for enhancement of image features, such as edges. Additional filtering, contrast equalization, and gradation steps may be employed for further enhancement of the image.
While such techniques provide useful mechanisms for certain types of image enhancement, they are not without drawbacks. For example, gains applied to decomposed images can result in inadvertent enhancement of noise present in the discrete pixel data. Such noise, when enhanced, renders the reconstructed image difficult to interpret, and may produce visual artifacts which reduce the utility of the reconstructed image, such as by rendering features of interest difficult to discern or to distinguish from non-relevant information.
Prior art methods such as that disclosed in U.S. Pat. No. 5,943,433 have also been employed for correcting non-uniformities, although not simultaneously with the above-described methods for feature enhancement. Prior art methods for correcting for non-uniformities include various intensity correction algorithms which correct surface coil images by dividing out an estimate of the surface coil's sensitivity profile. One such method is based on the assumption that distortion arising from use of surface coils generally varies slowly over space. In accordance with that prior art method, a low pass filtering operation is applied to the measured or acquired image signal. For this prior art method to be effective, however, the image signal must not contain sharp intensity transitions. Unfortunately, at least in MRI imaging, an air-lipid interface usually contains sharp intensity transitions which violate the basic assumption that the low frequency content in the scene being imaged is solely due to the inhomogeneity distortion from the surface coil's sensitivity profile.
Accordingly, certain prior art hybrid filtering techniques have been developed. Although these techniques have been effective in accounting for external transitions, they have not been particularly effective in accounting for significant internal transitions (e.g., transitions that occur between the edges of an organ or other tissue structure).
As stated before, acquired images are corrupted by slowly varying multiplicative non-uniformities. When such images are corrected using prior-art techniques, substantial noise amplification can occur, which hinders the visualization of salient features. Therefore, it is common to use less correction than optimal to prevent noise amplification. Besides using less correction, the image may be pre-filtered to reduce noise. Such pre-filtering, however, can also remove salient features from the image. Thus, the combination of pre-filtering and non-uniformity correction techniques has not been put into practice because the combination of prior-art methods has resulted in less-than-optimal images.
In image processing literature, several techniques are described to separately improve the SNR and non-uniformity in images. Many authors have described enhancing SNR in MRI images by spatial domain filtering. Likewise, several articles describe improving the shading by correcting for the non-uniformity in the images. Usually these two operations are treated as though they are disjointed operations.
R. Guillemaud and M. Brady have discussed simultaneous correction for noise and non-uniformity in IEEE Transactions in Medical Imaging, Vol. 16, pp. 238-251 (1997). These authors used the anisotropic diffusion-based technique proposed by G. Gerig et al., IEEE Transactions in Medical Imaging, Vol. 11, pp. 222-232 (1992), for noise reduction for both pre- and post-filtering with non-uniformity correction. They concluded that pre-filtering loses essential details in the non-uniformity corrected images. Therefore, Guillemaud and Brady chose to perform post-filtering of non-uniformity corrected images. This decision indicates that prior art methods obtain visibility of important weak structures at the expense of non-linear noise amplification.
What is needed is an automated method and apparatus which improves the visual quality of digital images. Particularly needed is an automated method and apparatus which reduce noise while correcting for non-uniformities in the image, resulting in better-quality images than were possible using prior art techniques.
BRIEF SUMMARY OF THE INVENTION
The present invention includes a method and apparatus for automatically correcting a digital image, and pa
GE Medical Systems Global Technology Company LLC
Horton Carl
Johnson Timothy M.
Quarles & Brady LLP
LandOfFree
Medical image enhancement using iteration to calculate an... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Medical image enhancement using iteration to calculate an..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Medical image enhancement using iteration to calculate an... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3210819