Noise reduction system

• US 5,526,446 A
• Filed: 09/24/1991
• Issued: 06/11/1996
• Est. Priority Date: 09/24/1991
• Status: Expired due to Term

• Current Assignee: Massachusetts Institute of Technology
• Original Assignee: Massachusetts Institute of Technology
• Inventors: Freeman, William T., Adelson, Edward H.
• Primary Examiner(s): Boudreau, Leo
• Assistant Examiner(s): KELLEY, CHRISTOPHER S
• Application Number:
  US07/764,567
• Time in Patent Office: 1,722 Days
• Field of Search: 382/54, 382/49, 382/42, 358/167, 346/616
• US Class Current: 382/275
• CPC Class Codes: G06T 2207/20016 Hierarchical, coarse-to-fine, multiscale or multiresolution image processing; Pyramid transform
• CPC Class Codes: G06T 5/10 using non-spatial domain filtering
• CPC Class Codes: G06T 5/70 Denoising; Smoothing
• CPC Class Codes: G06V 10/30 Noise filtering
• CPC Class Codes: G06V 10/449 Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
• CPC Class Codes: G06V 10/52 Scale-space analysis, e.g. wavelet analysis multi-scale boundary representations G06V10/42

Abstract

A technique is provided to remove noise from images and to enhance their visual appearance through the utilization of a technique which converts an image into a set of coefficients in a multi-scale image decomposition process, followed by modification of each coefficient based on its value and the value of coefficients of related orientation, position, or scale, which is in turn followed by a reconstruction or synthesis process to generate the enhanced image. Also contributing to the improved enhancement is a set of orientation tuned filters of a specialized design to permit steering, with the analysis and synthesis filters also having a self-inverting characteristic. Additionally, steerable pyramid architecture is used for image enhancement for the first time, with the steering being provided by the above orientation tuned filters. The utilization of related coefficients permits coefficient modification with multipliers derived through a statistical or neural-network analysis of coefficients derived through the utilization of clean and degraded images, with the modifiers corresponding to vectors which result in translating the degraded image coefficients into clean image coefficients, in essence by cancelling those portions of a coefficient due to noise. Further improvements include an overlay of classical coring on single coefficients. Thus, the subject technique provides improved image enhancement through the use of a multi-band or scale-oriented analysis and synthesis transform having improved coefficient modification, good orientation tuning, improved bandpass characteristics, and good spatial localization.

Claims

1. Apparatus for the enhancement of an image so as to remove noise and other artifacts therefrom, comprising:

means for scanning said image to produce a scanned image;

means coupled to the said scanning means for convolving a given neighborhood in said scanned image into a number of sets of transform coefficients, each set having a predetermined spatial scale corresponding to a predetermined spatial frequency band characteristic and each set of transform coefficients having been derived from convolving said image with a number of analysis filters;

means including estimates of orientation anisotropy and orientation strength of said scanned image in the neighborhood of the image associated with a transform coefficient at the given neighborhood for modifying each of said transform coefficients to eliminate those portions of the coefficient corresponding to noise by multiplying a coefficient by a modifier, said modifier being a number derived through the comparison of said estimates of orientation anisotropy and orientation strength with orientation anisotropy and orientation strength thresholds derived from related coefficients, said related coefficient being nearby in position to the coefficient being modified or nearby in scale to that of the coefficient being modified or at a different orientation than that of the coefficient being modified in which said thresholds are derived from analysis of multi-scale decomposition of a clean and a degraded image; and

,

means for reconstructing an enhanced image from said modified coefficients, said reconstructing means including a number of synthesis filters.

2. The apparatus of claim 1, wherein said analysis includes deriving cumulative clean and degraded image coefficient histograms and comparing them.

3. Apparatus for the enhancement of an image so as to remove noise and other artifacts therefrom, comprising:

means for scanning said image to produce a scanned image;

means coupled to said scanning means for convolving a given neighborhood in said scanned image into a number of sets of transform coefficients, each set having a predetermined spatial scale corresponding to a predetermined spatial frequency bandpass characteristic and each set of transform coefficients having been derived from convolving said image with a number of analysis filters;

means for modifying each of said transform coefficients to eliminate those portions of a coefficient corresponding to noise by applying a multi-input function to a transform coefficient, said function having at least two inputs, one of said inputs being the value of said transform coefficient and at least one other of said inputs being a value derived from a related transform coefficient, said related coefficient being nearby in position to the coefficient being modified or nearby in scale to that of the coefficient being modified or at a different orientation than that of the coefficient being modified, said related transform coefficient being associated with a portion of the scanned image in the same neighborhood as the portion of the scanned image from which said transform coefficient is derived to provide sets of coefficients that are all gathered from a given neighborhood in said image; and

,

means for reconstructing an enhanced image from said modified coefficients, said reconstructing means including a number of synthesis filters.

4. The apparatus of claim 3, wherein said multi-input function includes means for multiplying a transform coefficient with a number equal to 1 or zero depending on whether a threshold is exceeded or not, said threshold being a function of the orientation strength, orientation anisotropy and coefficient value of a scanned image in the neighborhood of the image associated with said transform coefficient.