Multiple processing of radiographic images based on a pyramidal image decomposition

• US 5,644,662 A
• Filed: 03/28/1995
• Issued: 07/01/1997
• Est. Priority Date: 02/11/1993
• Status: Expired due to Term

• Current Assignee: AGFA Healthcare NV (Agfa-Gevaert N.V.)
• Original Assignee: Agfa-Gevaert N.V.
• Inventors: Vuylsteke, Pieter Paul
• Primary Examiner(s): COUSO, YON JUNG
• Application Number:
  US08/412,355
• Time in Patent Office: 826 Days
• Field of Search: 382/302, 382/128, 382/132, 382/240, 382/260, 382/266, 382/168, 382/169
• US Class Current: 382/302
• CPC Class Codes: G06T 2207/10116 X-ray image
• 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 5/20 using local operators

Abstract

Multiple processing of radiographic images based on a pyramidal image decomposition.

A method of multiple processing a digital representation of a radiographic image is disclosed that is based on a pyarmidal decomposition of said image. A pyramidal decomposition is stored and retrieved to be applied to at least two different processing cycles, processed images are obtained by application of a reconstruction algorithm.

Claims

1. A method of generating more than one differently processed image version originating from a single radiation image, comprising the steps of:

1. applying a decomposition transform to said radiation image to decompose it into a multiresolution representation which represents localized image detail at multiple resolution levels,

2. storing said multiresolution representation into a memory,

3. producing at least two differently processed image versions by applying at least two individual processing cycles to said multi-resolution representation, each processing cycle comprising the steps of;

(i) retrieving said multiresolution representation from said memory,

(ii) modifying the multiresolution representation at at least one resolution level by applying a non-linear monotonically increasing odd modifying function with a slope that gradually decreases with increasing argument values to elements of the multiresolution representation that belong to a same resolution level,

(iii) obtaining a processed image representation by applying the inverse of said decomposition transform to the modified multiresolution representation,

(iv) applying said processed image presentation to a hard copy recorder or to a display device to control generation of density values of a processed image version of said original image.

2. Method according to claim 1 wherein said slope decreases gradually with increasing argument values except for a region of lowest absolute argument values where the slope is constant or increasing.

3. A method according to claim 1 wherein said multiresolution representation is obtained as a sequence of detail images at multiple resolution levels and a residual image at a resolution level lower than the minimum of said multiple resolution levels and wherein the number of pixels in the detail images decreases at each coarser resolution level.

4. A method according to claim 3 wherein the detail image at the finest resolution level is obtained as the pixel wise difference between the original image and an image obtained by low pass filtering the original image, and wherein the successive coarser resolution level detail images are obtained by taking the pixelwise difference between two low pass filtered versions of the original image, the second having a smaller bandwidth than the former.

5. A method according to claim 3 wherein the detail images at successive coarser resolution levels are obtained as a result of K iterations of the following steps:

a) computing an approximation image at a next coarser level by applying a low pass filter to the approximation image corresponding to the current iteration, ans subsampling the result in proportion to the reduction in spatial frequency bandwidth, using the original image as input to said low pass filter in the course of the first iteration;

b) computing a detail image as the pixelwise difference between the approximation image corresponding to the current iteration and the approximation image at a next coarser resolution level computed according to the method sub 4a;

both images being brought into register by proper interpolation of the latter image;

and wherein the residual image is equal to the approximation image produced by the last iteration,

and wherein said processed image is computed by iterating K times the following procedure starting from the coarsest detail image and the residual image;

computing the approximation image at the current resolution level by pixelwise adding the modified detail image at the same resolution level to the approximation image at the coarser resolution level corresponding to the previous iteration, both images being brought into register by proper interpolation of the latter image, using the residual image instead of the coarser approximation image in the course of the first iteration.

6. A method according to claim 5 wherein said subsampling factor is 2, and said low-pass filter has an impulse response which approximates a two-dimensional gaussian distribution.

7. A method according to claim 1 wherein said digital image representation is obtained by scanning a photostimulable phosphor screen that has been exposed to said radiation image with stimulating irradiation, detecting light emitted upon stimulation and converting said detected light into a digital representation.

8. A method according to claim 1 wherein said modifying function depends on the pixel value of a corresponding pixel in said original image.

9. A method according to claim 1 wherein said modifying function is not identical for each resolution level in said decomposition, such that modifications applied to finer resolution levels amplify image detail more than modifications applied to intermediate resolution levels.

10. A method according to claim 1 wherein different modifying functions are used for different resolution levels in said decomposition, such that modifications applied to coarser resolution levels amplify image detail less than modifications applied to intermediate resolution levels.

11. A method according to claim 1 wherein said multi-resolution representation is modified to an extent that depends on an estimation of the amount of variance in the image and in accordance with an estimated noise level.

12. A method according to claim 11 wherein said noise level is determined as the estimated noise level in each detail image.

13. A method according to claim 11 wherein a set of noise suppression functions are computed, each being associated with one of said detail images, said functions being monotonically non-decreasing in one variable and parametrically depending on said noise variance in a non-increasing monotonic way and said suppression functions being positive and asymptotically reaching a maximum value equal to one and wherein each detail image is attenuated by multiplying it with an associated noise suppression function evaluated at an abscissa equal to said local detail image variance.

14. The visual image produced by the process of claim 1.

15. The visual image produced by the process of claim 3.

16. The visual image produced by the process of claim 7.