Image processing method using a block overlap transformation procedure

• US 4,442,454 A
• Filed: 11/15/1982
• Issued: 04/10/1984
• Est. Priority Date: 11/15/1982
• Status: Expired due to Term

• Current Assignee: Eastman Kodak Company
• Original Assignee: Eastman Kodak Company
• Inventors: Powell, Philip G.
• Primary Examiner(s): Murray, Richard
• Application Number:
  US06/441,826
• Time in Patent Office: 512 Days
• Field of Search: 358/167, 358/166, 358/163, 358/135, 358/136, 358/36, 358/37, 358/133
• US Class Current: 348/618
• CPC Class Codes: G06T 2207/20016 Hierarchical, coarse-to-fine, multiscale or multiresolution image processing; Pyramid transform
• CPC Class Codes: G06T 2207/20021 Dividing image into blocks, subimages or windows
• CPC Class Codes: G06T 5/10 using non-spatial domain filtering
• CPC Class Codes: G06T 5/20 using local operators
• CPC Class Codes: G06T 5/70 Denoising; Smoothing
• CPC Class Codes: H04N 1/409 Edge or detail enhancement; Noise or error suppression
• CPC Class Codes: H04N 19/86 involving reduction of coding artifacts, e.g. of blockiness
• CPC Class Codes: H04N 5/21 Circuitry for suppressing or minimising disturbance, e.g. moiré or halo

Abstract

An image processing method is provided for reducing noise in a sampled image, particularly for reducing noise in an image divided into blocks of sampled image elements that are transformed by a linear procedure, such as the Walsh-Hadamard transform, and improved regarding visible noise by non-linear thresholding of the transform coefficients. By operating the process in a hierarchy of stages, each stage employing a block operating on image signals derived from a preceding stage, and by overlapping the blocks processed in each stage, the processed signal from each image element is the linear combination of many transform coefficients from each stage and from each overlapped block within each stage. Such a large number of contributions making up each processed image element assures that the processed image is generated without a characteristic block-like structure due to block transform processing while the wanted components of the image are rendered with minimal image loss or distortion.

Claims

1. A method of processing an original image in order to generate a processed image having fewer unwanted components than the original image, said method comprising the steps of

A. generating image signals related to the brightness of respective elements of the original image;

B. grouping the image signals into arrays of image signals corresponding to overlapping blocks of image elements, wherein each block includes at least one image element in common with at least one other block;

C. processing the arrays in a hierarchy of separate stages in which the image signals generated for each stage subsequent to the first represent derived signals responsive to combinations of large signals from a preceding stage, each stage comprising the steps of

(1) transforming each array of image signals by a set of independent functions into a corresponding set of coefficient signals representing combinations of image signals sensitive to general brightness and image detail, and

(2) subjecting the coefficient signals to modification in order to minimize the unwanted components in the processed image; and

D. generating a processed image from the sets of coefficient signals, some of which may be modified,

whereby the processed image is generated without a characteristic block-like structure due to block transform processing while the wanted components of the image are rendered with minimal image loss or distortion.

2. A method of processing an original image in order to generate a processed image having fewer unwanted components than the original image, said method comprising the steps of

A. generating image signals related to the brightness of respective elements of the original image;

B. grouping the image signals into image signal arrays having at least one mutually common signal and corresponding to similarly overlapping blocks of image elements, wherein each block includes at least one image element in common with at least one other block;

C. transforming each array of image signals by a set of independent functions into a corresponding set of coefficient signals representing combinations of image signals sensitive to general brightness and image detail;

D. subjecting the sets of coefficient signals to modification in order to minimize the unwanted components in the processed image;

E. performing the foregoing steps (A) to (D) in a hierarchy of separate stages in which the image signals generated for each stage subsequent to the first comprise derived signals corresponding to combinations of image signals sensitive to the brightness of image signals processed in a preceding stage; and

F. generating a processed image from the sets of coefficient signals, some of which may be modified,

whereby the block-like structure characteristic of block transform processing is minimized in the processed image and the rendition of the wanted components of the image is improved with minimal image loss or distortion.

3. A method of block transform processing of a sampled image composed of image elements in order to generate processed image signals having fewer noise components and fewer visible artifacts introduced by the processing method, said method comprising the steps of

A. generating image signals related to the brightness of respective image elements;

B. grouping the image signals into arrays of image signals corresponding to blocks of image elements;

C. aligning the arrays of image signals such that each array includes at least one image signal in common with at least one other array whereby the aligned arrays correspond to overlapping blocks of image elements;

D. transforming each array of image signals by a set of independent functions into a corresponding set of coefficient signals representing (1) a general brightness within each block and (2) one or more estimations of spatial variations of brightness representative of noise components and image detail within each block;

E. modifying at least some of the sets of coefficient signals by (1) selecting certain coefficient signals by comparison to a threshold and (2) altering these selected coefficient signals;

F. performing the foreging steps (A) to (E) in a hierarchy of separate stages in which the image signals generated for each stage subsequent to the first comprise signals derived from combinations of image signals sensitive to the general brightness of the image signals processed in a preceding stage; and

G. generating processed image signals from the sets of coefficient signals, some of which are modified,

whereby noise components are reduced in the processed image without generating the block-like structure characteristic of block transform processing, image detail is rendered with little image loss or distortion and the unwanted artifacts of processing are reduced.

4. The method as claimed in claim 3 in which the step of modifying at least some of the sets of coefficient signals comprises the steps of

A. determining one or more thresholds that represent one or more coefficient magnitudes of such values that greater coefficient magnitudes rarely occur due to noise alone;

B. selecting corresponding coefficient signals that are less than the said one or more thresholds; and

C. altering these selected coefficient signals by reducing their values.

5. The method as claimed in claim 3 in which

A. the step of generating the image signals additionally comprises generating a full-band image signal composed of image signals that are not grouped into arrays or applied to the hierarchy of stages;

B. the step of modifying at least some of the sets of coefficient signals comprises the steps of

(1) determining one of more thresholds that represent one or more coefficient magnitudes of such values that greater coefficient magnitudes rarely occur due to noise alone;

(2) selecting corresponding coefficient signals that are greater than the said one or more thresholds, and

(3) altering these selected coefficient signals by reducing their values; and

C. the step of generating processed image signals comprises generating noise signals from the sets of coefficient signals, some of which are modified, and subtracting the noise signals from the full-band image signal.

6. A method of transform processing of image signals related to the brightness of respective elements of a sampled image composed of a plurality of such image elements, said method comprising the steps of

A. processing the image signals in two or more separate stages, wherein the first stage comprises the steps of

(1) grouping the image signals into arrays of image signals corresponding to blocks of image elements,

(2) aligning the arrays of image signals such that some of the image signals are included in more than one array, whereby the aligned arrays correspond to overlapped blocks of image elements,

(3) transforming each array of image signals by a set of independent functions into a corresponding set of coefficient signals representing (a) an average brightness within each block and (b) one or more estimations of spatial variations of brightness representative of noise and image detail within each block,

(4) modifying at least some of the sets of coefficient signals by (a) selecting certain coefficient signals by comparison to a threshold and (b) altering these selected coefficient signals, wherein the subsequent stages comprise the steps of

(1) generating derived image signals from combinations of image signals representative of the brightness within localized areas of the image processed in the preceding stage,

(2) grouping the derived image signals into arrays of derived image signals corresponding to blocks of derived image elements,

(3) aligning the arrays of derived image signals such that some of the derived image signals are included in more than one array, whereby the aligned arrays correspond to overlapped blocks of derived image elements,

(4) transforming each array of derived image signals by a set of independent functions into a corresponding set of derived coefficient signals representing (a) an average brightness within each block of derived image elements and (b) one or more estimations of spatial variations of brightness representative of noise and image detail within each block of derived image elements,

(5) modifying at least some of the sets of derived coefficient signals by (a) selecting certain derived coefficients by comparison to a threshold and (b) altering these selected coefficient signals;

B. generating a processed image from the sets of coefficient signals processed in each of the stages,

whereby noise is reduced without generating the block-like structure characteristic of block transform processing and image detail is rendered with little image loss or distortion.

7. The method as claimed in claim 6 in which the step of generating derived image signals comprises using the coefficient signals representing the average brightness within the blocks of the preceding stage as the derived image signals.

8. The method as claimed in claim 6 in which the step of generating derived image signals comprises filtering the image signals processed in the preceding stage to provide filtered image signals representing a general brightness of the image signals processed in the preceding stage.