Method for compensating for void-defects in images

• US 4,698,843 A
• Filed: 08/19/1985
• Issued: 10/06/1987
• Est. Priority Date: 08/19/1985
• Status: Expired due to Term

• Current Assignee: RCA Corporation (General Electric Company)
• Original Assignee: RCA Corporation (General Electric Company)
• Inventors: Burt, Peter J., Adelson, Edward H.
• Primary Examiner(s): Boudreau, Leo H.
• Assistant Examiner(s): Couso, Jose L.
• Application Number:
  US06/766,932
• Time in Patent Office: 778 Days
• Field of Search: 364/723, 364/724, 364/725, 364/572, 364/825, 358/133, 358/160, 358/166, 382/17, 382/22, 382/41, 382/43, 382/49, 382/50, 382/54, 382/55
• 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 5/77 Retouching; Inpainting; Scratch removal
• CPC Class Codes: H04N 19/63 using sub-band based transform, e.g. wavelets
• CPC Class Codes: H04N 19/895 in combination with error concealment

Abstract

A multistage pyramid analyzer, modified to include extrapolation of the input to each stage, together with a pyramid synthesizer, may be used for filling in one or more void defects in an image to provide a natural looking processed image.

Claims

1. An image processing method for filling in at least one void defect in a relatively high-resolution sampled original image, wherein said image occupies a region of n-dimensional space, where n is an integer having a value of at least one, and said void defect is comprised of a plurality of contiguous zero-valued samples occupying a certain sub-region of said region, whereby the perimeter of said void sub-region forms a boundary between the zero-valued samples of said void sub-region and image-defining valued samples surrounding that void sub-region;

and wherein said sampled original image in each dimension thereof has a given relatively high sample density;

said method comprising the steps of;

(a) employing the image-defining valued samples surrounding said void sub-region to derive therefrom extrapolated-valued samples and replacing only those zero-valued samples that form the perimeter of said void sub-region with said extrapolated-valued samples, thereby to derive a modified image in which said void sub-region of zero-valued samples is reduced in size and the perimeter of said reduced-size void sub-region has a perimeter that is situated a single sample distance from the perimeter of said void sub-region prior to said replacement,

(b) low-pass filtering and sub-sampling said modified image to derive a lower resolution sampled image having a sample density in each dimension thereof which is one-half of that of said modified image.

(c) substituting said lower resolution sampled image for said sampled original image and repeating steps (a) and (b) as many times as is required to derive a processed sampled image in which said void sub-region is completely filled in.

2. The method defined in claim 1, wherein step (c) further includes the step of employing expansion and interpolation to derive said processed sampled image with samples having the same given relatively high sample density in each dimension thereof as said sampled original image.

3. The method defined in claim 1, wherein n is equal to two.