Method and apparatus for resizing images using the discrete cosine transform
AssigneeSarnoff Corporation
InventorStephen Anthony Martucci
Life of this patent
Estimating the term…
Abstract
A method and apparatus using the discrete cosine transform (DCT) to resize an image. To reduce an image, the method and apparatus exploits the convolution-multiplication property of the DCT to implement the anti-aliasing filter in the DCT domain, then the filtered coefficients are operated on to produce DCT coefficients of the reduced-size image. To increase an image, the DCT coefficients of the original image are operated on to produce coefficients of the larger image, then the convolution-multiplication property of the DCT is used to implement an anti-imaging filter. These DCT operations can be applied to the image as a whole or can be applied to non-overlapping blocks of the image.
Claims9
A method of resizing an input image, said method comprising the steps of:
- (a) performing a discrete cosine transform (DCT) on the input image to produce a plurality of DCT coefficients;
- (b) manipulating said plurality of DCT coefficients to effect filtering of the input image by pointwise multiplying said plurality of DCT coefficients by a plurality of transform coefficients of a two-dimensional lowpass filter-right-half;
- and downsampling said plurality of filtered DCT coefficients by folding said plurality of filtered DCT coefficients in both dimensions and applying pointwise subtraction;
- and
- (c) performing an inverse discrete cosine transform (IDCT) on the manipulated DCT coefficients to produce a resized image of the input image.
A method of resizing an input image, said method comprising the steps of:
- (a) performing a discrete cosine transform (DCT) on the input image to produce a plurality of DCT coefficients;
- (b) manipulating said plurality of DCT coefficients to effect filtering of the input image by (b1) upsampling said plurality of DCT coefficients;
- and (b2) filtering said plurality of sampled DCT coefficients;
- and
- (c) performing an inverse discrete cosine transform (IDCT) on the manipulated DCT coefficients to produce a resized image of the input image.
A method of resizing an input image containing a plurality of blocks, said method comprising the steps of:
- (a) performing a discrete cosine transform (DCT) on the input image to produce a plurality of DCT coefficients for each of said plurality of blocks;
- (b) manipulating said plurality of DCT coefficients for each of said plurality of blocks to effect filtering of the input image;
- and
- (c) performing an inverse discrete cosine transform (IDCT) on the manipulated DCT coefficients to produce a resized image of the input image.
Apparatus for resizing an input image, said apparatus comprising:
- a first transforming means for performing a discrete cosine transform (DCT) on the input image to produce a plurality of DCT coefficients;
- a filter for filtering said plurality of DCT coefficients by pointwise multiplying said plurality of DCT coefficients by a plurality of transform coefficients of a two-dimensional lowpass filter-right-half;
- a sampler for sampling said plurality of filtered DCT coefficients by folding said plurality of filtered DCT coefficients in both dimensions and applying pointwise subtraction;
- and
- a second transforming means for performing an inverse discrete cosine transform (IDCT) on the sampled DCT coefficients to produce a resized image of the input image.
Apparatus for resizing an input image, said apparatus comprising:
- a first transforming means for performing a discrete cosine transform (DCT) on the input image to produce a plurality of DCT coefficients;
- a sampler for sampling said plurality of DCT coefficients by extending said plurality of DCT coefficients in two dimensions;
- and
- a filter for filtering said plurality of sampled DCT coefficients by pointwise multiplying said plurality of sampled DCT coefficients by a plurality of transform coefficients of a two-dimensional lowpass filter-right-half;
- and
- a second transforming means for performing an inverse discrete cosine transform (IDCT) on the sampled DCT coefficients to produce a resized image of the input image.
Dependencies are inferred from the claim text.