Circuit for convolving a set of digital data

A digital signal filter circuit for convolving a two-dimensional array of at least n by n data samples with a kernel of n coefficients where the coefficients have as an envelope a polynomial function of degree g, comprising;

• a source of weighting values that are functions of the coefficients of said polynomial;
• first means for extracting from the array of data samples a first set of q+1 contiguous data samples in a first dimension;
• first means for storing n-(q+1) intermediate first dimension data samples following said set of q+1 contiguous data samples;
• second means for extracting from the array of data samples a second set of q+1 contiguous data samples in a first dimension that follow said intermediate first dimension data samples;
• first means for linearly combining said first dimension data samples of said first and second sets with the weighting values to obtain a linear combination of (q+2) sequential convolved first dimension samples with binomial coefficients;
• first means for recursively selecting said first dimension convolved samples from said linear combination of (q+2) sequential convolved samples to obtain a first dimensional convolvement;
• means for storing said first dimensional convolvement;
• third means, responsive to said means for storing said first dimensional convolvement, for extracting from said first dimensional convolvement a first set of q+1 contiguous data samples in a second dimension;
• second means for storing n-(q+1) intermediate second dimension data samples following said set of q+1 contiguous data samples in a second dimension;
• fourth means for extracting from said first dimensional convolvement a second set of q+1 contiguous data samples in a second dimension that follow said intermediate second dimension data samples;
• second means for linearly combining said second dimension data samples of said first and second sets with the weighting values to obtain a linear combination of (q+2) sequential convolved second dimension samples with binomial coefficients;
• second means for recursively selecting said second dimension convolved samples from said linear combination of (q+2) sequential convolved samples to obtain a second convolvement of the first dimensional convolvement;
• and
• means, responsive to said second means for recursively selecting, for storing said second dimensional convolvement to obtain a two-dimensional convolvement of said array.

A digital signal filter circuit for convolving a two-dimensional array of at least n by n data samples with a kernel of n coefficients where the coefficients have as an envelope a polynomial function of degree g, comprising;

• a source of weighting values that are functions of the coefficients of said polynomial;
• first means for extracting from the array of data samples a first set of q+1 contiguous data samples in a first dimension;
• first means for storing n-(q+1) intermediate first dimension data samples following said set of q+1 contiguous data samples;
• second means for extracting from the array of data samples a second set of q+1 contiguous data samples in a first dimension that follow said intermediate first dimension data samples;
• first means for linearly combining said first dimension data samples of said first and second sets with the weighting values to obtain a linear combination of (q+2) sequential convolved first dimension samples with binomial coefficients;
• first means for recursively selecting said first dimension convolved samples from said linear combination of (q+2) sequential convolved samples to obtain a first dimensional convolvement;
• means, responsive to each first means for recursively selecting;
• for storing n+2q+2 series of convolved first dimension data samples;
• second means for linearly combining at least n+(q+1) data samples in a second dimension of said n+2q+2 series of convolved first dimension data samples with the weighting values to obtain a linear combination of q+2 sequential convolved second dimensional samples with binomial coefficients;
• and
• second means, responsive to said second means for linearly combining, for recursively selecting said second dimension convolved samples from said linear combination of (q+2) convolved samples to obtain a two-dimensional convolvement of said array.