Scaled forward and inverse discrete cosine transform and video compression/decompression systems employing the same

• US 6,694,342 B1
• Filed: 10/16/2000
• Issued: 02/17/2004
• Est. Priority Date: 06/13/1997
• Status: Expired due to Fees

• Current Assignee: Sun Microsystems Incorporated (Oracle Corporation)
• Original Assignee: Sun Microsystems Incorporated (Oracle Corporation)
• Inventors: Mou, Alex Z.
• Primary Examiner(s): Malzahn, David H.
• Application Number:
  US09/690,351
• Time in Patent Office: 1,219 Days
• Field of Search: 708/402, 382/250
• US Class Current: 708/402
• CPC Class Codes: G06F 17/147 Discrete orthonormal transforms, e.g. discrete cosine transform, discrete sine transform, and variations therefrom, e.g. modified discrete cosine transform, integer transforms approximating the discrete cosine transform G06F17/145 takes precedence
• CPC Class Codes: G06T 9/007 Transform coding, e.g. discrete cosine transform

Abstract

A system and method of a forward and/or inverse discrete cosine transform in a video system. In one embodiment, an array of DCT transform coefficients are converted to a two dimensional array of spatial data. The array of DCT transform coefficients are first operated upon by a pre-scale computation unit (implemented in either hardware or software) which multiplies a set of predetermined pre-scale constants with the input coefficients. The pre-scale constants multiplied by the input DCT coefficient matrix form a symmetric pre-scale array. Upon pre-scaling using the symmetric pre-scale factor array, an intermediary array is composed by performing intermediary calculations upon each column vector of the pre-scaled array. The output of this intermediary calculation is composed to form an intermediary array. Subsequently, a set of calculations are performed row-wise upon each row vector of the intermediary array to thereby form the output array of spatial data. In one implementation, the array of pre-scale coefficients (i.e., the symetric pre-scale factor array) may be represented as the result of a matrix multiplication M×U×M, where the array U consists of a set of coefficients all equal to 1, and wherein the array M is an array of coefficients wherein the equal row-column diagonal of coefficients include a set of pre-scale constants and all other coefficients in the array are equal to 0. The pre-scale constants includes a set of constants expressed by the cos(π/16), where n=1, 2, 3 and 4. In another embodiment, the calculations are reversed to perform a forward discrete cosine transform operation.

Claims

1. A method for performing a forward discrete cosine transform (DCT) upon a matrix of image data in a video system, the method comprising:

performing a column-wise calculation upon each column of said matrix of image data to obtain an intermediary array;

performing a row-wise calculation upon each row of said intermediary array to obtain an unscaled DCT array; and

multiplying said unscaled DCT array element-wise with corresponding elements of a symmetric post-scale factor array to produce an output DCT array which represents the image data in a frequency domain;

wherein said symmetric post-scale factor array is an array expressible by the product M×

U×

M, wherein U is a matrix of coefficients each being equal to 1, wherein M is an array having a set of post-scale constants along a diagonal and all zeros off the diagonal, wherein said set of post-scale constants are taken from the collection of values defined by cos(nπ

/16), where n=1, 2, 3, 4.

2. The method for performing a forward discrete cosine transform as recited in claim 1 further comprising retrieving each element of said matrix of data from a memory.

3. A method for performing a forward discrete cosine transform (DCT) upon a matrix of image data in a video system, the method comprising:

performing a row-wise calculation upon each row of said matrix of image data to obtain an intermediary array;

performing a column-wise calculation upon each column of said intermediary array to obtain an unscaled DCT array; and

multiplying said unscaled DCT array element-wise with corresponding elements of a symmetric post-scale factor array to produce an output DCT array which represents the image data in a frequency domain;

wherein said symmetric post-scale factor array is an array expressible by the product M×

U×

M, wherein U is a matrix of coefficients each being equal to 1, wherein M is an array having a set of post-scale constants along a diagonal and all zeros off the diagonal, wherein said set of post-scale constants are taken from the collection of values defined by cos(nπ

/16), where n=1, 2, 3, 4.

4. The method for performing a forward discrete cosine transform as recited in claim 3 further comprising retrieving each element of said matrix of data from a memory.

5. A discrete cosine transform calculation unit for a video system, configured to perform a discrete cosine transform (DCT) upon a matrix of image data, said discrete cosine transform calculation unit comprising:

a column vector computation unit configured to perform a column-wise calculation upon each column of said matrix of image data to obtain an intermediary array;

a row vector computation unit configured to perform a row-wise calculation upon each row of said intermediary array to obtain an unscaled DCT array; and

a postscale computation unit configured to multiply said unscaled DCT array element-wise with corresponding elements of a symmetric post-scale factor array to produce an output DCT array which represents the image data in a frequency domain;

wherein said symmetric post-scale factor array is an array expressible by the product M×

U×

M, wherein U is a matrix of coefficients each being equal to 1, wherein M is an array having a set of post-scale constants along a diagonal and all zeros off the diagonal, wherein said set of post-scale constants are taken from the collection of values defined by cos(nπ

/16), where n=1, 2, 3, 4.

6. A discrete cosine transform calculation unit for a video system, configured to perform a discrete cosine transform (DCT) upon a matrix of image data, said discrete cosine transform calculation unit comprising:

a row vector computation unit configured to perform a row-wise calculation upon each row of said matrix of image data to obtain an intermediary array;

a column vector computation unit configured to perform a column-wise calculation upon each column of said intermediary array to obtain an unscaled DCT array; and

a postscale computation unit configured to multiply said unscaled DCT array element-wise with corresponding elements of a symmetric post-scale factor array to produce an output DCT array which represents the image data in a frequency domain;

wherein said symmetric post-scale factor array is an array expressible by the product M×

U×

M, wherein U is a matrix of coefficients each being equal to 1, wherein M is an array having a set of post-scale constants along a diagonal and all zeros off the diagonal, wherein said set of post-scale constants are taken from the collection of values defined by cos(nπ

/16), where n=1, 2, 3, 4.