Post-filter for removing ringing artifacts of DCT coding

Claim 2

Apparatus for use in a video signal decoding system which digitally decodes data values that have been encoded using a quantized spatial frequency compression technique, the system including filtering apparatus which suppresses a range of noise after image reconstruction, the apparatus comprising: means for receiving decoded data values which describe a portion of an image frame; and anisotropic diffusion filtering means including: means for calculating a threshold value for the portion of the image frame; and an adaptive anisotropic filter which selectively combines respective image components, responsive to the threshold value, to produce diffused image components representing the noise-suppressed image.

Claim 3

Apparatus in claim 1 wherein the threshold value is a predetermined fraction of a maximum gradient value in the portion of the image frame.

Claim 4

Apparatus in claim 2 further comprising means for determining the maximum gradient value, said means comprising: comparator means for comparing the magnitude value of a current pixel to the respective magnitude values of four neighboring pixels to determine a largest magnitude value and a smallest magnitude value; and subtracter means for subtracting the smallest magnitude value from the largest magnitude value to determine the maximum gradient value.

Claim 5

Apparatus in claim 1 wherein the filtering is calculated according to the equation ##EQU2## wherein "I(t 1 )" is the filtered value, "λ" is a numerical stability condition, "g i " is a conductance for a current pixel, "ΔI i " is a magnitude difference between the current pixel and a neighboring pixel, and "I(t 0 )" is the magnitude of the current pixel.

Claim 6

Apparatus in claim 1 wherein the filter is implemented using a read-only memory (ROM) that is programmed according to the equation ##EQU3## wherein "I(t 1 )" is the filtered value, "λ" is a numerical stability condition, "g i " is a conductance for a current pixel, "ΔI i " is a magnitude difference between the current pixel and a neighboring pixel, and "I(t 0 )" is the magnitude of the current pixel.

Claim 7

Apparatus for use in a video signal decoding system which digitally decades data values that have been encoded using a quantized spatial frequency compression technique, the system including filtering apparatus which suppresses a range of noise after image reconstruction, the apparatus comprising: means for receiving decoded data values which describe a portion of an image frame; anisotropic diffusion filtering means for filtering the received data values to selectively suppress signal edge components having values less than a threshold value, wherein the threshold value is a predetermined fraction of a maximum gradient value in the portion of the image frame; and means for controlling the anisotropic diffusion filtering means by a conductance value determined from the maximum gradient value in the portion of the image frame.

Claim 8

Apparatus for use in a video signal decoding system which digitally decades data values that have been encoded using a quantized spatial frequency compression technique, the system including filtering apparatus which suppresses a range of noise after image reconstruction, the apparatus comprising: means for receiving decoded data values which describe a portion of an image frame; anisotropic diffusion filtering means for filtering the received data values to selectively suppress signal edge components having values less than a threshold value, wherein the threshold value is a predetermined fraction of a maximum gradient value in the portion of the image frame; and means for controlling the anisotropic diffusion filtering means by a conductance value determined from the maximum gradient value in the portion of the image frame.

Claim 9

Apparatus in claim 6 wherein the conductance value is determined by a clipped straight line approximation.

Claim 10

Apparatus in claim 7 wherein the conductance value is calculated by the equation g(gradient)=C.sub.1 + C.sub.2 /k!*gradient wherein "g" is the conductance value, "gradient" is the maximum gradient value, "k" is the threshold value, and "C 1 " and "C 2 " are constants.

Claim 11

Apparatus in claim 8 wherein C 1 equals 1.21 and C 2 equals -0.85576.

Claim 12

Apparatus for use in a video signal decoding system which digitally decodes data values that have been compressed using quantized spatial frequency components, the system including filtering apparatus which suppresses a range of noise after image reconstruction, the apparatus comprising: means for receiving decoded data values which describe a portion of an image frame; and anisotropic diffusion filtering means, including a first anisotropic diffusion filter and a second anisotropic diffusion filter, for filtering the received data values to selectively suppress signal edge components having values less than a threshold value, wherein the threshold value is a predetermined fraction of a maximum gradient value in the portion of the image frame means for determining the maximum gradient value, comprising: comparator means for comparing the magnitude value of a current pixel to the respective magnitude values of four neighboring pixels to determine a largest magnitude value and a smallest magnitude value; and subtractor means for subtracting the smallest magnitude value from the largest magnitude value to determine the maximum gradient value.

Claim 13

Apparatus for use in a video signal decoding system which digitally decodes data values that have been compressed using quantized spatial frequency components, the system including filtering apparatus which suppresses a range of noise after image reconstruction, the apparatus comprising: means for receiving decoded data values which describe a portion of an image frame; and anisotropic diffusion filtering means, including a first anisotropic diffusion filter and a second anisotropic diffusion filter, for filtering the received data values to selectively suppress signal edge components having values less than a threshold value, wherein the threshold value is a predetermined fraction of a maximum gradient value in the portion of the image frame means for determining the maximum gradient value, comprising: comparator means for comparing the magnitude value of a current pixel to the respective magnitude values of four neighboring pixels to determine a largest magnitude value and a smallest magnitude value; and subtractor means for subtracting the smallest magnitude value from the largest magnitude value to determine the maximum gradient value.

Claim 14

Apparatus in claim 10 in which the filtering means further includes: means for decreasing the threshold value by a factor of two for use by the second anisotropic diffusion filter.

Claim 15

In a video signal decoding system which digitally decodes data values that have been compressed using quantized spatial frequency components, a method for suppressing a range of noise after image reconstruction comprising the steps of: a) receiving decoded data values which describe a portion of an image frame; and b) anisotropic diffusion filtering the received data values to selectively suppress signal edge components including the steps of: calculating a threshold value for the portion of the image frame; and selectively diffusing the image components in the portion of the image frame responsive to the threshold value to produce diffused image components representing the noise-suppressed image.

Claim 16

In a video signal decoding system which digitally decodes data values that have been compressed using quantized spatial frequency components, a method for suppressing a range of noise after image reconstruction comprising the steps of: a) receiving decoded data values which describe a portion of an image frame; and b) anisotropic diffusion filtering the received data values to selectively suppress signal edge components including the steps of: calculating a threshold value for the portion of the image frame; and selectively diffusing the image components in the portion of the image frame responsive to the threshold value to produce diffused image components representing the noise-suppressed image.

Claim 17

The method of claim 12 wherein the threshold value is a predetermined fraction of a maximum gradient value in the portion of the image frame.

Claim 18

The method of claim 13 in which the maximum gradient value is determined by the steps of: comparing the magnitude value of a current pixel to the respective magnitude values of four neighboring pixels to determine a largest magnitude value and a smallest magnitude value; and subtracting the smallest magnitude value from the largest magnitude value to determine the maximum gradient value.

Claim 19

The method of claim 12 wherein the anisotropic diffusion filtering is calculated according to the equation ##EQU4## wherein "I(t 1 )" is the filtered value, "λ" is a numerical stability condition, "g i " is a conductance for a current pixel, "ΔI i " is a magnitude difference between the current pixel and a neighboring pixel, and "I(t 0 )" is the magnitude of the current pixel.

Claim 20

The method of claim 15 wherein the anisotropic diffusion filtering is performed using a ROM.

Claim 21

The method of claim 16 wherein the ROM is programmed according to the equation ##EQU5## wherein "I(t 1 )" is the filtered value, "λ" is a numerical stability condition, "g i " is a conductance for a current pixel, "ΔI i " is a magnitude difference between the current pixel and a neighboring pixel, and "I(t 0 )" is the magnitude of the current pixel.

Claim 22

In a video signal decoding system which digitally decodes data values that have been compressed using quantized spatial frequency components, a method for suppressing a range of noise after image reconstruction comprising the steps of: a) receiving decoded data values which describe a portion of an image frame; and b) anisotropic diffusion filtering the received data values to selectively suppress signal edge components having values less than a threshold value, wherein the threshold value is a predetermined fraction of a maximum gradient value in the portion of the image frame; and c) controlling the anisotropic diffusion filtering by a conductance value determined from the maximum gradient value in the portion of the image frame.

Claim 23

In a video signal decoding system which digitally decodes data values that have been compressed using quantized spatial frequency components, a method for suppressing a range of noise after image reconstruction comprising the steps of: a) receiving decoded data values which describe a portion of an image frame; and b) anisotropic diffusion filtering the received data values to selectively suppress signal edge components having values less than a threshold value, wherein the threshold value is a predetermined fraction of a maximum gradient value in the portion of the image frame; and c) controlling the anisotropic diffusion filtering by a conductance value determined from the maximum gradient value in the portion of the image frame.

Claim 24

The method of claim 18 wherein the conductance value is determined by a clipped straight line approximation.

Claim 25

The method of claim 19 wherein the conductance value is calculated by the equation g(gradient)=C.sub.1 + C.sub.2 /k!*gradient wherein "g" is the conductance value, "gradient" is the maximum gradient value, "k" is the threshold value, and "C 1 " and "C 2 " are constants.

Claim 26

The method of claim 20 wherein C 1 equals 1.21 and C 2 equals -0.85576.

Claim 27

In a video signal decoding system which digitally decodes data values that have been encoded using a quantized spatial frequency compression technique, a method for suppressing a range of noise after image reconstruction comprising the steps of: a) receiving decoded data values which describe a portion of an image frame; and b) anisotropic diffusion filtering the received data values to selectively suppress signal edge components having values less than a threshold value, including the steps of: filtering the received data values through a first anisotropic diffusion filter; and filtering the received data values through a second anisotropic diffusion filter.

Claim 28

In a video signal decoding system which digitally decodes data values that have been encoded using a quantized spatial frequency compression technique, a method for suppressing a range of noise after image reconstruction comprising the steps of: a) receiving decoded data values which describe a portion of an image frame; and b) anisotropic diffusion filtering the received data values to selectively suppress signal edge components having values less than a threshold value, including the steps of: filtering the received data values through a first anisotropic diffusion filter; and filtering the received data values through a second anisotropic diffusion filter.

Claim 29

The method of claim 22 in which step b) further includes the step of decreasing the threshold value by a factor of two for use by the second anisotropic diffusion filter.