초록 ▼

A ringing area detector classifies the input image into two regions: a mosquito noise region (i.e. filtering region) and a non-mosquito noise region (i.e. non-filtering region), and uses this classification information to adaptively remove the mosquito noise in a mosquito noise reduction system. The

A ringing area detector classifies the input image into two regions: a mosquito noise region (i.e. filtering region) and a non-mosquito noise region (i.e. non-filtering region), and uses this classification information to adaptively remove the mosquito noise in a mosquito noise reduction system. The mosquito noise reduction system includes a ringing area detector, a local noise power estimator, a smoothing filter, and a mixer. The ringing area detector includes an edge detector, a near edge detector, a texture detector, and a filtering region decision block. The ringing detection block detects the ringing area where the smoothing filter is to be applied. The local noise power estimator controls the filter strength of the smoothing filter. The smoothing filter smoothes the input image. The mixer mixes the smoothed image and the original image properly based on the region information from the ringing area detection block.

대표청구항 ▼

What is claimed is: 1. An apparatus for detecting ringing area in a decoded video sequence, comprising: a gradient computation unit that calculates a gradient of the video image; an edge detector that detects an edge region in the video image and generates an edge map, the edge detector using a mag

What is claimed is: 1. An apparatus for detecting ringing area in a decoded video sequence, comprising: a gradient computation unit that calculates a gradient of the video image; an edge detector that detects an edge region in the video image and generates an edge map, the edge detector using a magnitude of the gradient to determine if a pixel is an edge pixel; a near edge detector that uses the edge map to detect near-edge region and generates a near-edge map; a texture detector that detects texture in the video image and generates a texture map; and a filtering region decision unit that uses the near-edge map and the texture map to generate a filtering region map. 2. The apparatus of claim 1, wherein the edge detection unit detects the edge region by comparing the magnitude of the gradient of each pixel in the video image with a threshold value, and marks the pixel as an edge pixel if its gradient is greater than the threshold value. 3. The apparatus of claim 1, wherein the near edge detector detects the near-edge region by counting the edge pixels within a block in the video image, wherein the block contains the current pixel, and marks the current pixel as a near-edge pixel if the number of edge pixels within the block exceeds certain threshold value. 4. The apparatus of claim 3, wherein said block used for near-edge detection overlaps other blocks in the video image. 5. The apparatus of claim 3, wherein said block used for near-edge detection is non-overlapping with other blocks in the video image. 6. The apparatus of claim 1, wherein the texture detection unit detects the texture region by comparing the magnitude of the gradient of each pixel in the video image with a threshold value, and marks the pixel as a texture pixel if its gradient is greater than the threshold value. 7. The apparatus of claim 1, wherein said filtering region decision unit generates the filtering region map by excluding the texture region from the near-edge region. 8. A noise reduction system for reducing mosquito noise in a decoded video sequence, comprising: a ringing area detection unit apparatus as claimed in claim 1 that generates a filtering region map for the video image; a local noise power estimator that estimates an equivalent local additive Gaussian noise power from the video image; a smoothing filter that removes unwanted noise from the video image based on said estimate; and a mixer that selects either the result of the smoothing filter or the original input video image as output based on the filtering region map. 9. The system of claim 8, wherein said local noise power estimator comprises: a high-pass filter that extracts the noise component of the video image signal F(i,j); a local standard deviation calculator that calculates the local standard deviation σh of the high-pass filtered signal; and a converter that converts the local standard deviation σh of the high-pass filtered signal into a equivalent local additive Gaussian noise power σn as said estimate. 10. The system of claim 9, wherein said local standard deviation calculator calculates the local standard deviation σh of the high-pass filtered signal HF(i,j) over a r×s window in the video image as: σ h ⁡ ( i , j ) = 1 r × s ⁢ ∑ n = - r r ⁢ ∑ m = - s s ⁢  HF ⁡ ( i + m , j + n ) - μ h ⁡ ( i , j )  wherein μ h ⁡ ( i , j ) = 1 r × s ⁢ ∑ n = - r r ⁢ ∑ m = - s s ⁢ HF ⁡ ( i + m , j + n ) . 11. The system of claim 9, wherein said converter converts the local standard deviation σh of the high-pass filtered signal into a equivalent local additive Gaussian noise power σn as: σn=Bσh0.45, wherein B is global filter strength parameter which controls the overall smoothing level of the smoothing filter. 12. The system of claim 8, wherein said smoothing filter comprises a weighted sigma filter that filters the input signal as: F NR ⁡ ( i , j ) = 1 N ⁢ ∑ n = - r r ⁢ ∑ m = - r r ⁢ w m , n · F ⁡ ( i + m , j + n ) , ⁢ wherein F NR ⁡ ( i , j ) ⁢ ⁢ is ⁢ ⁢ the ⁢ ⁢ filtered ⁢ ⁢ signal ; w m , n = { 2 ⁢ if ⁢ ⁢  F ⁡ ( i + m , j + n ) - F ⁡ ( i , j )  < C 1 ⁢ σ n , 1 ⁢ if ⁢ ⁢  F ⁡ ( i + m , j + n ) - F ⁡ ( i , j )  < C 2 ⁢ σ n , 0 ⁢ otherwise ; ⁢ ⁢ N = ∑ n = - r r ⁢ ∑ m = - r r ⁢ w m , n , and C1, C2 are predetermined constants with 0<C1<C2, and σn is the local noise power. 13. The system of claim 8, wherein said noise smoothing filter comprises a minimal mean square error filter that filters the signal as: F NR ⁡ ( i , j ) = μ ⁡ ( i , j ) + max ⁡ ( σ 2 ⁡ ( i , j ) - σ n 2 , 0 ) max ⁡ ( σ 2 ⁡ ( i , j ) - σ n 2 , 0 ) + σ n 2 · [ F ⁡ ( i , j ) - μ ⁡ ( i , j ) ] , ⁢ wherein F NR ⁡ ( i , j ) ⁢ ⁢ is ⁢ ⁢ the ⁢ ⁢ filtered ⁢ ⁢ signal ; μ ⁡ ( i , j ) = 1 r 2 ⁢ ∑ n = - r r ⁢ ∑ m = - r r ⁢ F ⁡ ( i + m , j + n ) , ⁢ σ ⁡ ( i , j ) = 1 r 2 ⁢ ∑ n = - r r ⁢ ∑ m = - r r ⁢  F ⁡ ( i + m , j + n ) - μ ⁡ ( i , j )  , and σn is the local noise power. 14. The system of claim 8, wherein said mixer selects the result of the smoothing filter as the output in the filtering region, and selects the original input signal as the output in the non-filtering region. 15. A noise reduction system for reducing mosquito noise in decoded video image sequence, comprising: a ringing area detection apparatus as claimed in claim 1 that generates a filtering region map for a video image; a low-pass filter that smoothes the filtering region map to generate a smooth transition between a filtering region and a non-filtering region; a local noise power estimator that estimates the equivalent local additive Gaussian noise power from the video image; a smoothing filter that removes unwanted noise from the video image; and a soft switching unit that gradually switches the smoothing filter on and off based on the smoothed filtering region map. 16. The system of claim 15, wherein said soft switching unit determines the output FOUT(i,j) of the system as: FOUT(i,j)={tilde over (R)}(i,j)·FNR(i,j)+(1−{tilde over (R)}(i,j))·F(i,j), wherein {tilde over (R)}(i,j) is the smoothed filtering region map, FNR(i,j) is the output of the smoothing filter, and F(i,j) is the original input video image. 17. An apparatus for detecting a ringing area in a decoded video sequence, comprising: a gradient computation unit that calculates a gradient of the video image; an edge detector that detects an edge region in the video image and generates an edge-map, the edge detector using a magnitude of the gradient to determine if a pixel is an edge pixel; a near edge detector that uses the edge map to detect near-edge region and generates a near-edge map; a texture detector that detects texture in the video image and generates a texture map; and a filtering region decision unit that generates a filtering region map based on the near-edge map, the texture map, and the edge map. 18. The apparatus of claim 17, wherein said filtering region decision unit generates the filtering region map by excluding the texture region from the near-edge region but keeping the edge region in the near-edge region. 19. An apparatus for detecting ringing area in a decoded video sequence, comprising: a gradient computation unit that calculates a gradient of the video image; an edge detector that detects an edge region in the video image and generates an edge-map, the edge detector using a magnitude of the gradient to determine if a pixel is an edge pixel; a first binary filter for filtering the edge-map; a near edge detector that uses the edge map to detect near-edge region and generates a near-edge map; a texture detector that detects texture in the video image and generates a texture map; and a second binary filter for filtering the texture map; and a filtering region decision unit that generates a filtering region map based on the filtered near-edge map and the filtered texture map. 20. The apparatus as claimed in claim 19, wherein the first binary filter comprises a mathematical morphology closing operator, and the second binary filter comprises another mathematical morphology closing operator. 21. An apparatus for detecting ringing area in a sequence of decoded video frames, comprising: a texture detector that detects texture in a current video frame and generates a texture map; a gradient computation unit that calculates a gradient of the video image; an edge detector that detects an edge region in the current video frame and generates an edge map, the edge detector using a magnitude of the gradient to determine if a pixel is an edge pixel; a near edge detector that uses the edge map to detect near-edge region in the current video frame and generates a near-edge map; a memory unit that stores the near-edge map of the previous video frame; and a filtering region decision unit that generates a filtering region map based on the delayed near-edge map of a previous frame and the texture map of the current frame. 22. A method of detecting a ringing area in a video image indicating mosquito noise therein, comprising: employing a noise reduction processor for: performing a gradient computation to determine the magnitude of a gradient in the video image; classifying the input image into a mosquito noise filtering region and a non-mosquito noise region using the gradient computation; and using the classification information to adaptively remove the mosquito noise from the video image. 23. The method of claim 22, wherein the step of classifying further includes the steps of detecting a ringing area in the video image by: detecting edge regions in the video image and generating an edge map; detecting near-edge regions in the video image using the edge map, and generating a near-edge map; detecting texture in the video image and generating a texture map; and generating a filtering region map representing a mosquito noise filtering region, based on the near-edge map and the texture map. 24. The method of claim 23, before the step of detecting an edge region, such that the step of edge detection further includes comparing the gradient of each pixel in the video image with a threshold value and marking the pixel as an edge pixel if its gradient is greater than the threshold value. 25. The method of claim 23 wherein the steps of detecting near-edge regions further includes the steps of counting the edge pixels within a block in the video image, wherein the block contains the current pixel, and marking the current pixel as a near-edge pixel if the number of edge pixels within the block exceeds certain threshold value. 26. The method of claim 25 wherein said block overlaps other blocks used for near-edge detection in the video image. 27. The method of claim 25 wherein said block is non-overlapping with other blocks used for near-edge detection in the video image. 28. The method of claim 23, before the step of detecting texture, wherein the step of detecting texture further includes the steps of detecting a texture region by comparing the gradient of each pixel in the video image with a threshold value, and marking the pixel as a texture pixel if its gradient is greater than the threshold value. 29. The method of claim 23 wherein the step of generating a filtering region map further includes the steps of excluding the texture region from the near-edge region. 30. A method of reducing mosquito noise in a digital video image, comprising: employing a noise reduction processor for: performing a gradient computation to determine the magnitude of a gradient in the video image; classifying the input image into a mosquito noise filtering region and a non-mosquito noise region using the gradient computation; using the classification information to adaptively remove the mosquito noise from the video image; generating a filtering region map for the video image; determining a local noise power estimator that estimates an equivalent local additive Gaussian noise power from the video image; using a smoothing filter to remove noise from the video image based on said estimate; and selects either the result of the smoothing filter or the original input video image as output based on the filtering region map. 31. The method of claim 30, wherein the step of determining the local noise power estimate further includes the steps of: using a high-pass filter to extract the noise component of the video image signal pixels F(i,j); calculating a local standard deviation σh of the high-pass filtered signal; and converting the local standard deviation σh of the high-pass filtered signal into a equivalent local additive Gaussian noise power σn as said estimate. 32. The method of claim 31, wherein the step of local standard deviation calculation further includes the steps of calculating the local standard deviation σh of the high-pass filtered signal HF(i,j) over a r×s window in the video image as: σ h ⁡ ( i , j ) = 1 r × s ⁢ ∑ n = - r r ⁢ ∑ m = - s s ⁢  HF ⁡ ( i + m , j + n ) - μ h ⁡ ( i , j )  wherein μ h ⁡ ( i , j ) = 1 r × s ⁢ ∑ n = - r r ⁢ ∑ m = - s s ⁢ HF ⁡ ( i + m , j + n ) . 33. The method of claim 31, wherein the step of converting further includes the steps of converting the local standard deviation σh of the high-pass filtered signal into a equivalent local additive Gaussian noise power σn as: σn=Bσh0.45, wherein B is global filter strength parameter which controls the overall smoothing level of the smoothing filter. 34. The method of claim 30, wherein said smoothing filter comprises a weighted sigma filter that filters the input signal as: F NR ⁡ ( i , j ) = 1 N ⁢ ∑ n = - r r ⁢ ∑ m = - r r ⁢ w m , n · F ⁢ ( i + m , j + n ) , ⁢ wherein F NR ⁡ ( i , j ) ⁢ ⁢ is ⁢ ⁢ the ⁢ ⁢ filtered ⁢ ⁢ signal ; w m , n = { 2 ⁢ if ⁢ ⁢  F ⁡ ( i + m , j + n ) - F ⁡ ( i , j )  < C 1 ⁢ σ n , 1 ⁢ if ⁢ ⁢  F ⁡ ( i + m , j + n ) - F ⁡ ( i , j )  < C 2 ⁢ σ n , 0 ⁢ otherwise ; ⁢ ⁢ N = ∑ n = - r r ⁢ ∑ m = - r r ⁢ w m , n , and C1, C2 are predetermined constants with 0<C1<C2, and σn is the local noise power. 35. The method of claim 30, wherein said smoothing filter comprises a minimal mean square error filter that filters the signal as: F NR ⁡ ( i , j ) = μ ⁡ ( i , j ) + max ⁡ ( σ 2 ⁡ ( i , j ) - σ n 2 , 0 ) max ⁡ ( σ 2 ⁡ ( i , j ) - σ n 2 , 0 ) + σ n 2 · [ F ⁡ ( i , j ) - μ ⁡ ( i , j ) ] , ⁢ wherein F NR ⁡ ( i , j ) ⁢ ⁢ is ⁢ ⁢ the ⁢ ⁢ filtered ⁢ ⁢ signal ; μ ⁡ ( i , j ) = 1 r 2 ⁢ ∑ n = - r r ⁢ ∑ m = - r r ⁢ F ⁢ ( i + m , j + n ) , ⁢ σ ⁢ ( i , j ) = 1 r 2 ⁢ ∑ n = - r r ⁢ ∑ m = - r r ⁢  F ⁡ ( i + m , j + n ) - μ ⁡ ( i , j )  , and σn is the local noise power. 36. The method of claim 30, wherein the step of selecting further includes the steps of selecting the result of the smoothing filter as the output in the filtering region, and selecting the original input signal as the output in the non-filtering region. 37. A method of reducing mosquito noise in decoded video image sequence, comprising: employing a noise reduction processor for: performing a gradient computation to determine the magnitude of a gradient in the video image; classifying the input image into a mosquito noise filtering region and a non-mosquito noise region using the gradient computation; using the classification information to adaptively remove the mosquito noise from the video image; generating a filtering region map for a video image; using a low-pass filter to smooth the filtering region map and generate a smooth transition between a filtering region and a non-filtering region; estimating a local noise power that estimates the equivalent local additive Gaussian noise power from the video image; using a smoothing filter to remove unwanted noise from the video image; and gradually soft switching the smoothing filter on and off based on the smoothed filtering region map to generate an output. 38. The method of claim 37, wherein the step of soft switching further includes determining the output FOUT(i,j) as: FOUT(i,j)={tilde over (R)}(i,j)·FNR(i,j)+(1−{tilde over (R)}(i,j))·F(i,j), wherein {tilde over (R)}(i,j) is the smoothed filtering region map, FNR(i,j) is the output of the smoothing filter, and F(i,j) is the original input video image.