IPC분류정보
국가/구분 |
United States(US) Patent
등록
|
국제특허분류(IPC7판) |
|
출원번호 |
US-0753548
(2007-05-24)
|
등록번호 |
US-8503811
(2013-08-06)
|
발명자
/ 주소 |
|
출원인 / 주소 |
|
대리인 / 주소 |
|
인용정보 |
피인용 횟수 :
0 인용 특허 :
3 |
초록
▼
A curvature-preserving filter having a null covariance matrix is applied to an input image to produce a denoised output image for output to a graphic display device or to a machine analysis tool. In one embodiment, the input image is a small kernel consisting of a limited number of pixels and the fi
A curvature-preserving filter having a null covariance matrix is applied to an input image to produce a denoised output image for output to a graphic display device or to a machine analysis tool. In one embodiment, the input image is a small kernel consisting of a limited number of pixels and the filter is applied to the input image by direct summation. In another embodiment, a digital image is input into an image processor that executes a Fourier transform to produce a Fourier-transformed signal. The curvature-preserving filter is applied to the Fourier-transformed signal in Fourier space to produce a denoised signal, then the denoised signal is transformed by an inverse Fourier transform to generate a denoised output image In an alternate embodiment, the filter further produces a deblurred signal by including an inverse point-response function.
대표청구항
▼
1. A method for reconstructing an image from an image source, comprising: inputting a digital image signal comprising a plurality of pixels into an image processing device programmed for executing a curvature-preserving filter, wherein the filter comprises at least zeroth, first and second moments,
1. A method for reconstructing an image from an image source, comprising: inputting a digital image signal comprising a plurality of pixels into an image processing device programmed for executing a curvature-preserving filter, wherein the filter comprises at least zeroth, first and second moments, the second moments comprising a null covariance matrix;filtering the image signal by the filter to produce a denoised image; andoutputting the denoised image to a graphic display device or a machine analysis tool,wherein the image processing device comprises a central processing unit, a microprocessor or a digital signal processor having software stored therein for, prior to the step of multiplying, executing a Fourier transform to produce a Fourier-transformed signal comprising vector wave numbers in Fourier space, and after multiplying the image signal by the filter, executing an inverse Fourier transform to produce the denoised image,wherein the filter has a Fourier representation comprising g~(k)/p~(k)=exp(12σ2k2-a4k4), where a is a scale length that determines a width of the filter, k is the vector wave number, k is the magnitude of k, and σ is a standard deviation of the point-response function. 2. The method of claim 1, wherein the filter kernel further comprises an inverse point-response function for deblurring. 3. The method of claim 1, wherein the image processing device comprises one or more field-programmable gate arrays or application specific integrated circuits configured to receive a kernel consisting of a small number of pixels, and the filter kernel is applied to the image signal by direct summation. 4. The method of claim 3, wherein the image signal comprises a raster data stream. 5. The method of claim 1, wherein the image signal comprises a full data frame. 6. The method of claim 1, wherein the filter kernel is a function of an even power of a magnitude of the vector wave numbers with power greater or equal to four. 7. The method of claim 1, wherein the filter kernel has a Fourier representation comprising {tilde over (g)}(k)=exp(−a4k4), where a is a scale length that determines a width of the filter kernel, k is the vector wave number, and k is a magnitude of k. 8. An image processing device for denoising an input image, the device comprising: an input device for receiving the input image and outputting a digital image signal;a processor for receiving the image signal, the processor programmed to apply a curvature-preserving filter to the image signal to generate a denoised signal, wherein the filter comprises at least zeroth, first and second moments, the second moments comprising a null covariance matrix;an output device for outputting the denoised signal to a graphic display device or a machine analysis tool,wherein the image processing device comprises a central processing unit, a microprocessor or a digital signal processor having software stored therein for, prior to the step of multiplying, executing a Fourier transform to produce a Fourier-transformed signal comprising vector wave numbers in Fourier space, and after multiplying the image signal by the filter, executing an inverse Fourier transform to produce the denoised image,wherein the filter has a Fourier representation comprising g~(k)/p~(k)=exp(12σ2k2-a4k4), where a is a scale length that determines a width of the filter, k is the vector wave number, k is the magnitude of k, and σ is a standard deviation of the point-response function. 9. The image processing device of claim 8, wherein the filter kernel further comprises an inverse point-response function for deblurring. 10. The image processing device of claim 8, wherein the image processing device comprises one or more field-programmable gate arrays or application specific integrated circuits configured to receive a kernel consisting of a small number of pixels, and the filter kernel is applied to the image signal by direct summation. 11. The image processing device of claim 10, wherein the image signal comprises a raster data stream. 12. The image processing device of claim 8, wherein the image signal comprises a full data frame. 13. The image processing device of claim 8, wherein the filter kernel is a function of an even power of a magnitude of the vector wave numbers with power greater or equal to four. 14. The image processing device of claim 8, wherein the filter kernel has a Fourier representation comprising {tilde over (g)}(k)=exp(−a4k4), where a is a scale length that determines a width of the filter kernel, k is the vector wave number, and k is a magnitude of k.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.