Adaptive PSF estimation technique using a sharp preview and a blurred image
View Patent ↗An adaptive motion estimation and deblurring technique for acquired digital images includes acquiring multiple digital images with a moving digital image acquisition device that includes an image sensor, including a relatively sharp, underexposed reference image and a blurred image. Anb initial approximate point spread function (PSF) is estimated corresponding to the moving of the device. A different DC offset point is determined and a second PSF is calculated based on the different DC offset point.
1. A method of adaptive motion estimation and deblurring of acquired digital images, comprising:
acquiring multiple digital images with a moving digital image acquisition device that includes an image sensor, including a relatively sharp, underexposed reference image and a blurred image;
estimating an initial approximate point spread function (PSF) based on a first DC offset point corresponding to the moving of the device during the acquiring of the multiple digital images;
generating a second DC offset point that is different from the first DC offset point and calculating a second PSF based on the second DC offset point;
calculating a Cepstrum and refining the second PSF to generate a third PSF based on the Cepstrum;
precisely determining a position of a first PSF point; and
deblurring at least one of the digital images using a best found PSF based on the third PSF and first PSF point.
2. The method of claim 1 , wherein the estimating comprises using a pair of gyro sensors or accelerometers, or a combination thereof.
3. The method of claim 1 , wherein the offset comprises a measure of a yaw or a pitch of the image sensor, or both.
4. The method of claim 1 , wherein calculating said Cepstrum comprises measuring a linearity of the PSF.
5. The method of claim 1 , wherein calculating said Cepstrum comprises computing a Radon transform.
6. The method of claim 1 , further comprising changing a PSF fit that does not pass a distance test from the initial approximate PSF.
7. One or more non-transitory computer readable media having code embedded therein for programming one or more processors to perform a method of adaptive motion estimation and deblurring of acquired digital images, wherein the method comprises:
acquiring multiple digital images with a moving digital image acquisition device that includes an image sensor, including a relatively sharp, underexposed reference image and a blurred image;
estimating an initial approximate point spread function (PSF) based on a first DC offset point corresponding to the moving of the device during the acquiring of the multiple digital images;
generating a second DC offset point that is different from the first DC offset point and calculating a second PSF based on the second DC offset point;
calculating a Cepstrum and refining the second PSF to generate a third PSF based on the Cepstrum;
precisely determining a position of a first PSF point; and
deblurring at least one of the digital images using a best found PSF based on the third PSF and first PSF point.
8. The one or more computer readable media of claim 7 , wherein the estimating comprises using a pair of gyro sensors or accelerometers, or a combination thereof.
9. The one or more computer readable media of claim 7 , wherein the offset comprises a measure of a yaw or a pitch of the image sensor, or both.
10. The one or more computer readable media of claim 7 , wherein calculating said Cepstrum comprises measuring a linearity of the PSF.
11. The one or more computer readable media of claim 7 , wherein calculating said Cepstrum comprises computing a Radon transform.
12. The one or more computer readable media of claim 7 , further comprising changing a PSF fit that does not pass a distance test from the initial approximate PSF.
13. A digital image acquisition device, comprising:
a lens and an image sensor for acquiring a digital image;
a processor; and
one or more processor-readable media having code embedded therein for programming the processor to perform a method of adaptive motion estimation and deblurring of acquired digital images, wherein the method comprises:
acquiring multiple digital images with a moving digital image acquisition device that includes an image sensor, including a relatively sharp, underexposed reference image and a blurred image;
estimating an initial approximate point spread function (PSF) based on a first DC offset point corresponding to the moving of the device during the acquiring of the multiple digital images;
generating a second DC offset point that is different from the first DC offset point and calculating a second PSF based on the second DC offset point;
calculating a Cepstrum and refining the second PSF to generate a third PSF based on the Cepstrum;
precisely determining a position of a first PSF point; and
deblurring at least one of the digital images using a best found PSF based on the third PSF and first PSF point.
14. The device of claim 13 , wherein the estimating comprises using a pair of gyro sensors or accelerometers, or a combination thereof.
15. The device of claim 13 , wherein the offset comprises a measure of a yaw or a pitch of the image sensor, or both.
16. The device of claim 13 , wherein calculating said Cepstrum comprises measuring a linearity of the PSF.
17. The device of claim 13 , wherein calculating said Cepstrum comprises computing a Radon transform.
18. The device of claim 13 , further comprising changing a PSF fit that does not pass a distance test from the initial approximate PSF.