|
a single image from the sequence recorded under turbulence conditions |
an average image computed from the complete sequence |
|
Wiener restoration of a single image from the sequence |
Our result: Wiener restoration of the average image |
Proposal Abstract Publications Results (turbulence) Results (super-resolution) Back to my Homepage
Restoration and Enhancement of
Image Sequences
Boaz Cohen
Ph.D. Research (in progress)
The Ph.D. research proposed here deals with three aspects regarding image sequence restoration and enhancement. One aspect is super-resolution restoration of an image from an image sequence of a still scene. The second is restoration of an image from an image sequence of a still scene affected by turbulence. The third aspect deals with generation of an image with reduced speckle noise from an ultrasound image sequence. The new proposed super-resolution restoration algorithm is an extension of the algorithm presented by M. Irani and S. Peleg (CVGIP: Graphical Models and Image Processing, vol. 53, no. 3, pp. 231-9, 1991), by integrating polyphase filtering into the iterative phase of the super-resolution image restoration algorithm. A polyphase FIR Wiener filter approach will be examined, and the ‘Projective Transform’ model shall be used for the sub-pixel registration of the image sequence. The results of the resolution enhancements shall be evaluated by psychophysical tests. Turbulent conditions cause blur and local geometric distortions in images. The geometric distortions are caused by random displacements with a zero mean. The average image of a sequence of a still scene has more blur than the individual images in the sequence, but the geometrical distortions are averaged out, and thus the geometrical features are closer to the real scene. Preliminary results show that once the averaged image is restored, the result is far better than restoration results of the images in the sequence. Theoretical analysis of this approach will be performed for better understanding of the process and for proper implementation of the proposed technique. The possibility of applying a pre-processing stage and averaging only ‘high-quality’ images shall be examined. Ultrasound B-scan images play an important diagnostic role for physicians, but they have relatively poor quality. Most of the attempts to reduce the speckle noise in ultrasound images have been focused on single frame images. A number of attempts exploiting the temporal information have also been explored. Speckle reduction through local motion estimation and motion compensated order statistic filtering for ultrasound images will be examined. An attempt will be made to develop a motion estimation technique suited for the low SNR ultrasound image sequences.
Regarding restoration of an image with turbulence:
|
1. B. Cohen, V. Avrin, M. Belitsky, and I. Dinstein, "Generation of a restored image from a video sequence recorded under turbulence effects",Optical Engineering, vol. 36, no. 12, pp. 3312-3317, Dec. 1997. |
|
2. B. Cohen, V. Avrin, M. Belitsky, and I. Dinstein, "Restoration of an image representing a video sequence recorded under turbulence effects", in Proceedings of the SPIE, Applications of Digital Image Processing XX, San-Diego, CA, 30 July-1 Aug., 1997, vol. 3164, pp. 535-543. |
Regarding image resolution enhancement:
|
1. B. Cohen and I. Dinstein, "Resolution Enhancement by Polyphase Back-Projection Filtering", in Proceedings of the International Conference on Acoustics, Speech and Signal Processing, Seattle, Washington, May 12-15, 1998, vol. V, pp. 2921-2924. |
|
2. B. Cohen and I. Dinstein, "Image Resolution Enhancement by Polyphase FIR Wiener Filtering", in Proceedings of the SPIE, Applications of Digital Image Processing XXII, Denver, Colorado, July 18-23, 1999, vol. 3808, pp. 504-514. |
|
3. B. Cohen, V. Avrin and I. Dinstein, "Polyphase back-projection filtering for resolution enhancement of image sequences", to appear in Proceeding of the International Conference on Acoustics, Speech and Signal Processing, Istanbul, Turkey, June 5-9, 2000. |
|
4 . B. Cohen and I. Dinstein, "Polyphase back-projection filtering for image resolution enhancement", to appear in IEE Proceeding- Vision, Image and Signal Processing.(view abstract) (get a compressed postscript file) |
|
5. B. Cohen and I. Dinstein, "Polyphase FIR Wiener filtering for image resolution enhancement", submitted to IEEE Transactions on Image Processing.(view abstract) (get a compressed postscript file) |
Regarding ultrasound motion estimation:
|
1. B. Cohen and I. Dinstein, "Motion estimation in noisy ultrasound images by maximum likelihood", to appear in Proceedings of the 15th International Conference on Pattern Recognition, Barcelona, Spain, Sept. 3-8, 2000. |
|
2. B. Cohen and I. Dinstein, "New maximum likelihood motion estimation schemes for noisy ultrasound images", submitted to Pattern Recognition.(view abstract) (get a compressed postscript file) |
|
a single image from the sequence recorded under turbulence conditions |
an average image computed from the complete sequence |
|
Wiener restoration of a single image from the sequence |
Our result: Wiener restoration of the average image |
one of the low-resolution images
interpolation of the low-resolution image (by a factor of 4)
super resolution result according to the polyphase back-projection method
