Image analysis – Image transformation or preprocessing – Changing the image coordinates
Reexamination Certificate
1998-05-18
2001-02-13
Mehta, Bhavesh (Department: 2721)
Image analysis
Image transformation or preprocessing
Changing the image coordinates
C382S254000, C382S260000, C382S265000, C382S279000
Reexamination Certificate
active
06188804
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to image processing, and is more particularly concerned with the processing of sampled images that can have missing information to provide a full output image.
BACKGROUND OF THE INVENTION
There are many imaging and communications systems which employ an image which is not uniformly sampled, and from which a full image is reconstructed. The human visual system provides an example from nature wherein the spatial positions of the rods and cones provide random samples of the scene imaged on the retina, from which a full image is subsequently reconstructed. Another example of a system dealing with two dimensional signals or images is found in video and electronic still applications where charge coupled device (CCD) sensors are often used. CCDs may have defects or non-valid pixels, and thereby have incomplete information at some pixel locations. Additionally, in the case where a single sensor is used to capture multiple two-dimensional channels, for example the color channels of an image, then a color filter array (CFA) is often employed to discriminate between the color channels. The CFA acts as either a regular or random sampling device in each of the color channels and may, in addition have defects which results in additional incomplete information.
There are many situations in which random sampling of a signal has significant advantages over regular sampling, an excellent review of such advantages being provided by U.S. Pat. No. 4,574,311 entitled “Random Array Sensing Devices.” Highlighted in the above-mentioned application is the fact that aliasing can occur when a non-bandlimited signal is sampled. Aliasing is particularly noticeable (as low frequency periodic information) when a non-bandlimited high frequency periodic signal is sampled on a regular sampling grid. Aliasing may also occur when a non-bandlimited high frequency signal is randomly sampled, but in this case the artifact is manifest as broad band noise which is less objectionable.
Various approaches to solving the reconstruction problem have been presented. The above-mentioned U.S. Pat. No. 4,574,311 discloses a triangulation based linear interpolation method that utilizes the spatial distance between irregular samples to reconstruct a continuous image: This requires the calculation of multiple distances at each pixel location and limits the contribution to the reconstruction to the nearest neighbour samples.
In many reconstruction schemes, iterative methods are used, such as the Adaptive Weights Method, an iterative Fourier interpolation technique, described in Benedetto and Frazier, Wavelets: Mathematics and Applications, CRC Press, 1994, pp 336-359. A comprehensive review of approaches to reconstructing irregularly sampled images can be found in IEEE Transactions on Image Processing, Vol. 6, No. 4, April 1997, ‘Computationally Attractive Reconstruction of Bandlimited Images from Irregular Samples’ by Strohmer. Again, recursive or iterative methods are discussed and it is noted that in many cases it is difficult or impossible to extend the methods applicable to one-dimensional signals for use in the reconstruction of signals of two or more dimensions, such as images.
SUMMARY OF THE INVENTION
It is an object of the present invention to provide a full reconstruction from a sampled image which has missing information.
It is another object of the present invention to provide a method of obtaining an output reconstructed image from a sampled input image which overcomes the problems described above, and which avoids computationally expensive iterative or recursive solutions.
A still further object of the invention is to provide a method which is easily extendable to multiple color channel and multiple dimensional signals.
In accordance with one aspect of the present invention, there is provided a method of processing a sampled input image having pixels which can have missing information at pixel locations to reconstruct an output image therefrom, comprising the steps of:
a) providing a mask for the sampled image to identify valid pixel locations;
b) interpolating by convolution with at least one finite impulse response (FIR) filter(s) applied to the information at valid pixel locations to produce reconstructed information for non-valid pixel locations; and
(c) adaptively normalizing the reconstructed information for the non-valid pixel locations produced by applying the FIR filter(s) to the mask so that the valid pixels and the reconstructed information at non-valid pixel locations provide a reconstructed image.
ADVANTAGEOUS EFFECT OF THE INVENTION
By this method, any sampled signal, can be processed to reconstruct an interpolated output image or signal, which will generally be uniformly or regularly sampled. The method according to the invention is simpler, more robust and more easily implemented than those described in the prior art. In addition, the method is more easily and robustly extendable to two or more dimensions than the methods of the prior art, for use for example, where the signal is multi-dimensional, such as images or time-varying sequences of images.
REFERENCES:
patent: 4574311 (1986-03-01), Resnikoff et al.
patent: 4910694 (1990-03-01), Walther
patent: 5172227 (1992-12-01), Tsai et al.
patent: 5272536 (1993-12-01), Sudo et al.
patent: 5363318 (1994-11-01), McCauley
patent: 5812336 (1998-09-01), Spurbeck et al.
patent: 5960106 (1999-09-01), Tsuchiya et al.
Benedetto and Frazier, Wavelets: Mathematics and Applications, CRC Press, 1994, pp. 336-359.
IEEE Transactions on Image Processing, vol. 6, No. 4, Apr. 1997, ‘Computationally Attractive Reconstruction of Bandlimited Images from Irregular Samples’ by Strohmer.
Pratt, Digital Image Processing, John Wiley & Sons, 1978, pp. 111-120.
Image Processing, by D. Pearson, published by McGraw-Hill (1991), pp. 22 to 26.
Loveridge Jennifer C.
Weldy John A.
Eastman Kodak Company
Mehta Bhavesh
Owens Raymond L.
Patel Kanji
LandOfFree
Reconstructing missing pixel information to provide a full... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Reconstructing missing pixel information to provide a full..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Reconstructing missing pixel information to provide a full... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2612274