Image analysis – Color image processing – Color correction
Reexamination Certificate
1999-09-27
2003-04-22
Patel, Jayanti K. (Department: 2625)
Image analysis
Color image processing
Color correction
C382S173000, C382S248000
Reexamination Certificate
active
06553139
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates generally to a method for processing of pixels of an image segment by a computer, and in particular for signal extrapolation of brightness values and/or color values of pixels of a first image segment onto pixels of a second image segment, and to an apparatus for practicing the method.
2. Description of the Related Art
The coding of video signals in accordance with, for example, the picture coding standards H.261, H.263, MPEG1 and MPEG2 is frequently based on a block-oriented discrete cosine transform (DCT). However, these block-oriented image coding methods are not suitable for image coding methods which are not based on rectangular blocks but rather in which, for example, objects are segmented from an image and the image segments are coded. These methods are referred to as region-based or object-based image coding methods. In this case, digital images are segmented in a manner corresponding to the objects appearing in the scene. Separate coding of these segmented objects is carried out instead of the coding of picture blocks as in the case of block-based image coding methods. In this case, the coding is usually effected by modeling of the segmented objects and subsequent transmission of the modeling parameters of these segmented objects.
After the image information has been transmitted from a transmitter to a receiver, the individual objects of the image are reconstructed again in the receiver using the modeling parameters that have been transmitted.
One possibility for modeling the objects consists in a series expansion of the image function in accordance with a set of suitably selected base functions. The modeling parameters then correspond to the expansion coefficients of this image function. Such image modeling is the foundation of transform coding. If the intention is to code individual image objects having any desired boundary, a transformation for segments having any desired boundary (as a rule not convex) is necessary.
Two fundamental approaches have existed hitherto for such a transformation.
In the method which is described in the publication by M. Gilge, T. Engelhardt and R. Mehlan, Coding of arbitrarily shaped image segments based on a generalized orthogonal transform, Signal Processing: Image Communication 1, pp. 153-180, October 1989, the given image segment is first of all embedded in a circumscribing rectangle having the smallest possible dimensioning. For this rectangle it is possible to specify a discrete cosine transform (DCT) which is completely specified by the base functions of the transformation. In order to adapt this transformation to the segment shape, the base functions defined on the rectangle are successively orthogonalized with regard to the shape of the segment. The resulting orthogonal, shape-dependent base functions then form the segment-adapted transformation sought.
One disadvantage of this solution approach may be seen in the fact that the implementation of this method requires a great deal of computing power and of memory space. Furthermore, this known method has the disadvantage that reliable statements cannot be made about the suitability of the resulting transformation for the purpose of data compression since the transformation essentially depends on the orthogonalization sequence, and thus on the specific implementation of the method.
The publication by T. Sikora and Béla Makai, Shape-adaptive DCT for generic coding of video, IEEE Trans. Circuits and Systems for Video Technology 5, pp. 59-62, February 1995 describes a method in which the given image segment is transformed separately according to rows and columns. For this purpose, all the lines of the image segment are first of all aligned on the left and successively subjected to a one-dimensional horizontal transformation whose transformation length corresponds in each case to the number of pixels in the corresponding line. The resulting coefficients are then transformed a further time in the vertical direction.
This method conceals the disadvantage, in particular, that the correlations of the brightness values of the pixels (similarities of the pixels) cannot be completely utilized on account of the resorting of the pixels.
In order to improve this method disclosed in the two publications T. Sikora and Béla Makai, Shape-adaptive DCT for generic coding of video, IEEE Trans. Circuits and Systems for Video Technology 5, pp. 59-62, February 1995, and T. Sikora, S. Bauer and Béla Makai, Efficiency of shape-adaptive 2-D transforms for coding of arbitrary shaped image segments, IEEE Trans. Circuits and Systems for Video Technology 5, pp. 254-258, June 1995, describes a method in which a transformation for convex image segment shapes which is adapted for a simple image model is carried out. In this case, however, the only image segment shapes that are permitted are those which have no interruptions (holes) on traversing rows or columns.
The known methods described above furthermore have the disadvantage that owing to the variable transformation length, standard transformation methods and/or standard modules cannot be employed for carrying out the transformation.
Furthermore, I. Donescu et al., A Comparison of Efficient Methods for the Coding of Arbitrarily Shaped Image Segments, Proceedings of Picture Coding Symposium, Melbourne, pp. 13.-15.3. 1996, pp. 181-186, 1996, discloses assigning a predetermined, fixed brightness value to all the pixels of a square image segment with 8×8 pixels which do not belong to a first image segment which is at least partly contained in the square image segment.
This method primarily has two disadvantages.
Firstly, in the boundary region of the first image segment and the square image segment, it is possible for discontinuities of the signal profile of the brightness values to occur between pixels, thereby causing high-frequency spectral components, which leads to an undesirable and unnecessary, increased coding complexity and hence requirement for transmission capacity.
Secondly, this method is restricted to block-based methods with image segments having a square shape.
Since the standard image transformation methods cannot be employed, the methods disclosed in the publications M. Gilge, T. Engelhardt and R. Mehlan, Coding of arbitrarily shaped image segments based on a generalized orthogonal transform, Signal Processing: Image Communication 1, pp. 153-180, October 1989; T. Sikora and Béla Makai, Shape-adaptive DCT for generic coding of video, IEEE Trans. Circuits and Systems for Video Technology 5, pp. 59-62, February 1995; T. Sikora, S. Bauer and Béla Makai, Efficiency of shape-adaptive 2-D transforms for coding of arbitrary shaped image segments, IEEE Trans. Circuits and Systems for Video Technology 5, pp. 254-258, June 1995, incur considerable costs for encoding units which use the known methods described above.
Standard image transformation methods are disclosed in R. J. Clarke: Transform Coding of Images, Academic Press, London, pp. 72-134, 1985.
German Patent Document DE 41 36 636 A1 discloses a device for coding video signals which is used to implement signal extrapolation of a signal profile from a first picture area into a second picture area. The signal extrapolation is effected by mirroring the signal profile at d of the edge of the first picture area.
The publication by J.-R. Ohm, Digitale Bildcodierung [Digital ImageCoding], Springer, Berlin, ISBN 3-540-58579-6 pp. 32-49, 1995, demonstrates various possibilities for signal extrapolation: a periodic continuation, a symmetrical continuation and also a value-constant continuation of the signal profile.
All the methods disclosed in German Patent Document DE 41 36 636 A1 and the publication by J.-R. Ohm, Digitale Bildcodierung [Digital ImageCoding], Springer, Berlin, ISBN 3-540-58579-6 pp. 32-49, 1995, have the underlying disadvantage that in the event of a spectral transformation onto the entire signal, resulting from the signal profile and the extrapolated signal profile, an undesirably
Bayat Ali
Bell Boyd & Lloyd LLC
Patel Jayanti K.
Siemens Aktiengesellschaft
LandOfFree
Computer-assisted process and device for processing the... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Computer-assisted process and device for processing the..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Computer-assisted process and device for processing the... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3027869