Image analysis – Image compression or coding
Reexamination Certificate
2000-03-31
2003-10-07
Couso, Jose L. (Department: 2621)
Image analysis
Image compression or coding
Reexamination Certificate
active
06631213
ABSTRACT:
The present invention relates to a method and device for coding digital signals, to a method and device for decoding these same digital signals, and to systems using them.
The digital signals in question can be images, video signals, sound signals or data. Here the invention will be described particularly in its application to digital signals representing images.
In the field of digital signal coding, it is sometimes advantageous to use several coding techniques for coding the same set of data. In such case, by virtue of a selection method, based for example on an optimum transmission rate allocation criterion, the best coding technique is chosen and applied locally.
For the decoder to be able to correctly decode the coded signals which it receives, an item of information referred to as a “flag” is, conventionally, generally transmitted, making it possible to know which coding technique has been used on each signal portion.
Thus, for example, in accordance with the prior art, a digital image can be divided into blocks, and, for coding each block, it is possible to have a choice between coding by vector quantisation and coding by discrete cosine transform (DCT). Each block is coded by the most appropriate method within the terms of a criterion fixed a priori, and a flag is transmitted with the image signal, in order to indicate, for each block, whether vector quantisation or DCT has been used.
The flag is for example a word able to take as many different values as there are different coding techniques used for coding the blocks.
This way of proceeding according to the prior art has notably the drawback of requiring the transmission, from the coder to the decoder, of a large amount of information in addition to the coded signal proper, and therefore of reducing the compression rate of the signals, which consequently limits the transmission rate.
In an article entitled “High compression image coding using an adaptive morphological sub-band decomposition”, in Proceedings of the IEEE, vol. 83, No 2, February 1995, O. Egger, W. Li and M. Kunt proposed a technique for coding an image signal using an adaptive decomposition of the signal, which uses either linear filters, for the textured regions of the image, or morphological filters, for the other regions of the image. No “flag” is transmitted to the decoder.
On the other hand, this method of the prior art has the drawback of requiring the coder and decoder to both calculate the local variance for each pixel of the original image, in order to determine whether or not the region in question is a textured region, knowing that a texture has a high local variance in all directions.
This method therefore has the drawback of entailing high calculation cost due to the seeking of textured regions, given that the decision on the textured character is made pixel by pixel.
The aim of the present invention is to remedy the aforementioned drawbacks of the prior art, by eliminating the need to transmit, from the coder to the decoder, additional explicit information indicating the coding technique used and lessening the calculation cost, whilst making it possible to improve the compromise between the compression and distortion rates of the signals.
To this end, the present invention proposes a method of coding a digital signal, characterised in that:
(a) a signal simplification step is performed, during which a simplified version of the digital signal is determined, containing only part of the information contained in this digital signal;
(b) a step of coding the simplified signal is performed, during which the simplified version is coded by means of a predetermined single coding technique;
(c) a subtraction step is performed, during which this simplified version is subtracted from the digital signal so as to obtain a residual signal;
(d) a coding selection step is performed, during which, on the basis of each block of the simplified version, one coding technique, amongst a predetermined number M of coding techniques Ci, where i is an integer between 1 and M, is selected for each corresponding block of the residual signal;
(e) a residual signal coding step is performed, during which each block of the residual signal is coded by means of the previously selected coding technique.
Thus the choice of the coding technique is made on the basis of the simplified version of the original signal, and no flag directly bearing a mention of the coding technique used is transmitted from the coder to the decoder.
Given that the simplified version of the original signal contains a smaller quantity of information compared with the original signal, it offers great simplicity of coding. In addition, this simplified version being a subset of the total information contained in the original signal, the addition cost, compared with the cost necessary for coding the image in its totality, is nil, or low where there exists redundancy between the simplified version and the residual signal. This thus makes it possible to save on all or part of the transmission rate which, according to the prior art, was reserved for the transmission of flags.
The above simplified signal coding step (b) can be performed either at the end of the steps (a) of simplification, (c) of subtraction, (d) of coding selection or (e) of residual coding. This gives the present coding method great flexibility of implementation.
In a particular embodiment where the digital signal is an image signal, according to a first variant, the aforementioned simplified version can have the same spatial resolution as the original digital signal. For example, it can be a simplified image obtained by opening-closing followed by morphological reconstruction. For more details on the concept of morphological reconstruction, reference can usefully be made to the article by P. Salembier entitled “Morphological multiscale segmentation for image coding”, in Signal Processing, 1994, No 38, pages 359-386.
This type of operators had the advantage of eliminating all the objects smaller than a certain size, and of restoring the contours of the objects which have not been completely eliminated. This still further facilitates the coding of the simplified image.
In this particular embodiment, according to a second variant, the simplified version can have a spatial resolution less than that of the original signal.
Such a simplified image can for example have been obtained by wavelet decomposition of the original signal and definition of the simplified image as being a sub-band obtained by low-pass filtering during this wavelet decomposition.
The obtaining of such a simplified image has the advantage of guaranteeing a concentration, in the aforementioned low sub-band, of the information contained in the original image.
According to a first aspect of the coding method of the present invention, at the coding selection step (d), a prediction step is performed during which, on the basis of the simplified version, the most appropriate coding technique amongst the M coding techniques Ci is predicted for each block of the digital signal, and in addition steps are performed according to which:
(f) an optimised coding selection step is performed, during which, using a predetermined optimisation criterion, the most appropriate coding technique amongst the M coding techniques Ci is selected for each block of the residual signal, and this block is coded by means of the coding technique selected;
(g) a comparison step is performed, during which, for each block of the digital signal, the coding technique predicted at the coding selection step (d) and the coding technique selected at the optimised coding selection step (f) are compared; and
(h) a storage step is performed, during which, if the predicted and selected coding techniques are identical, there is stored, with a view to decoding, a flag according to which the prediction made at the coding selection step (d) is correct, and otherwise a flag is stored according to which this prediction is false.
Thus there is transmitted, from the coder to decoder, not directly information relating to the cod
Amonou Isabelle
Henry Felix
Canon Kabushiki Kaisha
Couso Jose L.
Fitzpatrick ,Cella, Harper & Scinto
LandOfFree
Methods and devices for coding and decoding digital signals,... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Methods and devices for coding and decoding digital signals,..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Methods and devices for coding and decoding digital signals,... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3127245