Image analysis – Image compression or coding – Pyramid – hierarchy – or tree structure
Reexamination Certificate
2000-02-24
2004-10-05
Do, Anh Hong (Department: 2624)
Image analysis
Image compression or coding
Pyramid, hierarchy, or tree structure
C702S075000, C703S002000, C703S005000, C704S204000, C704S229000
Reexamination Certificate
active
06801666
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention concerns digital signal filtering, such as the transformation of a digital signal into frequency sub-band signals.
2. Description of Related Art
Many digital filtering methods and devices are known. Analysis filterings and corresponding digital signal synthesis filterings are considered here by way of example.
These filterings are generally subsystems integrated into coding and/or decoding systems. They often require a large amount of random access memory or buffer memory space, for storing the data in the course of processing.
However, in practice, the size of the memory means is often less than the size which would be necessary for storing an entire set of data, for example of a digital image.
SUMMARY OF THE INVENTION
The present invention firstly provides a method and a device for transforming a digital signal which optimise the buffer memory occupation of the data in the course of processing.
Since the size of the memory means is often less than the size which would be necessary for storing an entire set of data, it is therefore necessary to “cut” the signal into blocks and to process the blocks one after the other.
However, between an analysis and the corresponding synthesis of a signal, other processings, such as quantization or entropic coding, are generally applied to said signal. These processings, combined with processing by blocks, cause degradation in the reconstructed signal.
The present invention also provides a method and a device for transforming a digital signal which processes the signal by blocks, whilst limiting the degradation in the reconstructed signal, where other processings are applied to the signal between its transformation and its reconstruction.
The considered filterings are implemented by trellis filters. For practical reasons, it is often necessary to modify the theoretical calculations during implementation.
For example, these filterings often require a large amount of random access memory or buffer memory space, for storing the data in the course of processing. The data are then processed by blocks, as previously exposed.
However, it is known that processing by blocks causes degradation in the reconstructed signal.
The present invention also provides a method and device for transforming a digital signal which limits the degradation in the reconstructed signal.
The invention proposes a method of analysis filtering of an original digital signal including original samples representing physical quantities, original samples of the digital signal being transformed by successive calculation steps into high and low frequency output samples, any sample calculated at a given step being calculated by a predetermined function of original samples, and/or previously calculated samples, the samples being ordered in increasing rank,
characterised in that:
the signal is processed by successive series of samples, the calculations made on any series not taking into account the samples in a following series, and in that said any series terminates in a low-frequency sample.
The invention also proposes an analysis filtering method of an original digital signal including original samples representing physical quantities, comprising the following steps:
dividing the original signal in order to form plural series of samples,
filtering the original samples in a predetermined order and series by series, in order to generate at least one series of high and low-frequency samples,
wherein the end of said at least one series of high and low-frequency samples is a low-frequency sample.
According to the invention, the buffer memory space required for the filtering is reduced, because it is not necessary to store simultaneously all the samples in the buffer memory when being filtered series by series.
According to an other effect, as shown in
FIG. 22
a
, where the end of a series is a high-frequency sample, it is impossible to utilise sample B when sample A is synthesized because sample B in next series has not been generated yet. As a result, a discontinuity between samples A and B is generated.
FIG. 22
b
where the end of a series is a low-frequency sample, when samples C and D are synthesized respectively, it is possible to use low-frequency sample in previous rank which is necessary for generating a new accurate low-frequency sample. Namely, the distortion is mainly controlled by low-frequency samples. Therefore, it is possible to refer low-frequency sample over the boundary to reduce the distortion. As a result, a discontinuity between samples C and D is not generated, so sample C will be used for the synthesis of D and E samples. This configuration considerably limits the degradations on bordering samples.
The invention also proposes a method of synthesis filtering of a digital signal including high and low-frequency interlaced samples obtained by applying the above analysis filtering method to original samples, wherein:
the signal is processed by successive series of samples in a predetermined order, and the end of the series is a low-frequency sample.
By such analysis method, the distortions on the synthesized signals, especially on borders of series, are cancelled. The synthesis then needs only small buffer memory and provides signal without any distortion.
The invention also proposes a method of synthesis filtering of a digital signal including high and low-frequency interlaced samples obtained by applying the above analysis filtering method to an original digital signal including samples representing physical qualities, samples being ordered by increasing rank,
characterised in that:
the signal is processed by successive series of samples, the calculations made on any series not taking into account the samples of a following series, and in that said any series terminates in a low-frequency sample.
In addition, the buffer memory space taken up by the data currently being processed is optimised, since the signal is processed by blocks. Thus complex filterings can be integrated into numerous appliances, without these requiring very large memories.
According to a preferred characteristic, on synthesis, each series of samples terminates after a last sample of a series determined at the time of an analysis filtering as defined above.
According to another preferred characteristic, both on analysis and on synthesis, said any series terminates in a low-frequency sample of the lowest resolution level. This configuration considerably limits the degradation in the reconstructed signal.
The invention also proposes a method of analysis filtering of an original digital signal including original samples representing physical quantities, original samples of the digital signal being transformed by successive calculation steps into high and low frequency output samples, any sample calculated at a given step being calculated by a predetermined function of original samples, and/or previously calculated samples, the samples being ordered in increasing rank,
characterised in that:
the signal is processed by first successive input blocks of samples, the calculations made on a first input block under consideration taking into account only the original or calculated samples belonging to the first input block under consideration,
the first input block under consideration and the first following input block overlap over a predetermined number of original samples.
According to preferred characteristics:
the start limit of the first input block under consideration is formed between a first original sample and a first output sample, passing successively from a previous sample to a following sample calculated according to the previous sample, the following sample having a rank equal to or greater than the previous sample,
the end limit of the first input block under consideration is formed between a second original sample and a second output sample, passing successfully from a previous sample to a following sample calculated according to the previous sample, the following sample having a rank equal to or lower than the p
Berthelot Bertrand
Henry Felix
Majani Eric
Canon Kabushiki Kaisha
Do Anh Hong
LandOfFree
Device and method for transforming a digital signal does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Device and method for transforming a digital signal, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Device and method for transforming a digital signal will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3332841