KR100941582B1 - Adaptive update method and updating apparatus for lifting based directional wavelet transform and recording medium therefor - Google Patents

Abstract

The present invention relates to a method and an update apparatus for adaptive update of lifting based directional wavelet transform and a recording medium therefor.

The present invention provides a multiphase sample characteristic analyzer which analyzes characteristics of even polyphase samples and predicted odd polyphase samples; A multiple update direction selector for selecting at least two update directions (multiple update directions) for removing aliasing in the low band; A residual sample position calculator for determining the positions of the plurality of predicted odd polyphase samples needed to remove aliasing; A directional wavelet designed based on the lifting scheme; including an updater configured to perform an update by using the multiple update direction selected by the multiple update direction selector and the positions of the plurality of predicted odd polyphase samples by the residual sample position calculator When the spatial correlation in the image is removed by using a transform, the coding efficiency can be increased by effectively eliminating aliasing of the low band signal.

Description

Adaptive update method and apparatus for lifting based directional wavelet transform and recording medium therefor {Method and Apparatus for Adaptive Update of Directional Wavelet Transform based on Lifting, and Recording Medium therefor}

The present invention relates to an adaptive update method and an update apparatus for a lifting-based directional wavelet transform, and a recording medium therefor. More specifically, the coding efficiency of the wavelet transform-based image data compression technique in the field of large-capacity image data compression technology is increased. Adaptive update method and updating apparatus of lifting based directional wavelet transform to effectively remove aliasing of low band signal when removing spatial correlation in image using directional wavelet transform designed based on lifting scheme And a recording medium therefor.

Image data compression technology based on wavelet transform solves the blocking phenomenon of conventional block-oriented data processing methods such as JPEG, and scales and progressive transmissions according to the environment of transmission and storage media. It is expected to be applied to JPEG2000, which has recently been standardized as an international standard, and Dirac, a video compression technology developed by the British BBC.

The wavelet transform technique recursively takes two one-dimensional filters consisting of a low band filter and a high band filter on an input image in a horizontal direction and a vertical direction to convert an image signal in a spatial domain into a wavelet domain. Sufficient directional vanishing moments can be provided in the horizontal and vertical directions, resulting in a high energy concentration for flat image signals.

In this case, the directional attenuation moment refers to a state in which energy of the wavelet transform coefficient is sufficiently collected in the low band subband, thereby minimizing energy transfer in the high band filtering direction.

Recently, it is recognized that the standard discrete wavelet transform structure has a limitation in showing sufficient energy concentration effect along lines, edges, and contours, which are inherent to image signals. Discrete wavelet transform techniques have been proposed.

To explain the fundamental reason of this trend in more detail, the conventional discrete wavelet transform structure is a wavelet transform is performed by filtering in two horizontal and vertical directions, so that when the contour or boundary of the image is determined to be out of these two directions, This is because there is not enough Directional Vanishing Moments in the band signal.

[Reference 1] Roberto H. Bamberger, and Mark JT Smith (A Filter Bank for the Directional Decomposition of Images: Theory and Design, IEEE Trans.On Signal Processing, vol. 40, no.4, pp.882-893, April , 1992)

[2] E. J. Candes ("Ridgelets: Theory and applications, Ph.D. dissertation, Dept. Statistics, Stanford Univ., Stanford, CA, 1998)

EJ Candes and DL Donoho ("Curvelets-a surprisingly effective nonadaptive representation for objects with edges", in Curve and Surface Fitting, A Choen, C. Rabut, LL Schmumaker, Eds.Saint-Malo: Vanderbit University Press, 1999)

In order to solve the above problem, [1] attempted to provide Directional Vanishing Moments to a high band signal using a filter bank having a frequency response in various directions as well as in the horizontal and vertical directions. [2] and [3] were able to significantly reduce the energy of the high band filter band by converting two-dimensional continuous signals based on the characteristic information (lines and curves) of the image.

[4] Vladan Velisavlijevic, Baltasar Beferull-Lozano, Martin Vetterli, and Pier Luigi Dragotti ("Directionlets: Anisotropic Multidirectional Representation With Separable Filtering", IEEE Trans. Image Processing, vol. 15, no.7, pp.1916-1933 , Jul. 2006)

However, there is a problem in that the above-described methods require excessive computation because it is impossible to independently implement a one-dimensional filter like the existing wavelets.

In order to solve the above problems and effectively express the contour information existing in the image, [4] designs an directional asymmetric base filter banks (Anisotropic Basis Filter Banks) and uses a lattice structure. After determining two dominant directions of the image, we proposed a method of applying one-dimensional low-band filter in each direction.

By applying one-dimensional filtering according to the dominant contour direction existing in the image and performing sub-sampling according to the filtering direction, this method can provide highly effective directional vanishing moment with only a similar amount of computation as the existing wavelet technique. have.

For example, the method disclosed in [Document 4] assumes that two contour directions of filtering direction 1 and filtering direction 2 exist in the image, and then performs sub-sampling after performing low band filtering in the direction of "filtering direction 1". To obtain a reduced sized image of 1/2 size, perform low-band filtering in the direction of `` filtering direction 2 '', and then perform sub-sampling in `` filtering direction 2 '' to generate a low-band signal. will be.

The low-band signal generated as described above becomes an image of which the width and length are reduced to 1/2 of the initial input image. However, since the low-band signal is sub-sampled in two different directions, the initial input image is symmetrical with the sub-sampling direction. It is given in a distorted form.

Accordingly, when compression coding is performed by recursively transforming such low-band images, scalability and progressive transmission, which are advantages of wavelet transform-based image compression schemes, cannot be provided.

Chuo-Ling Chang, and Bernd Girod ("Direction-Adaptive Discrete Wavelet Transform for Image Compression", IEEE Trans. Image Processing, vol. 16, pp. 1289-1302, May 2007)

[Document 6] System and method for image coding using hybrid directional prediction and lifting wavelet technology (Korean Patent Publication No. 10-2006-0049812)

[5] solves the above-mentioned problems (which cannot provide scalability and progressive transmission, etc.) by applying a one-dimensional directional filter to the `` prediction '' stage based on the lifting method, and provides high directional vanishing for high-band subbands. We designed a wavelet transform with low computational complexity that can provide Moments.

In addition, [Document 6] also proposes a lifting wavelet transform method using such a directional prediction.

As shown in FIG. 1, the lifting-based wavelet transform apparatus includes an analysis unit 10 (FIG. 1A) for converting an input signal into wavelet coefficients, and an integrated unit 20 (FIG. 1B) for converting wavelet coefficients into a reconstruction signal. The analyzer 10 includes a divider 11, a predictor 13, and an updater 15.

Referring to the one-dimensional wavelet transform, the divider 11 divides the one-dimensional signal into even and odd polyphase samples, and the predictor 13 divides the plurality of odd polyphase samples into a plurality of even numbers around. Prediction from a polyphase sample is performed to subtract the prediction from an odd polyphase sample to yield a residual polyphase sample.

Finally, the updater 15 generates an updated polyphase sample by predicting an even polyphase sample from the generated residual polyphase sample and adding the prediction value to the even polyphase sample.

In this case, if a specific type of filter is used for prediction and update, the updated even polyphase sample becomes a wavelet coefficient passing through the low band filter, and the predicted odd polyphase sample (residual polyphase sample) is a wavelet passing through the high band filter. It becomes a coefficient. These wavelet coefficients can be restored to the original signal by the integrating unit 20 of FIG. 1B.

The lifting-based wavelet described above can be extended and applied to a two-dimensional signal, which performs division / prediction / update again in the horizontal direction on the updated polyphase sample and residual polyphase sample generated after performing division / prediction / update in the vertical direction. To perform the wavelet transform on the 2D signal.

Document 5 introduces directional prediction to the transformation of a two-dimensional signal using the lifting-based wavelet described above, and provides a higher performance directional vanishing moment than a simple wavelet of the two-dimensional signal.

FIG. 2 is a diagram illustrating a configuration of a conventional update device in detail. The update device disclosed in FIG. 2 corresponds to an update unit of a lifting-based wavelet transform device, and will be referred to as an update device for convenience of description.

As shown in the drawing, the update device 30 includes an update direction selector 31, a residual sample availability checker 33, and an updater 35.

In more detail, the update direction selector 31 receives the prediction direction and determines the update direction. At this time, the update direction selector 31 selects one update direction in the same direction as the prediction direction.

The residual sample availability checking unit 33 receives the update direction from the update direction selector 31, and then checks the positions of the plurality of odd polyphase samples (residual polyphase samples) existing at the same position as the update direction. Then, a plurality of odd polyphase samples (residual polyphase samples) existing at the corresponding position are output as a result (predicted odd polyphase sample).

The updater 35 updates an even polyphase sample to be updated by a plurality of predicted odd samples (residual polyphase samples) by the residual sample availability checking unit 33. Here, the updater 35 updates only the samples that have affected the prediction of the even polyphase sample to be updated when the even polyphase sample is updated.

Referring to FIG. 3, the samples in the row with gray samples represent the predicted odd polyphase samples (residual polyphase samples) and the white samples represent the samples to be updated.

The dashed black arrow indicates the update direction (prediction direction: the same as the update direction and the prediction direction), and the gray arrow indicates the direction of dividing the odd polyphase sample and the even polyphase sample, that is, the sub-sampling direction.

If it is assumed that the sample consisting of a dotted line in the middle of the even polyphase samples is updated, the updating apparatus 30 may include samples existing in a direction coinciding with the update direction among the predicted odd polyphase samples (remaining polyphase samples). It is to update the even polyphase sample to be updated by filtering only the test samples, which are the samples that influenced the prediction of the even polyphase sample to be updated.

In the above-described method, the predicted odd signal is represented as a high band wavelet signal and the updated even signal is represented as a low band wavelet signal as in the case of a general lifting based wavelet.

The method of [5] provided higher directional vanishing moments in the high band subband, showing higher compression performance than the wavelet of the simple two-dimensional signal. However, this technique has a problem in that even polyphase samples are not updated efficiently when generating low band subbands, which causes difficulty in compression in the low band.

As shown in FIG. 3, it can be seen that the sub-sampling direction and the update direction for distinguishing the even polyphase sample and the odd polyphase sample are different from each other. That is, since the filtering direction and the sampling direction are different from each other, the edges are not well aligned in the low band, so when performing the wavelet transform on the low band, it is difficult to provide high directional vanishing moments for the high band signal.

In addition, since the division direction and the sampling direction of the frequency domain are different from each other by the directional filtering, an aliasing phenomenon occurs in the low band in the frequency domain, which causes a problem that the efficiency is lowered when the low band is encoded.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problems. When the spatial correlation in an image is removed by using a directional wavelet transform designed based on the lifting scheme, the present invention effectively removes aliasing of a low band signal. There is a technical problem to provide an adaptive update method and update apparatus for the based directional wavelet transform and a recording medium therefor.

In addition, the present invention is an even polyphase sample in consideration of samples other than the sample affecting the multiple update direction and prediction when the filtering direction and the sampling direction are different from each other in the directional prediction and lifting based wavelet technology. The purpose is to enable low pass filtering.

In addition, an object of the present invention is to reduce the amount of computation when performing the adaptive update of the lifting based directional wavelet transform.

The present invention for achieving the above object is an update device for providing an adaptive update of the lifting based directional wavelet transform,

A multiphase sample characteristic analyzer for analyzing characteristics of the even polyphase sample and the predicted odd polyphase sample; A multiple update direction selector for selecting one direction different from the prediction direction or at least two update directions (multiple update directions) to remove aliasing in the low band; A residual sample position calculator for determining the positions of the plurality of predicted odd polyphase samples needed to remove aliasing; And an updater configured to perform an update by using the multiple update direction selected by the multiple update direction selector and the positions of the plurality of predicted odd polyphase samples by the residual sample position calculator.

The multiple update direction selection unit includes prediction direction analysis means for analyzing a prediction direction; Sampling direction analyzing means for analyzing a sampling direction; And update direction determining means for combining the analysis result calculated from the prediction direction analysis means and the sampling direction analysis means and the analysis result calculated from the multiphase sample characteristic analyzer to determine the update direction.

The residual sample position calculation unit may include prediction direction analysis means for analyzing a prediction direction; Sampling direction analyzing means for analyzing a sampling direction; And a sample position for determining the positions of the plurality of predicted odd polyphase samples necessary for eliminating aliasing by combining the analysis results calculated from the prediction direction analysis means and the sampling direction analysis means with the analysis results calculated from the multiphase sample characteristic analyzer. It is preferable to include a calculation means.

Preferably, the sample position calculation means determines the positions of the plurality of odd polyphase samples in consideration of both the odd polyphase samples having an influence on the even polyphase samples and the odd polyphase samples having no influence.

The updater may perform filtering by assigning a weight to a specific sample, or perform filtering by a specific filtering coefficient.

Another embodiment of the present invention is an adaptive update method of lifting-based directional wavelet transform in an update device.

a) The updating device selects one update direction different from the prediction direction or selects at least two or more update directions based on the even polyphase sample and the predicted odd polyphase sample, the prediction direction and the sampling direction received from the lifting-based directional wavelet transform device. Selecting; b) selecting a plurality of predicted odd polyphase samples based on the even polyphase sample and the predicted odd polyphase sample, the prediction direction and the sampling direction; And c) a plurality of even polyphase samples using a plurality of predicted odd polyphase samples present in one update direction or at least two or more update directions different from the prediction direction and the plurality of predicted odd polyphase samples selected in step b). It includes; updating.

Prior to the step a), the lifting-based directional wavelet transform apparatus separates the even position data and the odd position data along the vertical direction or the horizontal direction into an even polyphase sample and an odd polyphase sample; Selecting a direction most similar to an edge of the image as a prediction direction for prediction; And predicting a plurality of odd polyphase samples from a plurality of neighboring even polyphase samples positioned in the prediction direction.

Yet another aspect of the present invention is to provide a recording medium for providing an adaptive update method of a lifting based directional wavelet transform.

a) selecting one update direction different from the prediction direction or selecting at least two or more update directions based on the even polyphase sample and the predicted odd polyphase sample, the prediction direction and the sampling direction; b) selecting a plurality of predicted odd polyphase samples based on the even polyphase sample and the predicted odd polyphase sample, the prediction direction and the sampling direction; And c) a plurality of even polyphase samples using a plurality of predicted odd polyphase samples present in one update direction or at least two or more update directions different from the prediction direction and the plurality of predicted odd polyphase samples selected in step b). It includes a function to update.

As described above, the adaptive update method and update apparatus of the lifting-based directional wavelet transform according to the present invention and the recording medium therefor use a directional wavelet transform designed based on the lifting method to remove spatial correlation in the image. There is an advantage in that coding efficiency can be increased by effectively eliminating aliasing of a signal.

In addition, the present invention is an even polyphase sample in consideration of samples other than the sample affecting the multiple update direction and prediction when the filtering direction and the sampling direction are different from each other in the directional prediction and lifting based wavelet technology. Since the low pass filtering can be performed, the wavelet transform can be recursively performed in the low band when the wavelet based image and the video are compressed.

In addition, the present invention has the effect of reducing the amount of computation when performing the adaptive update of the lifting-based directional wavelet transform.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

4 is a diagram showing in detail the configuration of the updating apparatus according to the present invention.

As shown, the updating apparatus 100 includes a multiphase sample characteristic analyzer 110, a multi-update direction selection unit 130, a residual sample position calculator 150, and an update unit 170.

In more detail, the multiphase sample characteristic analyzer 110 performs an even polyphase sample, a predicted odd polyphase sample, a prediction direction, and a sampling direction that are the results after the decomposition and prediction steps are performed by a lifting wavelet transform device (not shown). If is input, the aliasing characteristics of the even polyphase sample are analyzed by frequency analysis of the even polyphase sample and the predicted odd polyphase sample.

The multiple update direction selector 130 selects an update direction for removing aliasing in the low band.

For example, the multi-update direction selector 130 may include at least two or more based on the prediction direction inputted to the multiphase sample characteristic analyzer 110 and the sampling direction and the results analyzed by the multiphase sample characteristic analyzer 110. Either the update direction (multiple update direction) or one direction different from the prediction direction is selected to remove aliasing in the low band.

Residual sample position calculator 150 locates the plurality of predicted odd polyphase samples required to remove aliasing.

The updater 170 performs an update process using the update direction and the predicted odd polyphase samples. That is, the updater 170 makes the even polyphase sample have the overall characteristics of the original data by using the predicted odd polyphase sample, and removes the overlapped portion of the even polyphase sample by using the predicted odd polyphase sample. .

The updater 170 adapts to reduce aliasing in the low band by using the multiple update direction generated by the multiple update direction selector 130 and the positions of the odd polyphase samples predicted by the residual sample position calculator 150. The update process. In this case, the updater 170 performs filtering by assigning a weight to a specific sample or performs filtering by a specific filtering coefficient.

5 is a diagram showing in detail the configuration of the multiple update direction selection unit according to the present invention.

As shown, the multi-update direction selection unit 130 is a prediction direction analysis means 131 for analyzing the prediction direction input to the multi-phase sample characteristic analysis unit 100, sampling input to the multi-phase sample characteristic analysis unit 100 Update by combining the analysis results calculated from the sampling direction analysis means 133 and the prediction direction analysis means 131 and the sampling direction analysis means 133 and the multiphase sample characteristic analyzer 110 analyzing the direction. Update direction determining means 135 for determining the direction.

Here, the result of the update direction determining means 135 is not just one update direction coinciding with the prediction direction as in the conventional method proposed in [5], but two or more, i.e., multiple, update elimination directions. Generate direction as output.

6 is a view showing in more detail the configuration of the residual sample position calculation unit according to the present invention.

As shown, the residual sample position calculation unit 150 includes prediction direction analysis means 151 for analyzing the prediction direction, sampling direction analysis means 153 for analyzing the sampling direction, and prediction direction analysis means 151 and the sampling direction. Sample position calculation means 155 for locating the predicted odd polyphase samples necessary for eliminating aliasing by combining the analysis result calculated from the analysis means 153 and the analysis result calculated from the multiphase sample characteristic analyzer 110. It includes.

The sample position calculating means 155 uses only the predicted odd polyphase samples, which are samples in which the even polyphase sample to be updated has influenced the prediction (filtering) in order to update the even polyphase sample as in the method proposed in the related art [Filter 5] Instead of updating the even polyphase sample to be updated, we consider the predicted odd polyphase samples that the even polyphase sample to be updated to remove aliasing did not affect the prediction.

Accordingly, the sample position calculation means 155 identifies the predicted odd polyphase samples needed to remove aliasing regardless of whether the even polyphase sample to be updated has been affected in the prediction and outputs the positions of the samples as a result.

7 is a view for explaining an example of the update method according to the present invention.

Referring to FIG. 7, gray samples represent predicted odd polyphase samples and white samples represent samples to be updated, an arrow composed of gray dotted lines represents a prediction direction, and a gray line represents a sub-sampling direction.

"Update direction 1", "update direction 2", and "update direction 3" are the results output by the multiple update direction selector 130. The even polyphase sample to be updated according to each direction may update the even polyphase sample using not only the plurality of predicted odd polyphase samples located in the prediction direction but also the predicted odd polyphase samples present in the newly determined "update direction". .

Here, the number of update directions is not limited, and the length of each update direction may be determined differently.

For example, as shown in FIG. 7, "updating direction 1" uses six samples, "updating direction 2" uses four samples, and "updating direction 3" uses only two samples to select even samples. Can be updated.

The black rhombic samples are odd polyphase samples predicted using the output result of the residual sample position calculator 150, and these samples also affect even samples to be updated. That is, the black samples are used to update the even polyphase sample and the even polyphase using all samples available irrespective of the prediction, as well as the predicted odd polyphase samples in which the even polyphase sample to be updated affected the prediction. Update the sample.

FIG. 8 is a flowchart illustrating an example of an update method according to the present invention. An example in which the above-described method is applied to a 2D wavelet signal will be described.

After the one-dimensional directional wavelet is executed in the vertical direction, the one-dimensional directional wavelet is further performed in the horizontal direction. Here, the order of the vertical and horizontal directions can be changed.

First, the updating apparatus 100 divides the even position data and the odd position data along the vertical direction and separates the even polyphase sample and the odd polyphase sample (S101).

Next, in order to best predict odd polyphase samples with even polyphase samples, the updating apparatus 100 selects an edge direction of the image as a prediction direction for prediction (S103).

For example, the update apparatus 100 selects the direction most similar to the edge of the image as the prediction direction, as shown in FIG. 7.

The updating apparatus 100 predicts a plurality of odd polyphase samples from a plurality of neighboring even polyphase samples positioned in the selected prediction direction (S105).

The updating apparatus 100 selects one update direction different from the prediction direction or selects at least two or more update directions so that no aliasing occurs in the low band based on the even polyphase sample and the predicted odd polyphase sample, the prediction direction, and the sampling direction. (S107).

Subsequently, the updating apparatus 100 selects a plurality of predicted odd polyphase samples to remove aliasing in the low band based on the even polyphase sample and the predicted odd polyphase samples, the prediction direction, and the sampling direction (S109).

Thereafter, the updating apparatus 100 uses the plurality of predicted odd polyphase samples present in one update direction or at least two or more update directions different from the prediction direction and the plurality of predicted odd polyphase samples selected in step S109 to provide a plurality of even numbers. The polyphase sample is updated (S111).

The updating apparatus 100 separates even and odd polyphase samples in the horizontal direction (S113).

Subsequently, the update apparatus 100 selects a direction most similar to an edge of the image as a direction for prediction in order to best predict odd polyphase samples with even polyphase samples (S115).

The updating apparatus 100 predicts odd polyphase samples from neighboring polyphase samples existing in the selected prediction direction (S117).

The updating apparatus 100 selects one update direction different from the prediction direction or at least two or more updates using the information of the even polyphase sample and the predicted odd polyphase sample, the prediction direction, and the sampling direction so that no aliasing occurs in the low band. The direction is determined (S119).

Subsequently, the updating apparatus 100 selects a plurality of predicted odd samples to remove aliasing in the low band using the information of the even polyphase sample, the predicted odd polyphase samples, the prediction direction, and the sampling direction (S121).

Thereafter, the updating apparatus 100 uses the plurality of predicted odd polyphase samples present in one update direction or at least two or more update directions different from the prediction direction, and the plurality of even polyphase samples using the plurality of odd polyphase samples selected in step S121. Update (S123).

Steps S101 to S111 shown in FIG. 8 represent one-dimensional vertical transformation, and steps S113 to S123 represent one-dimensional horizontal transformation.

Since the direction of the update step disclosed in FIG. 8 is different from the direction of the prediction step, it may include samples other than the sample that influenced the prediction, and the selected residual polyphase samples may also include samples other than the sample that influenced the prediction. Can be.

The adaptive update mode method for aliasing removal of a lifting based directional wavelet according to the present invention can be implemented as computer readable codes on a computer readable recording medium. Computer-readable recording media include all types of recording devices that store data that can be read by a computer system. For example, ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage. Device and the like.

In addition, the recording medium may also be implemented in the form of a carrier wave (for example, transmission over the Internet), and may be distributed and distributed in a networked computer system so that computer-readable code is stored and executed in a distributed manner.

As such, those skilled in the art will appreciate that the present invention can be implemented in other specific forms without changing the technical concept or essential characteristics. Therefore, the above-described embodiments are to be understood as illustrative in all respects and not as restrictive. The scope of the present invention is shown by the following claims rather than the detailed description, and all changes or modifications derived from the meaning and scope of the claims and their equivalents should be construed as being included in the scope of the present invention. do.

As described above, the adaptive updating method and update apparatus of the lifting-based directional wavelet transform according to the present invention and the recording medium therefor use a directional wavelet transform designed based on the lifting method to remove spatial correlation in the image, and thus, low band. It is suitable for the high necessity to increase the coding efficiency by effectively eliminating aliasing of the signal.

1A and 1B are views showing the configuration of a conventional lifting-based wavelet transform apparatus;

2 is a diagram showing the configuration of a conventional update device;

3 is a view for briefly explaining a conventional update method;

4 is a diagram showing the configuration of an update device according to the present invention;

5 is a view showing in detail the configuration of the multiple update direction selection unit of FIG. 4 according to the present invention;

6 is a view showing in detail the configuration of the residual sample position calculation unit of FIG. 4 according to the present invention;

7 is a view for briefly explaining an update method according to the present invention;

8 is a flowchart illustrating an example of an update method according to the present invention.

<Explanation of symbols for the main parts of the drawings>

10: analysis unit 11: dividing means

13, 23: prediction means 15, 21: update means

20: integrated unit 25: joining means

30, 100: update device 31: update direction selector

33: check availability of remaining samples

35, 170: update unit 110: multi-phase sample characteristic analysis unit

130: multiple update direction selection unit 131, 151: prediction direction analysis means

133 and 153 sampling means analyzing means 135 update direction determining means

150: residual sample position calculation unit 155: sample position calculation means

Claims (8)

An update apparatus for providing an adaptive update of lifting based directional wavelet transform, comprising: A multiphase sample characteristic analyzer for analyzing characteristics of the even polyphase sample and the predicted odd polyphase sample; A multiple update direction selector for selecting one direction different from the prediction direction or at least two update directions (multiple update directions) to remove aliasing in the low band; A residual sample position calculator for determining the positions of the plurality of predicted odd polyphase samples needed to remove aliasing; An updater for performing an update by using the multiple update direction selected by the multiple update direction selector and the positions of the plurality of predicted odd polyphase samples by the residual sample position calculator; Update device comprising a. The method of claim 1, The multiple update direction selection unit, Prediction direction analysis means for analyzing the prediction direction; Sampling direction analyzing means for analyzing a sampling direction; And Update direction determining means for determining an update direction based on the prediction direction, the sampling direction calculated from the prediction direction analysis means and the sampling direction analysis means, and the even polyphase sample and the predicted odd polyphase sample calculated from the polyphase sample characteristic analyzer; ; Update device comprising a. The method of claim 1, The residual sample position calculation unit, Prediction direction analysis means for analyzing the prediction direction; Sampling direction analyzing means for analyzing a sampling direction; And A plurality of predictions necessary to remove aliasing based on the prediction direction, the sampling direction calculated from the prediction direction analysis means and the sampling direction analysis means, and the even polyphase sample and the predicted odd polyphase sample calculated from the polyphase sample characteristic analyzer; Sample position calculation means for identifying a position of the odd-numbered polyphase sample; Update device comprising a. The method of claim 3, And the sample position calculating means identifies the positions of the plurality of odd polyphase samples in consideration of both the odd polyphase samples having an influence on the even polyphase samples and the odd polyphase samples having no influence on the prediction process. The method according to any one of claims 1 to 4, The updating unit may perform filtering by assigning a weight to a specific sample or performing filtering by a specific filtering coefficient. An adaptive update method of lifting based directional wavelet transform in an update device, a) The updating device selects one update direction different from the prediction direction or selects at least two or more update directions based on the even polyphase sample and the predicted odd polyphase sample, the prediction direction and the sampling direction received from the lifting-based directional wavelet transform device. Selecting; b) selecting a plurality of predicted odd polyphase samples based on the even polyphase sample and the predicted odd polyphase sample, the prediction direction and the sampling direction; And c) update the plurality of even polyphase samples using a plurality of predicted odd polyphase samples present in one update direction or at least two or more update directions different from the prediction direction and the plurality of predicted odd polyphase samples selected in step b) Doing; Adaptive updating method of lifting-based directional wavelet transform, comprising the. The method of claim 6, Before step a), Separating, by the lifting-based directional wavelet transform device, into an even polyphase sample and an odd polyphase sample by dividing the even position data and the odd position data along a vertical direction or a horizontal direction; Selecting a direction most similar to an edge of the image as a prediction direction for prediction; And Predicting a plurality of odd polyphase samples from a plurality of neighboring even polyphase samples positioned in the prediction direction; The adaptive update method of the lifting-based directional wavelet transform further comprises. A recording medium for providing an adaptive update method of a lifting based directional wavelet transform, a) selecting one update direction different from the prediction direction or selecting at least two or more update directions based on the even polyphase sample and the predicted odd polyphase sample, the prediction direction and the sampling direction; b) selecting a plurality of predicted odd polyphase samples based on the even polyphase sample and the predicted odd polyphase sample, the prediction direction and the sampling direction; And c) update the plurality of even polyphase samples using a plurality of predicted odd polyphase samples present in one update direction or at least two or more update directions different from the prediction direction and the plurality of predicted odd polyphase samples selected in step b) Function; A computer-readable recording medium having recorded thereon a program for executing the program.

Images