KR101221495B1 - Contents Adaptive MCTF Using RD Optimization - Google Patents

Abstract

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an adaptive MCTF method, wherein motion information of a current block is retrieved, a high frequency band image is generated using the retrieved motion information, and LaGrand is performed using a rate-distortion theory for the current block. The sum E _B of permissible error ranges per pixel is minimized to minimize the Lagrangian cost function. Also, obtain the local variance (JND) per pixel of the current block, calculate the sum of the local variance (JND _sum ) in the current block and the extra data (U _sum ) to be added to even frames, and then allow the error per pixel. Use at least one or more of the sum of the range (E _B ) or the sum of the local variances in the current block (JND _sum ) to compare with the _sum of extra data (U _sum ) to be added to the even frames. According to the comparison result, the sum of the extra data to be added to the even frame for the current block remains as it is, or the sum of the extra data to be added to the even frame is changed using the sum of the local variances in the current block, or the screen is displayed for the current block. It is characterized by the inner coding. According to the present invention, when updating an adaptive MCTF (contents adaptive MCTF), the block having inappropriate motion information is selected by applying an optimal threshold value through the RD theory, and the image quality can be improved by switching to an intra coding mode. have.

Description

Contents Adaptive MCTF Using RD Optimization

The present invention relates to a content adaptive MCTF method, and more particularly, to select blocks having inappropriate motion information by applying an optimal threshold value through RD theory when updating an adaptive MCTF (contents adaptive MCTF). The present invention relates to an adaptive MCTF method using RD theory which can improve image quality by switching to intra picture coding mode.

As the number of terminals using video information such as wireless video call or broadcasting increases, the video data transmission environment is changing in various ways. Accordingly, for video transmission from one source video to various terminals, such as a TV, a DMB, or a personal terminal, a scalable video coding (SVC) technology that can be applied to a highly variable transmission environment is required. SVC is largely classified into spatial, image quality, and temporal scalability. The SVC is composed of one bit stream that has various spatial resolutions, image quality, and various frame rates. A technology that allows the stream to be received and restored.

Motion compensated temporal filtering (MCTF), which is a temporal scalability among the three scalability of the SVC, uses a lifting scheme and plays a basic role in supporting the temporal scalability of the SVC. MCTF has the purpose of providing time scalability by using wavelet technique and at the same time eliminating redundancy between frames, and it is designed to provide scalability for image quality simultaneously with 2D wavelet technique. The MCTF provides temporal scalability by filtering the image into a mean image with low frequency neighbor images and a difference image with high frequency neighbor images on the time axis.

However, the disadvantage of MCTF is that as the step increases, the motion information between frames cannot be found accurately. The reason is that as the MCTF step progresses, the correlation between the images decreases as the time interval between the original image and the reference image increases, and the correct motion information cannot be found due to the increase of the time interval between frames. This results in the energy to be transferred to the low frequency band remaining in the high frequency band, resulting in deterioration of image quality. In addition, as the level of the MCTF increases, such deterioration of image quality accumulates, resulting in a residual image generated from a time difference between frames, resulting in a deterioration of image quality.

Therefore, the first problem to be solved by the present invention is to select the block having the inappropriate motion information by applying the optimal threshold value through the RD theory when updating the content adaptive MCTF, and switch to the intra picture coding mode By providing an adaptive MCTF method that can improve the image quality.

Further, the present invention provides a computer-readable recording medium having recorded thereon a program for executing the above method on a computer.

The present invention comprises the steps of: retrieving motion information of the current block to achieve the first object; Generating a high frequency band image using the retrieved motion information; Calculating a sum E _B of permissible error ranges per pixel to minimize a Lagrangian cost function using a rate-distortion theory for the current block; Obtaining a regional variance (JND) for each pixel of the current block and calculating a sum (J _sum ) of local variances in the current block and extra data to be added to an even frame (U _sum ); Comparing at least one of the sum of allowed error ranges per pixel E _B or the sum of local variances JND _{sum in} the current block to a sum of extra data U _sum to be added to the even frames. step; And according to the comparison result, maintain the sum of the extra data to be added to the even frame with respect to the current block or change the sum of the extra data to be added to the even frame using the sum of local variances in the current block. In addition, the present invention provides an adaptive MCTF method using an RD theory, comprising: intra-picture encoding the current block.

According to an embodiment of the present invention, when the slope of the RD curve in the RD theory is λ, the sum of the allowable error ranges per pixel and the λ that minimize the Lagrangian cost function Can be selected.

In addition, by using λ for the current frame as λ for the next frame, it is possible to reduce the complexity of the operation of minimizing the Lagrangian cost function.

According to another embodiment of the present invention, when the sum of the allowable error range per pixel and the sum of the local variance in the current block is less than the sum of the extra data to be added to the even frame, It is preferable to encode in the picture.

If the sum of the allowable error range per pixel and the sum of the local variance in the current block is smaller than the sum of the extra data to be added to the even frame, it is determined that incorrect motion information is applied.

In addition, when the sum of local variances in the current block is greater than the sum of extra data to be added to the even frame, it is preferable to maintain the sum of extra data to be added to the even frame with respect to the current block.

According to another embodiment of the present invention, the sum of the allowable error range per pixel and the sum of the local variance in the current block is greater than the sum of the extra data to be added to the even frame and the area in the current block. If the sum of the variances is smaller than the sum of the extra data to be added to the even frame, it is preferable to change the sum of the extra data to be added to the even frame using the sum of the local variances in the current block.

In order to solve the above other technical problem, the present invention provides a computer readable recording medium having recorded thereon a program for executing the above-described adaptive MCTF method on a computer.

According to the present invention, when updating an adaptive MCTF (contents adaptive MCTF), the block having inappropriate motion information is selected by applying an optimal threshold value through the RD theory, and the image quality can be improved by switching to an intra coding mode. have. In addition, according to the present invention, the PSNR can be improved over the conventional MCTF technique using the local variance by using the R-D optimized threshold.

1 illustrates an MCTF using Lifting-Scheme.
2 shows a 5 / 3-tap wavelet MCTF encoding structure.
3 is a flowchart of an adaptive MCTF method using RD theory according to a preferred embodiment of the present invention.
4 illustrates duplication of updates due to the prediction process of MCTF and backward motion information.
FIG. 5 is a graph of PSNR and bit amount indicated by applying a foreman sequence to JSVM for each MCTF.
FIG. 6 is a graph of PSNRs and bit amounts indicated by applying football sequences to JSVM for each MCTF.

Prior to the description of the specific contents of the present invention, for the convenience of understanding, the outline of the solution of the problem to be solved by the present invention or the core of the technical idea will be presented first.

An adaptive MCTF method using the RD theory according to an embodiment of the present invention comprises the steps of: retrieving motion information of the current block; Generating a high frequency band image using the retrieved motion information; Calculating a sum E _B of permissible error ranges per pixel to minimize a Lagrangian cost function using a rate-distortion theory for the current block; Obtaining a regional variance (JND) for each pixel of the current block and calculating a sum (J _sum ) of local variances in the current block and extra data to be added to an even frame (U _sum ); Comparing at least one of the sum of allowed error ranges per pixel E _B or the sum of local variances JND _{sum in} the current block to a sum of extra data U _sum to be added to the even frames. step; And according to the comparison result, maintain the sum of the extra data to be added to the even frame with respect to the current block or change the sum of the extra data to be added to the even frame using the sum of local variances in the current block. Intra picture encoding is performed on the current block.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, these examples are intended to illustrate the present invention in more detail, it will be apparent to those skilled in the art that the scope of the present invention is not limited thereby.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will now be described more fully with reference to the accompanying drawings, in which preferred embodiments of the invention are shown. In addition, in describing the operation principle of the preferred embodiment of the present invention in detail, when it is determined that the detailed description of the known function or configuration and other matters related to the present invention may unnecessarily obscure the subject matter of the present invention, The detailed description is omitted.

In addition, throughout the specification, when a part is 'connected' to another part, it is not only 'directly connected' but also 'indirectly connected' with another element in between. Include. In addition, the term 'comprising' a certain component means that the component may be further included, without excluding the other component unless specifically stated otherwise.

MCTF used to provide temporal scalability of video in SVC plays a big role in providing scalability of picture quality with temporal expansion. However, as the prediction and update stages increase due to the characteristics of the MCTF, the time interval between frames to be analyzed becomes longer and the search for the correct motion information becomes difficult, resulting in deterioration of image quality.

1 illustrates an MCTF using Lifting-Scheme.

MCTF uses a lifting scheme in applying the filter, Figure 1 shows the MCTF and the reverse transformation using the lifting scheme. The lifting scheme has two steps. First, the motion information of the image is found during the prediction process, and a high frequency band image is generated using the image. In the second step, a low frequency band image is generated by using a high frequency band image in the updating process. In this process, the energy of the image is transferred to the low frequency band as much as possible, and a small amount of energy is maintained in the high frequency band to increase the encoding efficiency. Inverse transformation allows the original to be restored without any additional information. However, the disadvantage of MCTF is that as the step increases, the motion information between frames cannot be found accurately.

2 shows a 5 / 3-tap wavelet MCTF encoding structure.

Referring to FIG. 2, the relationship between step and time is shown for a 5 / 3-tap wavelet MCTF.

t-L represents the progression of the MCTF, F0-F7 represents the frame to be encoded. In each step, the solid line represents the low frequency frame, the dotted line represents the high frequency frame, and the frames requiring motion information prediction are connected by arrows.

As the MCTF step progresses as shown in FIG. 2, the correlation between the images decreases as the time interval between the original image and the reference image increases, and correct motion information cannot be found due to an increase in the time interval between frames. This results in the energy to be transferred to the low frequency band remaining in the high frequency band, resulting in deterioration of image quality.

In addition, as the MCTF step increases, such deterioration of image quality accumulates, resulting in afterimages occurring in the time difference between frames. By omitting the update, this noise can be prevented from accumulating, but the PSNR is degraded due to the omission of the update. This is because the information of the high frequency band required for restoration is missing due to the omission of the update.

In addition, the time interval between the original image and the reference image caused by the increase of the MCTF makes it difficult to search the correct motion information, which leads to a deterioration in the coding efficiency and the image quality. When a new object appears in the image, the macro block of the portion cannot find the correct motion information, and as a result, generates a large value in the high frequency band image, and adds a large noise to the low frequency band image. In this case, a more fundamental problem solving method is needed, not a change in the update function.

In general, in the image processing field, JND (just noticeable difference) is used to determine the direction of high frequency components in an image. The characteristics of JND are used to determine noise visibility in the field of image watermarking and to insert additional data, or to image watermarking using JND masks and wavelets. The use of JND is a method using the directivity of the high frequency components in the image or using noise visibility.

The present invention relates to a method of preventing image quality deterioration due to inter-screen encoding of blocks in which motion information cannot be found using JND during an update process of MCTF. JND, which is used to find the edge of the image, is used as a threshold of high frequency band image which is added during the update process of MCTF by finding the parts with high visibility and those without errors. This method uses noise to be more pronounced in the low frequency region where there is no high frequency component in the image. Through the present invention, it is possible to overcome the time difference between frames in the lower stages of the MCTF and to improve visual quality.

According to an embodiment of the present invention, the update is prevented from occurring by using the noise visibility discrimination power of JND.

Applying updates by setting JND to a threshold of U can prevent multiple updates from occurring in areas that exceed the threshold. In addition, JND can be used to prevent such redundant updates because assuming that U added during the update is noise, the update is skipped due to the addition of noise.

The adaptive MCTF uses JND, whose size varies depending on the pixel value in the image, as a threshold value of U to limit the update and reduce noise according to the characteristics of the corresponding region in the image. On the contrary, by using the characteristics of JND, it is possible to find blocks having a large value in the high frequency band by inaccurate motion information, and to improve the image quality by intra coding. Adaptive MCTF shows very good visual quality improvement in blocks with U close to JND. However, when U is greater than or equal to JND, the size of H [i] added to L [i] is limited. Therefore, the image quality is degraded when the MCTF is restored, and the degradation of the quality increases as the difference between U and JND increases. U having a value greater than or equal to JND can be seen as not finding the correct motion information as the difference with JND increases, because it means that many values remain in the high frequency band. Therefore, in the case of the adaptive MCTF which reduces this value, it is difficult to see that the inter-picture encoding is performed by the accurate motion information.

In general, the noise present in the picture is more prominent in the low frequency part, and this information can be used to determine the threshold of U. The adaptive MCTF uses local variance to set thresholds during the update process to prevent noise from accumulating. The regional variance value JND used in the adaptive MCTF is used to determine the high frequency part of the luminance image in the human visual system (HVS). The JND equation maintains the maximum value of the regional variance in the image and calculates the local variance according to the characteristics of the image. In addition, by using the local dispersion JND, the U applied a threshold value is added to the pixels in the low frequency region, thereby reducing the noise.

σ _s ² (m, n) is the variance in (m, n) of the image. Equation 1 shows how much noise is noticeable at the current position, and since the JND value becomes smaller in the low frequency part, it can be used as a threshold of the update function.

In the present invention, a block having wrong motion information is selected by using JND used in the adaptive MCTF, and the image can be improved by intra coding the corresponding block. Since JND is a value calculated for every pixel, it is defined as follows so that the sum of JND in a block can be calculated for processing for each block.

x and y are the coordinates in the block, s is the coefficient representing the degree of application of JND, and B _x and B _y are the horizontal and vertical dimensions of the block. The V value is an allowable error range of the U value per pixel. If the value of V is 0, intra-block coding is performed for all blocks where U is greater than or equal to a threshold value. As V increases, the value approaches the adaptive MCTF. E _B is a value calculated as the sum of the allowable error ranges V per pixel (E _B = B _x B _y V), which is used to specify the range of U values to which the adaptive update through JND can be applied without change.

As described above, JND is used to prevent duplication of updates. Applying updates by setting the JND as a threshold can prevent multiple updates from occurring in areas that exceed the threshold. In addition, JND can be used to prevent such redundant updates because assuming that U added during the update is noise, the update is skipped due to the addition of noise.

One of the considerations in the update process of the MCTF using the motion information is the multiple update to the multi-referenced pixel. In the update process of generating a low frequency band using a high frequency band, the reverse direction of motion information used to generate a high frequency band is used, which causes repetitive updates to one pixel.

In order to overcome the shortcomings of the MCTF, the adaptive MCTF overcomes the shortcomings of the MCTF by using JND, which is a local variance value, as the update threshold. Covered pixel detection algorithms help prevent multiple updates to a pixel by selecting only one connected pixel to prevent duplicate updates.

According to another embodiment of the present invention, when updating an adaptive MCTF using a regional variance value, an optimal threshold value is applied through an RD theory to select blocks having inappropriate motion information and to switch to intra picture coding mode to improve image quality. Can be obtained. That is, in the present invention, the block having the wrong motion information is selected by using JND, and the optimum E _{B is obtained} to improve the image quality by encoding the block in the screen.

In the following, the adaptive MCTF is extended to apply JND, which is a local variance value, as a threshold for updating, and at the same time, a block having inappropriate motion information is selected and switched to an intra picture encoding mode to obtain an improvement in image quality. Let's look at it.

3 is a flowchart of an adaptive MCTF method using R-D theory according to a preferred embodiment of the present invention.

In step 300, motion information of the current block is retrieved.

In step 310, a high frequency band image is generated using the retrieved motion information.

During the updating of the MCTF, a high frequency band image is generated and a low frequency band image is generated using the same. The high frequency band image H [i] may be calculated as shown in Equation 5 below.

Here, H [i] is a high frequency band image generated during a frame prediction process, and P is a result predicted using motion compensation between adjacent frames. s represents image data according to a frame number, and MV _{2i + 1 → 2i} represents motion information from an odd frame 2i + 1 to an even frame 2i. As a result, MC (·) represents a frame predicted by using compensation through motion information.

Meanwhile, the low frequency band image L [i] may be calculated as in Equation 6 below.

L [i] represents a low frequency band image, and U represents extra data to be added to an even frame. The extra data is added to even frames using H [i] generated from Equation 1 and backward motion information.

Referring to the structure of FIG. 1, H [i] is generated first and L [i] is generated. In this case, if L [i] is generated, U is added to the data of the video to be encoded. In the lifting scheme, the lossless inverse transform is enabled by using the U added during the updating process. However, since U can be recognized as noise, it is possible to reduce the visible noise by applying a threshold value according to the difference between the current pixel and the surroundings during the update process.

In operation 320, a sum E _B of permissible error ranges per pixel for minimizing a Lagrangian cost function is calculated using a rate-distortion theory for the current block.

Due to the application of the fixed E _B , the image quality is deteriorated due to the error of the threshold value in the region where the high frequency component is low. This error in encoding mode switching occurs more often in flat areas where visual noise is more noticeable. To solve this problem, we apply the rate-distortion (RD) theory as follows.

R-D theory is a theory for calculating the minimum transmission bit rate R for transmitting the required picture quality when distortion D is present, which is a prior art and thus a detailed description thereof will be omitted.

To solve Equation 7, we need to find λ, but we can simplify the problem by using the fact that λ is the slope of the RD curve at the selected optimal point that satisfies the characteristics of the constraint and convex RD curve. Therefore, the non-limiting problem is explained as follows using a constant λ, and it is necessary to solve for E _B ^* which minimizes the Lagrange's cost function Equation 8 for a given λ> 0.

가 주어진 λ값에 관해서 R과 같은 경우가 발생하게 되면 E_B ^*는 한정 문제에 대한 원하는 해법이 된다. 다시 말해서, R이 주어지면 λ를 구할 수 있고 이에 따라서 비한정 문제에 관해서 해를 구할 수 있으며

을 만족시키는 원하는 해법 E_B ^*를 얻을 수 있다. Note that the optimal solution may not exist when λ> 0, depending on the RD characteristic. For all non-negative λ, it can be said that the solution of the finite problem is the same as the solution of the corresponding nonlimiting problem. But

If R occurs for a given lambda value, then E _B ^* is the desired solution to the limiting problem. In other words, given R, we can find λ and thus solve for non-limiting problems,

We can obtain the desired solution E _B ^* that satisfies.

Therefore, we need to search for λ as well as E _B to obtain a finite minimum cost function. If changes in the scene between the frames remain consistent, the value of λ between the frames may be considered to be insignificant. This implies that the change in coded bit rate in each frame is small between frames, so using the previous λ for the next frame instead of finding λ for each frame reduces the complexity of the computer processing. You can. Since [lambda] can be interpreted as a quality index of coded frames, it is important to find an optimal [lambda] from the start of coding the first frame until a screen change occurs. With regard to the macroblocks, the distortion of the j th frame may be extended as follows.

Therefore, the Lagrange cost function is changed as follows.

,

는 j번째 프레임,

는 j번째 복원된 프레임을 의미하며 D와 D^-1은 DCT와 IDCT, Q와 Q^-1은 각각 양자화 및 역 양자화를 의미한다.

,

Is the jth frame,

Is the j th reconstructed frame, D and D ^-1 are DCT and IDCT, and Q and Q ^-1 are quantization and inverse quantization, respectively.

At step 330 to obtain a local dispersion (JND) per pixel of the current block, and calculates the sum _(sum U) of the redundant data added to the sum of the current block in the region variance _(sum JND) and the even frame.

The _sum of extra data to be added to the even frame using at least one of the sum of the allowable error range E _B per pixel or the sum of the local variance in the current block (JND _sum ) in steps 340 and 350. _sum ) Hereinafter, the description will be based on the sum of the extra data to be added to the even frame, but in some cases it may be replaced by the sum of the extra data to be added to the odd frame.

Specifically, when the sum of the allowed error range per pixel and the sum of the local variances in the current block is smaller than the sum of the extra data to be added to the even frame in step 340, the process proceeds to step 360. If proceeds to step 350.

In the case of a block having U above JND and relatively close to the JND value, the image quality is improved by reducing the value without changing the mode. In one block, JND _s (m, n) S of all the pixels in the block are added and compared to the sum U _{sum of the} U in the block, and used for mode determination. If unable to find a correct motion information increases, so that U is added to the update to the JND E _B + E _B + _sum as a threshold value or more has the JND _{sum sum} U is assumed that the motion information of the block is incorrect.

The sum of the U _sum Most of my U U is greater than or JND or if there is a very large part of the U that is, if the average is greater than the threshold JND of the pixel values in the block, the block the block is greater than E + _{B sum} S JND. In this case, the update process is omitted and the encoding mode of the block is switched to intra picture encoding. E _B may be changed as S according to the change of the transmission environment of the image, and may play a role in helping to easily adapt to the transmission environment. The update function used for adaptive MCTF using local variance is defined as follows.

The proposed update function divides U into three ranges using JND and E _B and encodes them for each situation. If U has a size within the range of E _B + JND _sum S and JND _sum S, if the adaptive MCTF is applied and U exceeding a certain range, that is, U is greater than E _B + JND _sum S, inaccurate motion information It is regarded as applied and changes the coding mode of the block to the intra coding mode. A block converted to intra picture coding is encoded in the same form as an existing intra block. Blocks not converted to intra picture coding are adaptively added according to JND to generate a low frequency band image. The value of JND is adaptively changed according to the image, and the change factor is changed according to the visibility of noise, resulting in visual quality improvement of low frequency region. In addition, even in the case of an image including a very large motion, it is possible to contribute to the improvement of the image quality by selecting a portion in which the image quality deteriorates due to wrong motion information and changing the encoding mode.

In step 350, if the sum of the local variances in the current block is smaller than the sum of the extra data to be added to the even frame, the flow proceeds to step 370, and if large, the flow proceeds to step 380.

In step 360, the current block is encoded in the screen.

In step 370, the sum of the extra data to be added to the even frame is changed using the sum of the local variances in the current block.

In step 380, the sum of the extra data to be added to the even frame is maintained for the current block.

4 illustrates duplication of updates due to the prediction process of MCTF and backward motion information.

Looking at the prediction process of the MCTF, the frame s [2i] is predicted using the frame s [2i + 1], and H [i] is generated using the prediction. In the process of adding H [i] to s [2i] by using backward motion information during update, a shaded portion in which the update is duplicated occurs.

The covered pixel detection algorithm causes such an update to occur only once, and if there are a majority of unupdated pixels in a block, the block is encoded in the picture. However, the shielded pixel detection algorithm requires a separate operation process for each pixel to determine an encoding mode.

FIG. 5 is a graph of PSNR and bit amount indicated by applying a foreman sequence to JSVM for each MCTF. FIG. 6 is a graph of PSNRs and bit amounts indicated by applying football sequences to JSVM for each MCTF.

As can be seen from the graph, the MCTF according to the present invention showed an improvement in PSNR of about 0.5 to 0.8 dB compared to the conventional method, and the number of final bits also increased by about 80 to 800 bits compared to the reference method, which is about 250 bits on average. Appeared to increase. This results in better PSNR results at similar compression rates and visually better picture quality as PSNR enhancements focus on image quality improvements in low frequency parts where noise is noticeable.

Embodiments of the present invention may be implemented in the form of program instructions that can be executed by various computer means and recorded in a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the medium may be those specially designed and constructed for the present invention or may be available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks, such as floppy disks. Magneto-optical media, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

In the present invention as described above has been described by the specific embodiments, such as specific components and limited embodiments and drawings, but this is provided to help a more general understanding of the present invention, the present invention is not limited to the above embodiments. For those skilled in the art, various modifications and variations are possible from these descriptions. Accordingly, the spirit of the present invention should not be construed as being limited to the embodiments described, and all of the equivalents or equivalents of the claims, as well as the following claims, belong to the scope of the present invention .

Claims (8)

Retrieving motion information of the current block;
Generating a high frequency band image using the retrieved motion information;
Calculating a sum E _B of permissible error ranges per pixel to minimize a Lagrangian cost function using a rate-distortion theory for the current block;
Obtaining a regional variance (JND) for each pixel of the current block and calculating a sum (J _sum ) of local variances in the current block and extra data to be added to an even frame (U _sum );
Comparing at least one of the sum of allowed error ranges per pixel E _B or the sum of local variances JND _{sum in} the current block to a sum of extra data U _sum to be added to the even frames. step; And
According to the comparison result, the sum of the extra data to be added to the even frame for the current block is maintained as it is, or the sum of the extra data to be added to the even frame is changed using the sum of local variances in the current block, Adaptive MCTF method using the RD theory, characterized in that it comprises the step of intra picture coding for the current block. The method of claim 1,
In the RD theory, when the slope of the RD curve is λ, an RD theory is selected, wherein the sum of the allowed error ranges per pixel is minimized to minimize the Lagrangian cost function. Adaptive MCTF method using The method of claim 2,
The adaptive MCTF method using the RD theory characterized by reducing the complexity of the operation to minimize the Lagrangian cost function (λ) for the current frame as λ for the next frame. The method of claim 1,
If the sum of the sum of the allowable error ranges per pixel and the sum of the local variances in the current block is smaller than the sum of the extra data to be added to the even frame, the intra-block coding is performed on the current block. Adaptive MCTF Method Using Theory. The method of claim 1,
The sum of the allowed error range per pixel and the sum of the local variances in the current block is greater than the sum of the extra data to be added to the even frame, and the sum of the local variances in the current block is added to the even frame. If greater than the sum of the data, the adaptive MCTF method using the RD theory, characterized in that to maintain the sum of the extra data to be added to the even frame for the current block. The method of claim 1,
The sum of the allowed error range per pixel and the sum of the local variances in the current block is greater than the sum of the extra data to be added to the even frame, and the sum of the local variances in the current block is added to the even frame. If less than the sum of the data, the adaptive MCTF method using the RD theory, characterized in that for changing the sum of the extra data to be added to the even frame using the sum of the local variance in the current block. The method of claim 4, wherein
When the sum of the allowed error range per pixel and the sum of the local variance in the current block is smaller than the sum of the extra data to be added to the even frame, it is determined that the incorrect motion information is applied. Adaptive MCTF Method Using RD Theory. A computer-readable recording medium storing a program for causing a computer to execute the method according to any one of claims 1 to 7.

Images