FI109394B - Method and apparatus for encoding and decoding a video image - Google Patents

Description

109394

Method and apparatus for encoding and decoding a video image

Area

The invention relates to a method and a device for encoding a video image formed by sequential motion towers. The invention further relates to a method and apparatus for decoding a video image formed from sequential still images.

Background

Video coding and decoding are used to reduce the amount of data so that it can be stored more efficiently on a storage medium or transmitted over a communication link. An example of a video encoding standard is MPEG-4 (Moving Pictures Expert Group), which aims to stream video over a wireless channel in real time. This is a very ambitious goal because, for example, if the image to be transmitted has a cif size of 15 (288 x 352 pixels) and a transmission rate of 15 frames per second, 36.5 million bits should be compressed to 64 kilobits every second. The compression ratio would then be very high at 570: 1.

To move an image, the image is typically divided into image blocks, the size of which is selected to fit the system. The block information generally comprises 20 information about the brightness, color, and position of the block in the image itself. The data in the image blocks is compressed block by block using the desired encoding method. Compression is based on the removal of less relevant data. The compression methods are divided into three main categories: Spectral Redundancy Reduction, Spatial Redundancy I · * '25 (Spatial Redundancy Reduction), and Temporal Redundancy Reduction (Temporal). Redundancy Reduction). Typically, various combinations of these methods are used for compression.

For example, a YUV color scheme is used to reduce spectral redundancy. The YUV model takes advantage of the fact that the human eye is more sensitive to variations in luminance, or luminance, than to changes in chrominance, or color. The YUV model has one luminance component (Y) and two chromium nance components (U, V). Two chrominance components can be called:; : also for cb and cr components. For example, the H.263 video coding standard luminance block is 16 x 16 pixels and each chrominance block is: 35 which covers the same area as the luminance block, 8 x 8 pixels. Single Snow I> ·

t I

I · • · 2 109394 The combination of a nance block and two chrominance blocks is referred to in this standard as a macroblock. Macro blocks are usually read line by line. Each pixel, in both the luminance and chrominance blocks, can get a value between 0 and 255, meaning that eight pixels are required to represent one pixel. For example, a value of 0 for a luminance pixel is black and a value for 255 is white.

For example, Discrete Cosine Transform (DCT) is used to reduce space redundancy. In discrete cosine conversion, the pixel representation of an image block is converted to a space frequency representation. In addition, only the signal frequencies that occur therein have high amplitude coefficients, and those signals which do not occur in the block have coefficients close to zero. The discrete cosine transform is basically a lossless transform and the signal is only disturbed by quantization.

An attempt is made to reduce temporal redundancy by taking advantage of the fact that sequential images generally resemble each other, so that instead of compressing each individual image, motion data of the blocks is generated. The basic principle is: to find the best possible coded reference block for the block to be coded, the motion between the reference block and the block to be coded is modeled and the calculated motion vector coefficients are transmitted to the receiver. The difference between the coded block and the reference block is expressed as a Prediction error component (Prediction error frame). The motion vector Prediction can be used to determine the motion picture Prediction previously stored in memory (reference picture, reference frame). This type of coding is called in- •: ·:: 25 ter coding, which means the equivalent between images in the same sequence of images; ' · ": Utilization of opportunities.

: Since it is of interest to the present invention to reduce spatial-.Y 'dundancy, we will now examine it in greater detail. Macro! the block is transformed into a discrete cosine transform by the formula: ··. 2 / v \ (2x + 1) mtt {ly + l) v; r 30 F (u, v) = - C («) C (v) £ £ / (x, Y) cos -—- cosY—— - (1)

N 2W 2 N

Y ·: where x and y are the coordinates of the original block, u and v are the coordinates of the transform block, N = 8 and:: ': - \ =, for all w, v = 0 c («), c (v ) = V2. (2) '·; · '[L, otherwise 3,109,394

Next, Table 1 shows an example of how an 8x8 pixel block is converted by a discrete cosine transform. The upper part of the table contains the unmodified pixels, and the lower part of the table the result of the discrete cosine conversion, with the first element of 1303, 5 ns. dc coefficient, describes the average size of pixels in a block, and the other 63 elements, so-called. ac coefficients, which represent pixel scattering in a block.

As can be seen from the pixel values in Table 1, they have a large dispersion. In this case, the result after the discrete cosine conversion also contains many different coefficients of ac.

10 249 241 248 255 248 251 250 246 215 249 199 214 236 193 218 220 118 219 175 135 176 118 172 186 87 187 230 123 182 172 178 205 112 154 166 122 164 154 143 159 124 115 112 122 110 113 116 111 120 121 122 122 124 123 122 121 _120 121 119 120 120 120 120 118_ 1303 -18 -3 -43 -32 -54 -22 28 335 -2 5 -20 -6 -26 -26 11 100 29 6 45 44 49 -8 -30 41 9 -9 39 25 42 25 -21 ·: ·: 68-27-19 -9-24 -4 39 11 9 -18 -6 -23 -23 -1 18 9: -54 10 10 -6 2 10 -9 -5 • # «_-18 5 3 2 5 -8 -11 1_ {·

Table 1 / Table 2 describes a block with a small number of pixels between the pixels ·· * 15. As can be seen in Table 2, the ac coefficients get the same values,; that is, the block is compressed very efficiently.

• · · · 4 109394 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 115 _115 115 115 115 115 115 115 115_ 924 0 0 0 0 0 0 0 00000000 0 0000000 00000000 oooooooo 00000000 00000000 _00000000_

Table 2

Next, the discrete cosine transform block is "" quantized, "i.e. in principle each element of it is divided by some constant. This constant may vary between different macroblocks. In addition, for ac coefficients, a larger divisor is generally used than for dc coefficients. The 'quantization parameter' from which these divisors are calculated is in the range 1-31. The more zeros a block receives, the better the block will be compressed, since no zeros are transmitted to the channel. For the quantized. ···. 10 blocks can still be made irrelevant to the present invention. and encoding methods and eventually converting them into a bit stream which is transmitted to the decoder. Within the encoder, the quantized blocks are still inverted. , secondary quantization, and inverse discrete cosine transform, and thus a reference / reference image is formed from which blocks of the following images can be predicted. Thus, in the future, the encoder transmits the difference data between the incoming block and the reference blocks. Here's how:: ': Compression efficiency improves.

•> · · 1> »5 109394

Quantization is a problem with video coding, because the greater the amount of quantization you have to use, the more data you lose from the image and the result is unpleasant to watch.

After decoding the bit stream and decoding methods irrelevant to this invention, the decoder does basically the same thing as the encoder did when generating the reference image, i.e. the blocks do the same thing as the encoder, but in reverse.

Eventually, the compiled video image is displayed and the result depends largely on the quantization parameter used. Namely, if any element of a block goes to zero in quantization, inverse quantization is no longer able to restore it.

Short description

An object of the invention is to provide an improved method for encoding a video image and an improved device for encoding a video image. It is a further object of the invention to provide an improved method of decoding video image and an improved device for decoding video image.

In one aspect, the invention provides a method for encoding a video image according to claim 1. In one aspect, the invention provides a method for decoding a video image according to claim 6. In one aspect, the invention provides a device for encoding a video image according to claim 16. In one aspect of the invention, there is provided a device according to claim 21 for decoding a video image.

A: Other preferred embodiments of the invention are contemplated in the dependent claims.

»» ··. ·. The invention is based on pre-processing in the encoder before discrete cosine conversion, which can be restored by inverse processing in the decoder. Preprocessing increases compression efficiency as preprocessing reduces noise / spread in the image. As discussed above in Tables 1 and 2, the efficiency of the discrete cosine transform increases with decreasing block dispersion. The quantum | thus used however, the data transmission capacity required by the channel remains approximately the same. In practice, this means: '·, that at the same data rate, better quality recorded video is recorded / transmitted using the preprocessing and inverse processing according to the invention; · · 35 quality video images than without the invention.

6 109394

List of figures

Preferred embodiments of the invention will be described by way of example with reference to the accompanying drawings, in which: Figure 1 shows prior art devices for encoding and decoding a video image 5; Figure 2 illustrates novel devices for encoding and decoding video images; Figure 3 illustrates the preprocessing used for encoding a video image; Fig. 4 illustrates the inverse processing used in video image decoding; Figure 5 shows prior art pixels at the interfaces between parts of the devices illustrated in Figure 1; Figure 6 illustrates novel pixels at the interfaces between parts of the devices 15 shown in Figure 2; Fig. 7 is a flowchart illustrating a method for encoding a video image formed by a sequence of still images; Fig. 8 is a flowchart illustrating a method in succession! decoding a video image formed from still images; Figure 9 illustrates the effect of preprocessing on pixels.

DESCRIPTION OF EMBODIMENTS Referring to Fig. 1, prior art devices for encoding and decoding a television image are described. The face of a person 100 is captured by a camcorder 102. The video image is formed by the camera 102 of 25 consecutive still images, one of which is a still image 104. The camera 102 generates a pixel, ··, matrix, e.g. and chrominance have their own matrices. The data stream 106 representing the image 104 in pixels is then passed to the encoder 108. Of course, it is also possible to construct one; A device 30 for transmitting data stream 106 to encoder 108, for example, via a data link '· 1 * · or, for example, from a computer storage medium. The purpose here is that the- | the uncompressed video image 106 is compressed by the encoder 108, e.g. for retransmission or storage.

. ···. Since in the devices of interest we are interested in compression to reduce the state redundancy of '' '35, only the relevant parts of encoder 109 1094 and decoder 120 are described. The operation of the other parts will be apparent to one of ordinary skill in the art from reference to standards and textbooks, such as the following references: - ISO / IEC JTC 1 / SC 29 / WG 11: 'Generic coding of audio visual ob-5 jects - Part 2: Visual', pages 178, 179, 281.

- Vasudev Bhaskaran and Konstantinos Konstantinides: '' Image and Video Compressing Standards - Algorithms and Architectures, Second Edition '', Kluwer Academic Publishers 1997, Chapter 6: '' The MPEG Video Standards ''.

Encoder 108 comprises discrete cosine conversion means 110 performs discrete cosine conversion on pixels of each still image 104 as described above. The data stream 112 generated by the discrete cosine transform is fed to quantization means 114 for performing quantization at a selected quantization ratio. Still other coding may be performed on the quantized data stream 116, which is not described herein. The compressed video image 15 formed by encoder 108 is transmitted using channel 118 to decoder 120.

It is not described here how the channel 118 is implemented as the various embodiments will be apparent to those skilled in the art. For example, channel 118 may be a fixed or wireless communication link. Channel 118 may also be interpreted as a transmission path through which a video image is stored on a storage medium, such as a laser disk, and through which the video image is then read from a storage medium and processed by decoder 120.

The decoder 120 comprises inverse quantization means 122 for decoding the quantization performed in encoder 108. Unfortunately, inverse quantization is no longer able to return a block element whose value in quantization:: 25 went to zero.

; · The inverse quantized data stream 124 is then passed to inverse disc: chalk cosine conversion means 126 for performing inverse discrete. ···. cosine conversion for 104 pixels of each still image. The resulting data stream 128 is then passed through a possible other decoding process to a display 30 1 30 showing a video image of still images 104.

The encoder 108 and decoder 120 may be located in various devices, such as computers, subscriber terminals of various radio systems, such as mobile stations, or other devices in which video image processing is desired.

; Encoder 108 and decoder 120 may also be connected to the same device as. 35 can then be called a video codec.

8 109394

Figure 5 illustrates prior art pixels at the interfaces 106, 112, 116, 124, and 128 of the components of the devices illustrated in Figure 1 as a test image using the first 8x8 luminance-5 blocks of the first image calendar sequence "calendar_qcif.yuv". The interface 106 illustrates the content of the data stream after the camera 102. The interface 112 illustrates the content of the data stream after the discrete cosine conversion means 110. The interface 116 illustrates the content of the data stream after the quantization means 114. The quantization ratio used is 17.

For simplicity, other known encodings are not used, i.e., the data stream of interface 116 is transmitted along channel 118 to decoder 120. Interface 124 describes the content of the data stream after inverse quantization means 122. As can be seen in Figure 5 by comparing the original data stream 112 before quantization and the reconstructed data stream 124 after inverse quantization, as a result of the quantization, the zero values of the ac components shown at interface 116 can no longer be restored. In practice, this causes the original image 106 before decoding and the image reconstructed by the inverse discrete cosine conversion means 126 described at the interface 128 to no longer match. Thus, noise that degrades image quality has appeared in the reconstructed image.

Figure 2 illustrates novel devices for encoding and decoding video images. Since the devices of Figure 2 are based on the prior art devices shown in Figure 1, only parts differing from the devices described in Figure 1 will be described below.

.i,: In encoder 108, the discrete cine conversion means 110 is connected to the:: ··: 25 preprocessing means 200 to modify the value of each pixel of the still image 104 before performing the discrete cosine conversion. Esiprosessoin-; The means 200 comprise three different means, which are illustrated in Figure 3.

The first means establish a change in pixel value between two successive pixel values. One skilled in the art can choose the appropriate and effective way to perform this mathematical operation. In one embodiment, the preprocessing means 200 generates a change in pixel '·: value by subtracting the value of the latter pixel from the former pixel value.

Iin value.

Other means produce a weighted change using »ti ·, * ·. 35 value changes and a predetermined weight factor. One skilled in the art can select the appropriate and effective way to perform this mathematical 9,109,394 operation. In one embodiment, the preprocessing means 200 generates a weighted change by dividing the pixel value by a change in weight factor.

The third means sets the value of the latter pixel modified by the weighted change to the value of the previous pixel. One skilled in the art can choose the appropriate and effective way to perform this mathematical operation. In one embodiment, the preprocessing means 200 generates a weighted change using the previous pixel value modified by adding a weighted change to the previous pixel value.

In one embodiment, the preprocessing means 200 performs preprocessing with the following formula or a mathematically equivalent formula: i yi = yn + ((xi-yn) / k) (3) where y, is the last pixel preprocessed, y, .i is the preceding pixel. , Xj is the last pixel unprocessed, and k is the weighting factor.

The weight factor k is preferably a number greater than one. In one preferred embodiment in this embodiment, the weighting factor is two, whereby formula 3 takes the form yi = yi-i + ((Xi-yi-i) / 2) (4)

One mathematically equivalent (i.e., same result) 20 way to perform preprocessing is to use a moving weighted average for the calculation. The weighted average formula is n ί Σ »λ! .. * = - *; - ψ- (5)! Σ «, Σ“ ..,.: / = 1h »V. To keep the value of the last pixel in the preprocessed

t I

, between the values of a pixel and the value of the latter, unprocessed, apply

. ··. 25 in the formula for the weighting factor Σαι = M, and since here M

'* "' / = 1 is two, so ai + a2 = 2.

In the decoder 120, the inverse discrete cosine conversion means 126: are connected to the inverse processing means 214 to remove the preprocessing of each pixel of the still image 104 of the inverse discrete cosine conversion i 1 <, *. 30 after running. The reverse processing means 214 comprises three different means described in Figure 4.

»I

The first means generate a change in pixel value between two successive pixel values. A person skilled in the art can choose the appropriate and effective way to perform this mathematical operation. In one embodiment, the reverse processing means 214 generates a change in pixel value by subtracting the value of the former pixel from the value of the latter.

Other means produce a weighted change using a change in pixel value and a predetermined weighting factor used to generate the weighted change in video coding. One skilled in the art can choose the appropriate and effective way to perform this mathematical operation. In one embodiment, the inverse processing means 214 10 generates a weighted change by multiplying the change in pixel value by a single reduced weight factor.

The third means sets the latter pixel value modified by the weighted change to the latter. One skilled in the art will be able to choose the appropriate and effective way to accomplish this Mon-15 themed operation. In one embodiment, the reverse processing means 214 generates a weighted change using the weighted change of the latter pixel by adding a weighted change to the value of the latter pixel.

In one embodiment, the inverse processing means 214 performs the inverse processing with the following formula, or a therapeutically equivalent formula thereof, ma-20:

Xi = yi + (yi-yi-i) (k-i) (6) where Xj is the non-preprocessed last pixel, y, is the preprocessed-V. * Processed last pixel, yM is the previous pixel, and k is the weight factor.

·; ··: The weight factor k is preferably a larger number. In a preferred embodiment; * · *; In this embodiment, the weighting factor is two, whereby formula 6 takes the form

Xi = yi + (yi-yi-i) (7) ··· 'As already stated above, preprocessing or inverse processing can be performed on both the luminance and chromium nance data of a still image.

The preprocessing means 200 or the inverse processing means 214 process the still image 104 in rows, columns, macroblocks, blocks, or in any other predetermined non-random manner. The main point is just that both encoder 108 and decoder 120 are familiar with the ··· processing method used. In principle, preprocessing can also be performed twice, for example first in rows and then in columns.

i i · 11 109394

The preprocessing means 200, the discrete cosine conversion means 110, the quantization means 114, the inverse quantization means 122, the inverse discrete cosine conversion means 126 and the inverse processing means 214 can be implemented as a computer program running in the processor. Thus, the computer program includes routines for performing the steps of the methods described below. For the purpose of selling a computer program, it may be stored on a computer storage medium such as a Compact Disc Read Only Memory (CD-ROM). The computer program may be designed to operate also on a normal general purpose personal computer, laptop computer, computer network server, or other prior art computer.

Such means 200, 110, 114, 122, 126 and 214 may also be implemented as a hardware solution, for example as one or more Application Specific Integrated Circuits (ASICs), or as operating logic built from discrete components. In selecting the means of implementation, one skilled in the art will consider, for example, the processing power required and the manufacturing cost. Various software and hardware hybrid implementations are also possible.

Figure 6 illustrates novel pixels at the interfaces 202, 204, 206, 210, 212, 216 between the parts of the devices illustrated in Figure 2. The test lane uses the same luminance block as in Figure 5, i.e., the first 8x8 luminance of the test sequence. - *: ··: blocks. In the device of Figure 2, the content of the data stream 106 after the camera 102 is the same as that of the device of Figure 1; therefore, the data stream 106. is not re-depicted in Figure 6. The interface 202 illustrates the contents of the data stream. after processing means 200. The interface 204 illustrates the contents of the data stream. '' ··, after the discrete cine conversion means 110. Interface 206 illustrates the content of the data stream after quantization means 114. The quantization ratio used is 11.

. For simplicity, other known encodings are not used, i.e., the data stream of interface 206 is transmitted along channel 208 to decoder 120. Interface 210 describes the content of the data stream after inverse quantization means 122. Interface 212 describes the contents of the data stream in inverse discrete cinematic ···. after lifting means 126. The interface 214 illustrates a data stream after the reverse processing means 214.

12 109394

In Figure 5, the absolute value of the difference between each pixel difference of the original image block 106 from the camera and the reconstructed block 128 at the decoder 120 is obtained by calculating 677. This figure illustrates the noise introduced into the block during coding and decoding. Similarly, in FIG. -10 sessions. Thus, with the new encoder 108 and decoder 120 of FIG. 2, the quality of the reconstructed image is better because a lower quantization ratio could be used while the required data transmission capacity of channel 118, 208 would remain approximately the same, since the compression efficiency would remain nearly the same. In Figure 5, the compression ratio is about 1:25 and in Figure 6, the compression ratio is about 1:23.

Figure 9 illustrates the effect of preprocessing on pixels. From pixel 5, eight original pixels of the first row of the interface 106 and pixel 8 of the first line of the pixel preprocessed pixel of interface 620 are selected as pixels. The horizontal axis has a pixel number from 1 to 8, 20 and the vertical axis represents a pixel value from 0 to 120. The solid line represents the original pixels and the dashed line represents the preprocessed pixels. As shown in Figure 9, preprocessing moderates the variation in pixel values.

Applicant has conducted experiments using visual comparison of image: ·: Peak Signal Noise Ratio Estimator (PSNR) calculated using the formula: • '· 2552: .'-i psnr = lOlog ——-, ( 8) ::: where the image size is MxN, f (x, y) is the pixel of the original image and r (x, y) is the pixel of the image to be compared. In the comparison, psnr is calculated as the average of the p · ns for luminances' · "· and chrominances.

30 When comparing compression efficiency, the compressed video image size is simply used. Preprocessing is used as a weighting factor. · · \ Second and rounded to the nearest whole number.

Abbreviations used in the tables: - Stream name filename 13 109394 - Use of pre enc pre-processing in encoder ("+" = used, = not used) - Use of pre dec pre-processing in decoder - Size of compressed video image in bytes 5 - Quantization parameter used in QP - Psnr for decoded video psnr compared to original video psnr - Size rate compression ratio (Size unprocessed /

Size Processed) 10 - Psnr rate psnr ratio (Psnr unprocessed /

Psnr processed).

Let's first look at the sequence "calendar_qcif.yuv", which encodes the first 20 images.

Stream name Pre Pre size QP Psnr Size Psnr __enc Dec_____rate rate calendar qcif.yuv__ 448258 1 47.23 calendar qcif.yuv __ + __ 269639 1 29.98 0.6 1.58 calendar qcif.yuv __ + + 269639 1 40.29 0.6 1.17 15

Table 3

Using quantization parameter 1 and preprocessing only in code V, which is very similar to traditional preprocessing, results in a compression efficiency of · .. ·: 20 40%, but the image quality is significantly reduced (58%). If pre-processing; v: decoding is added to the decoder, the picture quality will only be reduced by 17%.

. · ·: Stream name Pre Pre Size QP Psnr Size Psnr __enc Dec_____rate rate "calendar qcif.yuv __-__ 16752 15 28.30 __- calendar qcif.yuv __ + __-__ 7941 15 25.92 0.47 1.09 calendar qcif.yuv __ + + 7941__15 26.73 0.47 1.06. ·. Calendar qcif.yuv __ + + 14378 10 28.11 0.86 1.01 * · *: 25 Table 4. ·;; After increasing the quantization to fifteen, it is noticed that the power of pre-processing in size, ···. Der already doubles, > * »14 109394 whereas the decoder loses its power compared to the data driven by the lower quantization The test data run by the quantization parameter 10 is essentially unprocessed: the power increases by about 15% while the image quality remains virtually unchanged.

5

Stream name Pre Pre Size QP Psnr Size Psn enc Dec rate r ________rate calendar qcif.yuv __-__ 4904 30 25.53 calendar qcif.yuv __ + ___ 4757 22 25.08 0.97 1.02 calendar qcif.yuv __ + + 4757 22 25.55 0.97 1.00

Table 5 10 When really high quantization is used, it is seen that the preprocessing power decreases. Now the same image quality is achieved with about 3% more compression.

Referring now to the flowchart of Figure 7, a method for encoding a video image formed from a sequence of still images is described. Execution of the method begins at block 700 where the encoder begins to receive the video image to be encoded. In block 702, the first unprocessed part of the first image is processed.

In block 704, counter i is set to one.

| ", In block 706, the first pixel preprocessed value (yi) 20 is set to the unprocessed value (xi).

•: In block 708, the value of counter i is incremented by one.

'·! Then, in block 710, the actual preprocessing is started, that is, before performing the discrete cosine conversion, it is repeated in turn: a preprocessing for each pixel of a still image, where: * '' 'and a predefined weighting factor, and:. - setting the latter pixel to the value of the previous pixel modified by weighted change 30.

. The preprocessing can be performed, for example, as described in block 710 • t # /. · ·. using formula 3 already described.

15 109394

In block 712, it is tested whether the portion underneath the image processing has already been completed, i.e., in our example, the value of counter i, increased by one, is greater than the last pixel sequence number of the portion of the picture. For example, if a cif-sized image is processed line by line, here we test whether 5 i + 1> 352. If the last pixel of the portion of the image has not yet been processed, then back to block 708, according to step 714, from which execution is continued, incrementing the value of counter i by one. If the last pixel of the image entered at block 710 discussed, let the direction of arrow 716 in accordance with block 718 in which it tests whether the whole image processing.

10, if all elements of the image has not yet been addressed, we go to block 720 as indicated by arrow 702, which is continued by reading the next part of the image esiprosessoimaton.

If parts of the image have been processed, we go from block 718 according to arrow 722 to block 724, which is carried out diskreettikosinimuunnos 15 724 unmodified pixels of a still image. The discrete cosine conversion may be performed in block 724, either for the entire image at a time, or then in accordance with the broken line 724 for each preprocessed portion of the image in turn before processing the next portion of the image in block 702.

Then, in block 726, it is tested whether the previous preprocessed image 20 was the last of the video sequence in question. If you did not, let's go according to arrow 728 to block 702, which will be continued by the processing of the first esiprosessoimaton the next part of the image processing. If videokuvasekvenssin.,:: Pre-processing was completed, we go from block 726 in accordance with an arrow 730: 1 'in block 732, where the embodiment of the method is terminated.

; The method may be modified by the following dependent patents; .· according to the requirements. Because their contents are described above at codec 108. · · ·. context, the explanation is no longer repeated here.

Finally, referring to FIG. 8, a method for decoding a video image formed from sequential still images is illustrated.

,. Execution of the method is started in block 800 where the decoder '; / ri starts receiving decoded video.

In block 802, an inverse discrete cosine conversion 802 is performed on the still image pixels.

Then, in block 804, the first pre-pro- portion of the image is processed.

1 In block 806, the initial value of counter i is set to one.

I t I

16 109394

In block 808, the preprocessed value (xi) of the first pixel is set to the preprocessed value (yi).

In block 810, the value of counter i is incremented by one.

Then, in block 812, the actual inverse-5 processing is started, i.e., after performing the inverse discrete cosine conversion, each pixel of the still image is repeatedly subjected to inverse processing, whereby: a predefined weighting factor used to generate the weighted change; and - setting the latter pixel to the value modified by the weighted change.

Reverse processing can be performed, for example, using the formula 6 already described in block 812.

| In block 814, it is tested whether the portion under processing of the image has already been completed, i.e., in our example, the value of counter i, increased by one, is greater than the last pixel sequence number of the portion of the image. If 20 is not yet processed the last pixels of the image in accordance with the go direction of the arrow 816 back to block 810, where execution continues with the increasing value of the counter i by one. If the last pixel of the image entered at block 812 discussed, let the direction of arrow 818 in accordance with a block 820 in which it tests whether the size of the Ku-Va processed.

f 25, if all elements of the image has not yet been addressed, we go from block 820 according to arrow 824 to block 804, where processing is continued by monitoring; 11; a raw preprocess part of an image. As shown in Fig. 8, here · 1 ·. da has an optional dashed block 802, i.e., the inverse discrete cosine transformation can be performed on all pixels of the image before inverse. 30 secondary processing, or alternately for each portion of the image to be reversed »» ·;,.: Secondary processing.

·; · 'If all the parts of the image have been processed, block 820 of the arrow go; 822 to block 826 for testing whether the previous preprocessed image was the last of the video sequence in question. If not, we go the direction of arrow j \ 35 828 according to block 802, where processing is continued by carrying out a reverse diskreettikosinimuunnos the next image pixel. If videokuvasekvens- 17 109394 käänteisesiprosessointi sequence was completed, we go from block 826 according to arrow 830 to block 832, where the method embodiment is terminated.

The method may be modified according to the appended dependent claims. Since their contents have been described above in connection with decoder 5 120, the description will not be repeated here.

; Fig. 7 does not illustrate the quantization performed in Fig. 7 after block 724, and likewise Fig. 8 does not depict the inverse quantization performed prior to block 802. It should be noted that the sequence of procedures of the methods need not be as described. For example, if the real-time requirement 10 is not high and the device has enough memory, the discrete cine transform according to block 724 may be performed at any one time on more than one picture stored in memory.

Although the invention has been described above with reference to the example of the accompanying drawings, it is clear that the invention is not limited thereto, but that it can be modified in many ways within the scope of the inventive idea set forth in the appended claims. 1

»» I

· ·

Claims (27)

A method of encoding a video image formed by one of the following still images, comprising: performing a discrete cosine conversion (724) for the non-converted pixels of the still image; characterized by decreasing before the discrete cosine transformation is performed, a pre-processing (710) is repeated in sequence for each pixel of the still image, where: a change of the pixel value between two consecutive pixels 10 is formed; a weighted change is formed using the pixel value change and a predetermined weighting number; and as the value of the latter pixel, the processed value of the preceding pixel is set using the weighted change. The method according to claim 1, characterized by the change of the pixel value being formed by subtracting the previous pixel value from the value of the latter pixel. Method according to any of the preceding claims, characterized in that the weighted change is formed by dividing the change of the pixel value by the weighting number. 4. A method according to any of the preceding claims, characterized by using the weighted change, the processed value is calculated by adding the weighted change by adding the weighted change to the value of the preceding pixel. according to any of the preceding claims, characterized in that the preprocessing is carried out with the following formula or with: ... · a mathematically equivalent formula: yi = yi-i + ((Xi-yi-i ) / k) where yj is a preprocessed later pixel, · 30 Yu is a preceding pixel, Xj is a non-preprocessed later pixel, k is a weighting number: 6. Method of decoding a video image formed by each other the following still images, comprising: performing an inverted discrete cosine conversion (802) for its still-transformed pixels; characterized by decreasing after the inverted discrete cosine conversion one executed, repeated in order of each pixel of the still image, an inverted preprocessing (812) where: a change of the pixel value between two consecutive pixels values is formed; a weighted change is formed using the pixel value change and a predetermined weighting number used to form the weighted change when encoding the video image; and 10 as the value of the latter pixel is set the processed value of the latter pixel using the weighted change. Method according to claim 6, characterized in that the change of the pixel value is formed by subtracting the previous pixel value from the value of the latter pixel. Method according to any of the preceding claims 6-7, characterized in that the weighted change is formed by multiplying the pixel value change by the weighting number reduced by one. Method according to any of the preceding claims 6-8, characterized in that by using the weighted change the processed value of the latter pixel is formed by adding the weighted change to the value of the latter pixel. A method according to any of the preceding claims 6-9; characterized in that the inverted preprocessing of the pixel is performed: by the following formula or by a mathematically equivalent formula: Xi = yi + (yryi-i) (k-i) k '. where Xj is a non-preprocessed later pixel, '/ · y, is a preprocessed later pixel; y, -i is a preceding pixel, / k is a weighting number. Method according to any of the preceding claims, characterized in that the weighting number is one tai greater than one. Method according to claim 11, characterized in that the weighting number is two. Method according to any of the preceding claims, characterized in that the preprocessing or inverted preprocessing of the pixel is performed in a still image in a row, column, macroblock, block or in some other predetermined non-random way. A method according to any of the preceding claims, characterized in that the preprocessing or inverted preprocessing of the pixel is performed on the luminance data of the still image. Method according to any of the preceding claims, characterized in that the preprocessing or inverted preprocessing of the pixel is performed on the chrominance data of the still image. An apparatus for encoding a video image formed by consecutive still images, comprising discrete cosine conversion means (110) for performing discrete cosine conversion at still image pixels (104); characterized in that the preprocessing means (200) are connected to the discrete cosine conversion means (110) to process each pixel value of the still image (104) before performing the discrete cosine conversion; the preprocessing means (200) comprises: means for forming a change in pixel value between two consecutive pixel values; means for forming a weighted change using the pixel value change and a predetermined weighting number; means to set as the value the latter pixel the processed value of the preceding pixel by using the weighted change. Apparatus according to claim 16, characterized in that the processing means (200) form the change of the pixel value by subtracting the previous pixel value from the value of the latter pixel. . Apparatus according to any of the preceding claims 16-17, characterized in that the preprocessing means (200) form the weighted; 'The change by dividing the pixel value change by the weighting number. | 19. Apparatus according to any one of the preceding claims 16-18, 30 characterized in that the preprocessing means (200), using the weighted change, form the processed value of the preceding pixels by ': adding the weighted change to the value of the preceding pixels. . Apparatus according to any one of the preceding claims 16-19, characterized in that the preprocessing means (200) perform the preprocessing of the following formula or with a mathematically equivalent formula: 26i = yi-i + ((Xi-yi) -i) / k) where y, is a preprocessed lateral pixel, yi-i is a preceding pixel, Xi is a non-preprocessed lateral pixel, 5k is a weighting number. An apparatus for decoding a video image formed by one of the following still images, comprising means (126) for inverted discrete cosine conversion to perform inverted discrete cosine conversion on the pixels of the still image (104); Characterized by inverted preprocessing means (214) connected to inverted discrete cosine conversion means (126) to remove the preprocessing of each pixel value of the still image (104) after performing the inverted discrete cosine conversion; The inverted preprocessing means (214) comprises: means for forming a pixel value change between two successive pixel values; means for forming a weighted change using the pixel value change and a predetermined weighting number used to form the weighted change when encoding the video image; means to set as the value the latter pixel the processed value of the latter pixel using the weighted change. Apparatus according to claim 21, characterized in that the inverted preprocessing means (214) form the change of the pixel value by subtracting the previous pixel value from the value of the latter pixel. . . Apparatus according to any of the preceding claims 21-22, characterized in that the inverted preprocessing means (214) are automated; 'makes the weighted change by multiplying the change in pixel value by | "weighting number decreased by one. Apparatus according to any of the preceding claims 21-23, characterized in that the inverted preprocessing means (214), by using the weighted change, form the processed value of the latter pixel by adding the weighted change to the value of the latter pixel. Apparatus according to any one of the preceding claims 21-24, characterized in that the inverted preprocessing means (214) perform the inverted preprocessing of the following formula or with a mathematically equivalent formula: Xi = yi + (yryi-i) (ki) where Xj is a non-preprocessed later pixel, y is a preprocessed later pixel, Ym is a preceding pixel, k is a weighting number. Apparatus according to any of the preceding claims 16-25, characterized in that the weighting number is one tai greater than one. 27. Apparatus according to claim 26, characterized by the weighted number being two. 28. Apparatus according to any of the preceding claims 16-27, characterized in that the preprocessing means (200) or the inverted preprocessing means (214) process the still image in a row, column, macro block, block or in some other predetermined non-random way. Apparatus according to any of the preceding claims 16-28, characterized in that the preprocessing means (200) or the inverted preprocessing means (214) process the luminance data of the still image. Apparatus according to any of the preceding claims 16-29, characterized in that the preprocessing means (200) or the inverted preprocessing means (214) process the chrominance data of the still image. I · * * ♦ t »*