FI98109B - Method for improvement of the quality of video compression - Google Patents

Description

98109

METHOD FOR IMPROVING THE QUALITY OF VIDEO PRINTING

The invention relates to a method for improving the quality of video printing as defined in the preamble of claim 1.

5 Analog video is an increasingly important image source.

In addition to video professionals, more and more amateurs are documenting the reality around us. In many situations, recorded images are desired in printed form - for publications, private use, to identify subjects in surveillance applications, etc. Despite the poor image quality, current analog video recordings are used due to low freezing costs and short production times and when no other image sources are available.

High-frequency consumer video technology (Super-VHS or Hi8) allows pre-press or print-house customers to use their own video equipment. This excludes advanced technologies such as 20 HDTVs. Despite the technical superiority of HDTV (resolution is about 5 times that of current video), current video technology will dominate for the next few years for both historical and economic reasons. Therefore, techniques for improving current video prints are also relevant from a general point of view.

Another technology excluded from this invention is professional broadcast recording devices, which • naturally give video prints a better quality than consumer S-video. On the other hand, the savings are lost, which is the main reason for using video printing technology. It is also questionable whether there is. market for such expensive video prints.

0 In this high price range, scanning cameras might be a more attractive choice due to their diverse resolution compared to video.

• · · '. Permanent video prints and video prints • · 2 98109 have its roots in image memories developed in the 1970s industry. These devices offered the ability to convert video images to digital format. But their uses were different, 5 thus mainly conversion from one video format to another. Dainippon Printing developed the first commercial video printing system - VPS - in 1982. The VPS system used exclusive hardware and transmission level 1 '' recorders. Later-10 min, several companies introduced cheaper video copiers for business and home use. These ready-to-deliver machines produce continuous-color prints that need to be resized and rasterized for publications. Video copiers do not allow for quality improvement or digital input to publishing systems. Modern video printing systems are an integral part of publishing systems that use superimposable image capture devices. There are a large number of image capture devices for both Macintosh and PC-20 computers. Hijackers are also available for UNIX workstations. The temporal processing of video signals with successive images has been studied extensively, for example for video compression. Desktop video is becoming more and more · * popular with image captureers that digitize, * 25 store and compress real-time video. But • ·; temporal processing is not aimed at improving quality.

«* ·· There are also special systems on the market for semi-automatic real-time enhancement of mobile video. These expensive systems employ techniques

30 such as adaptive comb filtering PAL / RGB

decoding, noise filtering as well as spatially. · R (2D and ID) that the temporal and inter-image noise-. median filtering of the peaks. But due to the lack of information from the video source a • »ί a priori, they do not apply any image degradation modeling to video coding and •« *. *. recording stages. The goal of these video systems is to produce moving video in an acceptable manner

IM

......--------- T _______ - _______________..........

9 98109 image quality - no photos on paper.

Image quality is improved by addition or restoration. In contrast, an increase in restoration requires a mathematical model of image degradation. These 5 models can compensate for varying degrees of deterioration.

If modeling is adequate, the restoration will naturally significantly improve quality compared to any ad hoc addition. In the present invention, the video encoding and recording model is the basis of the restoration. The model cat-10 covers two main sources of image degradation: PAL encoding of the YUV signal and S-video recording.

The PAL video signal consists of a luminance signal Y and a chrominance signal C, which is QAM-encoded for the subcarrier at the frequency fc. In the composite video, Y and C 15 are mixed together, while in the S-video, the signals are separate. The amplitude of C is: C = Usincot ± Vcostot, (1) 20 where the color difference signals U = 0.493 (BY) and V = 0.877 (RY, (o = 2nfc. In the following discussion, we omit the scales of U and V for the simplicity of the notation). The sign of V changes for each line in the field, • * :: because the phase of V changes 180 ° from line to line Y / C- * · When decoding 25 signals to YUV, the phase of V changes back and the resulting horizontal chrominance - the line is averaged with the line preceding the same field ·: * .This basic principle of PAL reduces • color tone errors caused by phase distortions, e.g., in transmission.This also reduces the difference between U and V • · · mixing, which increases when the upper • · * '· * * sideband is set to 5 MHz.This encoding causes a small man-made product, usually. · **. there is a certain hue shift every second between the horizontal lines in a uniform color field, • · • · * ί.ί · with a significant V component. Calculating the average also causes a small irregularity in the frequency * 1 98109 response.

Carrier frequency in PAL en: fc = 0.25 * 1135 * fLIME + 0.5 * fFIELD = 4.43361875 Mhz (2) 5

As a result of this frequency selection by the subcarrier, the successive lines in the field shift slightly more than 90 °. The displacement exceeding 90 ° is added together over a field of 180 °. This arrangement makes the close 10 lines of the two fields of the image in opposite phases. More specifically, line 1 of field 1 is opposite to line 1 of field 2, etc. This creates a chessboard pattern that is so typical of digitized color video images due to the remnants of the subcarrier in the digitized signal.

15 Because there are an odd number of lines (625) in the TV picture, a 90 ° offset between successive lines results in a 90 ° offset between successive pictures (offsets between fields compensate for each other).

Overall, the subcarrier frequency pattern needs 8 to 20 field (4 image) periods to repeat itself. As can be seen in Figure 1, the subcarriers in the figure are in opposite phases in every other figure (1 and 3). Furthermore, the sign of V is the same in opposite lines. The temporal processing developed below »relies on these observational works.

• The bandwidth of the 5 MHz PAL / G signal limits the luminance resolution (52 μ3 / 3 ^ ϋνίη € η line) * 5 Mjak soo / s = 260 cycles / line = 520 TV line. Correspondingly, the 1 MHz chrominance bandwidth limits the resolution to 52 ·. 30 episodes / line = 104 TV lines.

Standard consumer video (VHS, 8 mm) and enhanced S-video (S-VHS, Hi8 mm) are a compromise between recording time, I ·: size image quality and price. Much longer recording time, more compact size and lower price, 35 compared to professional recorders, have been achieved by 9 * I. * lowering the picture quality. This deterioration is below the model nett. The modeling is for 5-VHS, but with the smallest modifications of 5 98109, the Hi3 mm standard also applies. The recording portion of the camcorder is shown in Figure 2. Only aspects of the video restoration issue are included.

The S-camcorder CCD camera has one 5 sensor chip. In our deterioration model, the dimensions of the sensor matrix are M rows x N columns. Typical dimensions are M = 575, N = 750 CCD elements or pixels. The values of these pixels I (i, j) are the source of the imaging process. According to the Nyqvist criterion, the maximum 10 reproducible spatial frequencies are horizontally ti fH / N / 2 and vertically fv = M / 2. Higher frequencies cause spurious repetition interference and should therefore be eliminated by the lens system.

In the mosaic color coding used in camcorders, every other horizontal sensor line registers a periodic series of cyan (Cy) and yellow (Ye), and every other line registers magenta (Mg) and green (G). YUV pixel values are calculated over a 2x2 pixel group.

20 Y (i, j) = (1/4) (Cy + Ye + Mg + G) = (1/4) [I (i, j) +1 (i, j + 1) + I (i + 1 , j) +1 (i + 1, j + 1)] U (i, j) = (1/4) (Cy + Ye-Mg-G) = (1/4) [I (i, j) + 1 (i, j + 1) -25 I (i + 1, j) -I (i + 1, j + 1)] (3) V (i + 1, j) = (1/4) {Ye + G-Cy-Mg) = (1/1) [I (i + 1, j) +1 {it-1, j +1) -. Ki + 2, j) -I (i + 2, j + 1)]

30 i = 1.2, ..., M, j = 1,2, ..., N

It is easy to see that (3) produces false color defects: especially on the vertical edges of the image.

The U and V signals are calculated for every other »35 lines. Missing U and V lines are interpolated by copying the corresponding corresponding lines (using a single-line vii -. * Ve element). The maximum chrominance frequency fz received from the camera part 98109 / - * 3 is thus fCv = fv / 2, fCH = fH / 2. (4) 5 Y is calculated for each vertical and horizontal line, but the moving average of the two sensor elements in both directions attenuates the Y frequencies especially around the zero points fYH / 2 and fYV / 2 and the maximum of fYH and f ^ in below the hair. In addition, 10 cameras require optical low-pass filtering to reduce the color array Aree phenomenon when spatial frequencies higher than fCH are reversed back to the U and V basebands. Elimination of the Moaree phenomenon means that 25% of the Y-frequency band is taken out.

15 The maximum luminance frequency fL given by the camera is therefore lower than the Nyqvist frequency: fLV = M / (2k!), FLH = N / (2k1) (5) 20 where l <kt <2

Both Y and matrixed RGB components are gamma-corrected to correct for the nonlinear relationship between voltage and monitor light intensity according to the following equation: 2 5 γ'-γ7 "R '= R ™ (6) G' = B '= B» "m 30 A typical TV monitor is 2.2. Gamma correction increases pixel intensity in midtones.

‘J The C signal is generated after gamma correction and band limiting. Applying (1) and (3) we get for chrominance: 35 7 98109 C (i, j) = U (i, j) sin cot + V (i-2, j) cosot, i = even (7) C (i, j) = U (i-2, j) sincot-V (i, j) coswt, i = odd This means a decrease in the resolution of C in the vertical direction, which causes incorrect coloring at the edges.

The Y signal passes through vertical (VAPC) and horizontal aperture correction (HAPC), where a second derivative is added to it for edge enhancement. To prevent too radical overshoot, the Y signal is locked. Prior to modulation, high Y frequencies are amplified (under- and pre-emphasis filtering) to improve the S / N ratio of the recordings.

The Y signal is recorded in the FM modulated format with a deviation between 5.4 and 7 MHz (Figure 3). The side bands, together with the baseband, occupy the range from 1.2 MHz to 9 MHz. According to the rule of thumb, 80 TV lines / 1 MHz, the lower fundamental frequency limit (5.4 MHz) gives a resolution of more than 432 horizontal TV lines compared to 20 na for about 240 lines in composite video. The improved resolution is the result of accurate magnetization of the tape.

. However, noise makes it difficult to achieve this resolution in practice. The fLH measured with the video system used in the present invention was 380 TV lines = 190 divisions with 25 dB / line -6 dB attenuation or 73% PAL / G bandwidth.

··· With the recording chrominance, the subcarrier frequency is low • ·; *; nenee to 626, 953 kHz and the chrominance bandwidth is limited to 1 MHz. According to the above-mentioned thumb rule 30, the horizontal color resolution is then 80 • · 1 * 1 lines / MHz * 0.627 MHz = 50 TV lines. This limits the second • · • · *. line of the theoretical frequency fCH = 50 cycles. Jär- • · ·

· The fCH measured in our system was 32 cycles / line -6 dB

t ** ': with damping.

. *. 35 The AM-modulated recorded chrominance contains »· · *** / a considerable amount of frequency noise n, mainly due to the unstable ♦ · ** ··} head-to-tape contact. The S / N ratio for S-8 98109 VHS is standardized to be at least 48 dB compared to 43 dB in consumer video. This corresponds to 2.5 of the 256 average noises of grayscale. Thus, the chromium-nance can be divided into two parts 5 C (i, j) = Cf (i, j) + n (f) (8) where Cf is band-limited (f <f CH) and the noise term is assumed to be above the frequency range fCH.

10 Figure 3 shows that the AM-modulated chrominance occupies the frequency range below the luminance. Unlike VHS, S-VHS has no overlap between frequency bands. The modulated Y and C signals are frequency multiplexed and applied to recording heads of at least 3 on camera-15 recorders and between two and four on recorders. One track holds one field on the tape.

The heads are oriented slightly differently from the track - the difference is 12 °. This directional angle coding reduces the magnetic crosstalk between the strips 20 and the fields adjacent to the strip. Crosstalk occurs because consumer video recorders do not use guard bands. However, directional angle coding does not reduce chrominance crosstalk between fields due to the low chrominance frequency. The resulting interference can extend asymptotically over several lines.

In VHS and S-VHS, crosstalk in chrominance bands is reduced by the color phase rotation method (Figure 4). In a double-ended system, it works as follows. The phase of the chrominance lines of field 2 (= channel 2) shifts to -90 ° for each line. The phase shift is the opposite during playback. Lines i-2 and i of the same field are summed and the opposite phase crosstalk signals are canceled. For vertical color gradients. The compensation does not work perfectly because the opposite phase crosstalk signals differ in amplitude. Thus o 98109 C (i, j) = 0.5 [C (i-4, j) -C (i, j) j + d {Cfi-1, j), C (i ^ 1, j)} ( 9) where the function D {} describes uncompensated crosstalk from adjacent stripes i-1 and i + 1.

5 Due to the phase rotation system, uncompensated crosstalk has a periodic structure that repeats itself after 8 horizontal lines. Figure 4 shows that the period includes transitions (e.g., i + 2 m i + 3) from lines where the crosstalk and useful signal are in phase (saturated color), to lines where the crosstalk and useful signal are in opposite phases (gray color) . This manifests as a wrong coloring that varies from line to line.

During playback, the low-pass filter separates the Y and 15 C signals. The Y-s signal passes through a "double limiter", a drop-off compensator, a noise limiter, and a post-attenuation circuit, C is time-corrected for both frequencies and phase oscillations caused by tape speed variations 20 (vibration).

When C is decoded into U and V, the adjacent horizontal lines in the same field are averaged as described above: C (i, j) = 0.5 [C (i-2, j) + C (i, j )] (10)

Combining (8), (9) and (10) we get C '(i.j = -i.C / (i-21, j) + D {cf 0 - 1.J). C, (i + l.j)} + n (11) 4 i-o 30: Equation (ll) means that the original chrominance. on line i affects the recorded chrominance up to vii but i + 6 - a significant loss in vertical color resolution. Due to Equation (7), there is an * · 35 effect even at i + 8. In addition, there is a non-. * Compensated field crosstalk that extends beyond the 8 10 98109 line. The result is a visible downward movement of the color combined with a line-varying false coloring, especially at the horizontal edges, with saturated colors. In addition, saturation decreases whenever the signals of the 5-lane lines are averaged to give a lighter image.

Equation · '(11) can be thought of as representing the vertical average of a 7-dimensional filter and has a frequency response of 6 zeros at 2π / 7, 4π / 7, 10 ... 12π / 7. The pixels move down the image by seven pixels.

The object of the invention is, on the basis of the modeling of the previous pieces, to eliminate the above-mentioned disadvantages and to improve the quality of video printing.

15 In a method for improving the quality of video printing, a PAL video signal consists of a luminance signal Y and a chrominance signal C and an amplitude C = Usino) t + Vcostöt. According to the invention, the method comprises the following steps: - averaging four consecutive images in the PAL period with respect to Y, U and V, - smoothing the horizontal chrominance scvelt.a-; club low-pass filter to reduce noise at the chrominance line, at a frequency higher than 't 25 upper bandwidth limit fCH, ·: - vertical filtering of chrominance low-pass filters-: * Tinella, choosing improving horizontal chrominance resolution • · V. ' using more accurate luminance information, - increasing saturation by converting RGB to HL5 and ··· i 'back, and: - compensating for gamma correction by the camcorder.

35 Since the 4.43 MHz carrier modulates the chrominance * * * · * m * 1 *. ' the remnants of the subcarrier produce a known chessboard · * · ’pattern for the reproduced UV and RGB signal. Repeated 11 98109 tu Y also contains residues of the subcarrier. Since the Y and C signals are separate in S-video and not mixed with each other as in normal consumer video, this cross-luminance effect should not appear in 5 theories. Even if the Y and C signals are mixed together during tape-to-tape recording, interference should not occur because the frequency of the subcarrier is lower, so that C reserves the frequency range below Y without overlap.

10 That reduction of the applicant is not entirely successful in practice. Parts of the 4.43 MHz subcarrier are retained in the downconverted C signal, which is mixed with Y in the recording. When repeated, these subcarrier residues propagate to the Y signal because they are in the lower sideband of Y 15.

Disturbance is greatest with saturated colors. This is clear, bearing in mind that in an unrecorded PAL signal, the chrominance is higher than the luminance in all saturated base colors. For a blue similar to 20, the chrominance is up to 4 times higher. In the experimental arrangement, the amplitude of the residues in Y was 6% over the entire range of the similar blue signal.

By averaging the four ku-25 periods of the PAL period for only Y, U, and V, the subcarrier residues become smaller because the subcarriers of the opposite-phase image pairs (1,3) and (2,4} attenuate each other (Fig. 1).

Υ '= ΣΥ · (12) 30 for U and V, respectively, where t refers to the image number of the PAL sequence.

; Successful filtering requires either that the view be static over four images or that variable areas be omitted from the oois averaging. Only: 35 static views are considered in this invention.

«

Calculating averages also improves the S / N ratio, 98109 1. ίί.

but this improvement is insignificant compared to the reduction in the co-carrier.

Due to the low bandwidth of chrominance compared to luminance digitized horizontal chrominance lines U and V, too many samples are taken. Therefore, it is confident to apply a low-pass filter to the chrominance line to reduce noise, at a frequency higher than the upper bandwidth limit fCH (see (8)). Filtration also reduces the false parasitic coloring that occurs in (3).

A non-recursive finite impulse response (FIR) filter is applied over the 2K + 1 pixel convolution window: U '(ij) ~ Σ hn (k) U (i, j -k), (13)

i-C

15 similarly to V.

Because small vibrations are more important than a narrow transition range from the passband to the cut-off band, a rectangular window filter is not suitable. Instead, a Kaiser filter is selected with an adjustable tradeoff between the side beam plane and the transition width. The horizontal pulse response hH is calculated by multiplying the ideal low-pass frequency response by the infinite-length pulse response h-, with the window function w:: 25 hH0) = wWhj (j) ·> * · </) = -—--.- K <j <K 1141 io (a) •. = 0 elsewhere Ä (xV 1 30 where I0 (x) = l + Y - —- TI ^ 2J n!. · L. _ 13 98109 is a modified first-order and zero-order Bessel function. A and K are determined from the vibration d and of the variation width, i.e. the frequency response fs (cut-off band cut-off) - fp (pass-through cut-off) 5 In our video system, where 512 samples / line are sampled, the following design criterion applies: fCH = 32 cycles / line and ends at 10 f, = 64 cycles / line The Kaiser algorithm calculates a symmetric filter 17 with a constant 16-bit hH (k) with the frequency response described in Figure 5.

Figure 6 shows the application of the designed filter to the U-line. The filter clearly reduces noise 15 without significantly changing the shape of the curve.

Color drop and miscoloration, expressed in Equations (9) and (10), must be addressed to obtain acceptable image quality. Restoration based on the inversion of the deterioration pattern did not prove to be successful because the noise in the reproduced image caused a cumulatively increasing restoration error as the restoration progressed down the image. Instead, low-pass filtering data was effective. The first zero point of C, 2p / 7, was chosen as the cut-off frequency of the passband in fp 25 (II) to reduce the side beams modeled.

: Since fp = 575/7 = 82 cycles / line, fs was chosen to be 140 • cycles / line. The maximum vibration d is again armed. -27.96 dB, giving a Kaiser filter of 17 constant.

The filtered pixels are shifted 7 lines up in Figure 30.

* 8 U '(i-7J) = Yik. (K) U (i-kJ), (15) * - * *: respectively Ville.

This filter "lifts" the dropped colors and reduces the color change caused by the crosstalk of the stripe in '35 consecutive lines. The cancellation of the track crosstalk is obvious because the section of line 98109 + A * of phase cycle 8 fits in the filter window. However, high pass filtering reduces saturation, as mentioned above.

As mentioned above, the recorded chrominance horizontal resolution fCH is only 32 parts per 5 lines / line compared to 190 cycles per line for luminance.

This means that changes in chrominance are slow, causing colors to run at the vertical edges, which is especially noticeable when the luminance is low. However, the horizontal chrominance cell can be improved by using more accurate luminance information. Such Healing Methods have been in use for some time in circuits made for advanced analog and digital televisions. These circuits detect and sharpen chromium-15 nuance changes. The sharpened change occurs at the end of the original change period. The change in luminance is delayed so that it occurs at the same time as the change in sharpened chrominance. However, this method of edge shift to the right is typically 8 to 20 by 10 pixels, which distorts the geometry of the subjects.

In the present invention, a new method was developed (Figure 7): 1) The horizontal U and V equalization intervals on line i are observed from the sliding difference: 25,: IU (i, j + p) -U (i, j) I> T -> the equalization interval s starts from · j • · • · «·; IU (i, j + p) -U (i, j) I <T —► the equalization interval s ends at 30 j + p, respectively Ville • · • · ···. where p is, e.g., 8 pixels. T is the threshold, e.g., 15 • · "... 35 shades of gray out of 256.

• · · I *

· 2) The interval length is limited to a maximum of L

• · · »· • ♦ ··· 98109 pixels (eg 25 pixels).

3) The transmission intervals of successive lines touching each other are combined with the change range s.

4 4) Detection of the position ys (j) (inflection point) of the maximal first derivative of smoothed Y in the change-tar line on line j.

5 5) The locations of the turning points in the median are filtered in the change region s to avoid spike-like deviations of the sharpened edges. The kernel size is R pixels.

R / 2 y'sU) = medianiysUJ)} k.-R / 2 15 6) The new change interval is concentrated (y '(j) -1 ^ Y's (j) + 1) (similar to V).

Due to spatial chrominance low-pass filtering in both the camcorder functions 20 (Equations 8-11) and the restoration algorithms (Equations 12-14), saturation is significantly reduced, resulting in light colors. For this reason, satiety improves; by converting RGB to HLS and back. Typically. S increases by 20-40%.

. The final step in the enhancement according to the invention is the compensation of the gamma correction by the camcorder (equation (6)). RGB values pass through a series of tone curves • · ’· determined from (6) as well as * also by estimating the gamma gp of the printing process. gp is estimated by measuring the intensity of the black and white wedge of the printed sheet. The compensation is thus * * tm: 'R = R * ·· ym:: G * ··· • jm 5' = B * • · • * ·· * • · · ♦ * · 16 98109

The system of Figure 8 was formed to implement the method of the invention. The video was recorded on a Panasonic S-VHS camcorder NV-MS70E. Read items were imaged for 30-60 s. Thus, the video print operator had 5 to select the correct view to capture. The tape was performed on a Panasonic AG7330 S-VHS desktop recorder. The S-video signal then passed through the PAL / RGB converter VIDI / O BOX manufactured by Truevision. The Y signal circulated through the converter, yielding four YRGB outputs, of which we used 10 YRBs. The alternative to this would have been to digitize all four YRGB signals, but this would have required a four-channel capture device, which is more expensive. The G component was omitted because it is the most stable component to be mathematically calculated from 15 of the other three signals. The digitizing and processing card SIEPPO, developed jointly by VTT and TT Oy, was used in the system. The card works on a PC equipped with an AT bus. SIEPPO is based on a Texas Instruments graphics processor. As the card begins to 20 digitize the line from the sync pulse, it immediately recovers from the noisy line. Therefore, it behaves better than phase-locked image captors when the video signal quality is poor. SIEPON 3 Mb memory. allows consecutive 512 * 512 images to be stored in YRBt · 25 format. Digitization accuracy is 8 bits for sample and • · *. per color component. The measured A / D conversion was very ‘· ** linear, within ± 1%. The required sampling rate ♦ · · * · ti 2 * ίω = 380 samples / line is easily achieved as 512 samples are taken from each horizontal * · · il '30 line. In the vertical direction, the lowest 63 lines are omitted.

; Signal processing is applied according to the invention. Four consecutive images are captured and the time average of the YRB components is calculated on the capture * * ♦ * · * * 35 card. The PC calculates the color difference pixels U = B-Y and V = R-Y. U- ♦ · · * ... · and V-pixels are filtered horizontally and vertically, and color changes are sharpened in the horizontal ♦ »* • * • · • ♦ · 17 98109 horizontal direction. The image components Y, U and V are then converted back to YRB. The green value is then calculated according to the modified video equation: 5 G = (l / 0.587) (Y-0.239R-0.11B) (17)

Finally, the color saturation is increased, the gamma correction added to the camcorder is compensated, and the aspect ratio is changed to 4: 3. Now the RGB image is still ready for process-10 sounding, e.g. in a page layout program.

The results obtained with the video printing method according to the invention show a significant improvement in color quality. In particular, the subcarrier image and vertical color omission have been effectively removed.

15 The improvement was made possible by modeling the image degradation process in S-VHS camcorders and setting up a restoration system on this model. Because the chrominance component in the recording is more distorted than the luminance, most of the restoration is associated with 20 chrominance.

The invention clearly shows that video printing is a suitable technology for low quality products, such as some sales catalogs that contain small or medium-sized images.

»

Claims (7)

1. A method for improving the quality of a video printing, where a PAL video signal consists of a luminance signal Y and a chrominance signal C, where the amplitude of C is Usincot + Vcosot and the mean values are calculated from a PAL series four successive images in relation to Y, U and V, may be drawn therein, including the stages: - equalization of the horizontal chrominance by applying a low-pass filter to reduce the noise of the chrominance line, at a higher frequency than the upper bandwidth limit fCH, - vertical filtering of the chrominance with a low-pass filter, by selecting as the passband cut-off frequency fp 15 C's first zero point for reducing the side cones, - improving the horizontal chrominance resolution by using more accurate luminance information, saturation by transferring RGB to HLS 20 and back and offset by the camera tape recorder ·. gamma correction. A method according to claim 1, characterized in that the view is static for one: 25 times of four images. 3. A method according to claim 1, characterized in that the changed areas are left out of the calculation of the mean. 4. A method according to claim 1, characterized in that the improvement of the horizontal chrominance resolution contains the following stages: observing horizontal U and V stabilization intervals in applicable differences, 35. limiting the intervals to a certain amount of pixels. - joining of the conversion intervals from adjacent contact lines to a conversion region, - observing a maximum equilibrium Y's maximum derivative in a line's conversion interval, 5. median filtration of the inflexion point blending diverting points for avoidance points. deviation, and - centering on the new conversion intervals. 5. A method according to claim 1, characterized in that gP is estimated by measuring the intensity of the black and white wedge of the printed sheet. 6. A method according to claim 1, characterized in that, instead of a rectangular window filter, a Kaiser filter with a tradeoff between the sidewall plane and the conversion width is selected instead of a rectangular window filter. 7. A method according to claim 1, characterized in that, in the leveling of the vertical chrominance, instead of a right-angled window filter, a Kaiser filter with adjustable gear is selected; a tradeoff between the sidewall plane and the conversion width. }