KR101231332B1 - Method of and apparatus for processing image data for display by a display device - Google Patents

Abstract

A method of processing image data for display by a display panel of a display device is provided. Image pixel data representing an image is received. In a first processing step, the pixel data is processed to produce a multiview effect to the viewer. In the second processing step, for each of the plurality of subsets (kernels) of the pixel data, each subset (kernel) has the same number of pixel groups P _{x -1} , P _x , P _{x +1} , P _{x +2} . Each pixel group P _{x -1} , P _x , P _{x + 1} , P _{x +2} includes at least one pixel, and new pixel data P _x 'is included in the subset (kernel). Derivation (operation) of at least one of the pixel group Px of the subset (kernel) according to the pattern (pattern #) of the pixel data of the pixel group P _{x -1} , P _x , P _{x +1} , P _{x +2} )do. The derivation (operation) of the new pixel data P _x 'is done in such a way that it is possible to distinguish at least one such pattern from its inverse.

Description

Method and apparatus for processing image data for display by a display device {METHOD OF AND APPARATUS FOR PROCESSING IMAGE DATA FOR DISPLAY BY A DISPLAY DEVICE}

The present invention relates to a method and apparatus for processing image data for display by a display device.

Active matrix liquid crystal display devices are known which can switch between public and private display modes or direct multiple images to different viewers. In the first (public) mode, such displays behave in common as standard displays. A single picture is displayed by the device for the widest possible viewing angle range with optimum brightness, picture contrast and resolution for all viewers. In the second (personal) mode, the main image can only be recognized within the reduced range of the viewing angle, usually centered on being perpendicular to the display surface. The viewer associated with the display from outside this reduced angular range perceives a second masking picture that obscures the main picture or a deteriorated main picture to be unreadable.

In GB2413394 (Sharp), a switchable personal device is constructed by adding one or more extra liquid crystal layers and polarizers to the display panel. The inherent viewing angle dependence of this redundant element can be altered by electrically switching the liquid crystal in a known manner. Devices using this technology include Sharp's Sh851i and Sh902i mobile phones.

However, this kind of display can only selectively attenuate the brightness of the main image for an observer outside the reduced field of view and cannot display a reconstructible color video picture to the side viewer. Multiview displays based on parallax barrier technology with this capability are disclosed in US 2007/0296874 A1 and US7154653.

Both of these methods have disadvantages in that they need to add extra devices to the display to provide the ability to electrically switch the viewing angle range. This increases costs and increases the volume, which is very undesirable for mobile display applications such as mobile phones and laptop computers. A parallax barrier based display can also display only half of the display's pixels in each viewing area even in public mode, so that the effective image resolution is half of the base panel.

An example of a display with private mode performance without any display hardware complexity is Sharp's Sh702iS phone. This utilizes the manipulation of the image data displayed on the LCD of the mobile phone, in conjunction with the angular data-emitting characteristics inherent in the liquid crystal mode used for the display, and displays the display from a position off center of the personal mode where the displayed information is unreadable. Create to the viewer to observe. However, the image quality displayed to the viewer on the legitimate axis in the personal mode is somewhat lowered.

Multiview displays based solely on image processing techniques that can provide detailed reconfigurable subpictures in a personal mode that do not require additional optics on the display panel are described in GB2428152A1 and GB patent application 0804022.2. In such a display, the main image data is manipulated in a manner dependent on the second masking image, so that when the modified image data is displayed on the panel, the masking image is recognized by the off-axis viewer.

As described in GB patent application 0804022.2, the high spatial frequency modulation applied to the main image to produce a multiview effect is an undesirable color artifact that becomes apparent where any image characteristic is present in the main image. May cause). The application substantially reduces the appearance of these artifacts by applying an image processing filter as described in GB patent application 0701325.3 (published as GB-A-2445982) to the main image to blur the fine features in the image. It explains that it can be improved. However, standard image blurring filters, including those described for use in multiview displays, such as the above disclosure, are not designed to specifically address specific color artifacts associated with image data manipulation multiview processes. As a result, they show that for all picture types it is not effective enough to remove color artifacts with minimal blurring of the main image. Specifically, the filter cannot correct color artifacts around black text and white text with the same set of parameters.

Thus, with a substantially unchanged public mode, i.e., full brightness, resolution, contrast, viewing range, etc., for an equivalent standard display panel, the quality of the main image is produced by the multiview image process when viewed by the main viewer. It would be desirable to provide a multiview display with multiview or personal mode capabilities based on an image processing method that is optimized by including image processing steps to specifically correct the shape of color artifacts.

According to a first aspect of the present invention, there is provided a display image processing method for a display by a display panel of a display device, the method comprising: receiving image pixel data representing an image; Processing the pixel data to generate, and in the second processing step, for each of the plurality of subsets of the pixel data, each subset includes the same number of pixel groups, each pixel group including at least one pixel Inducing new pixel data for at least one of the subset of pixel groups in accordance with the pattern of pixel data in the subset of pixel groups, wherein inducing the new pixel data comprises at least one such pattern. It is done in a way that can be distinguished from the inverse.

The first processing step is a change in the on-axis emission that is not localized through the spatial averaging and not recognizable by the on-axis viewer, and the off-axis emission that can be perceived by the off-axis viewer because it is not locally balanced through the spatial averaging. This is done in consideration of the characteristics of the display panel to introduce a change.

According to a second aspect of the present invention, there is provided a method for processing image data for display by a display panel of a display device, the method comprising receiving image pixel data representing an image and, in a first processing step, locally through spatial averaging. A pixel that considers the characteristics of the display panel to introduce a change in on-axis emission that cannot be perceived by the on-axis viewer in balance, and a change in off-axis emission that can be perceived by the off-axis viewer because it is not locally balanced through spatial averaging. Processing data, and in a second processing step, for each of the plurality of subsets of pixel data, each subset comprising an equal number of pixel groups, each pixel group comprising at least one pixel; For at least one of the pixel groups of the subset according to the pattern of pixel data in the subset of pixel groups Deriving new pixel data.

In the first processing step, any increase in light emission introduced on the axis for one of the pair of pixel groups perceived by the viewer when having a single light emission through spatial averaging is applied to the other of the pair of pixel groups. Substantially matched by a substantially equivalent reduction in luminescence.

The final light emission of one of the pair of pixel groups may be close to the maximum light emission, or the final light emission of the pair of other pixel groups may be disposed near the minimum light emission.

Deriving the new pixel data may include selecting a filter according to a pattern of the pixel data, and applying a selected filter to at least a portion of the pixel group of the subset to derive the new pixel data.

The method may include comparing a pattern of pixel data with a plurality of predetermined patterns and inducing the new pixel data according to a result of the comparing step.

Each predetermined pattern is associated with a corresponding respective filter, and the method includes determining a matching pattern in the comparing step, and selecting the associated filter for use in deriving the new pixel data. can do.

The comparing step may include determining, for at least one pixel group or each of which new pixel data is derived, a luminance measurement of the pixel group for at least one immediately adjacent pixel group, wherein the measurement is performed with high luminance and Distinguish low luminance.

The method includes assigning, for each pixel group of the subset, the pixel group to one of a predetermined set of levels according to the pixel data, and in the comparing step, comparing the pattern of the assigned level with a predetermined pattern. It may include.

The method may include calculating a measurement based on pixel data of at least a portion of the pixel group of the subset, and assigning the pixel group to one of the predetermined levels may be performed according to the calculated measurement. .

According to a third aspect of the present invention, there is provided a display image processing method for a display by a display panel of a display device, the method comprising: receiving image pixel data representing an image; Processing the pixel data to generate, and in the second processing step, for each of the plurality of subsets of the pixel data, each subset includes the same number of pixel groups, each pixel group including at least one pixel Calculating a measurement based on pixel data of at least a portion of the pixel group of the subset, and for each pixel group of the subset, the pixel group according to the pixel data and depending on the calculated measurement Assigning to one, comparing a pattern of the assigned level with a plurality of predetermined patterns, and determining the comparison According to a step of inducing a new pixel data for at least one of the pixel groups of the subset.

The measurement may be an average value.

The method may include assigning the group of pixels to one of a predetermined set of levels based on a comparison between the pixel data and the calculated measurement.

The method may comprise, for example, converting the pixel data into apparent luminance values for use in the derivation step based on a gamma square law function.

The method may include using at least one conversion parameter determined according to a pattern of the pixel group of the subset.

The pixel group of the subset may be adjacent and extend substantially in one dimension.

The group of pixels in the subset may comprise adjacent two-dimensional arrays.

Each of the plurality of subsets may be spaced apart from one another by one pixel group.

The deriving of the new pixel data may utilize knowledge of the processing performed in the first processing step.

The new pixel data can be derived in a manner that takes into account a pattern of pixel data representing an image characteristic that is likely to cause artifacts as a result of the processing performed in the first processing step.

The deriving step determines whether at least one pixel group from which new pixel data is derived or each of them forms part of a pattern of pixel data representing an image characteristic that is likely to cause artifacts as a result of the processing performed in the first processing step. It may include the step.

The pattern of pixel data exhibiting bright features on a relatively dark background may be processed differently from the pattern of pixel data exhibiting dark features on a relatively bright background in the derivation step.

The deriving step may include determining from a pattern of pixel data whether at least one pixel group or each of which new pixel data is derived forms each of at least one of the following image features, wherein the next image feature Is a dark or tread line with a width of a single pixel group, a pixel group adjacent to a dark or light line with a width of a single pixel group, a left edge of a dark or light line with a width of two pixel groups, one or two pixels Dark-bright checkerboard patterns with group pitches and diagonal lines.

The second processing step may be performed before the first processing step.

The second processing step may be performed after the first processing step.

The new pixel data may be derived for a single pixel group of the subset of pixel groups.

Each pixel group may include a mixed color pixel group of color component pixels, and the image data processing method is applied to each color component pixel in turn.

The mixed color pixel group may include red, green, and blue color component pixels.

The first processing step may be done to interleave a set of received image data representing each different picture to present a different picture to the viewer at different respective viewing positions.

According to a fourth aspect of the invention, there is provided an apparatus configured to perform a method according to any one of the first to third aspects of the invention.

According to a fifth aspect of the invention, there is provided a display device comprising the device according to the fourth aspect of the invention.

According to a sixth aspect of the present invention, a program is provided for controlling an apparatus to perform a method according to any one of the first to third aspects of the present invention. The program may be included on a carrier medium. The carrier medium may be a storage medium or a transmission medium.

The characteristics of the display panel under consideration can be its signal voltage for the on-axis emission response, and the change in emission introduced on the axis is arranged so as to be locally balanced through the spatial average so that it is not recognizable by the on-axis viewer. The panel has a nonlinear off-axis emission versus on-axis emission relationship, so that the change introduced in the off-axis light emission is not locally balanced through the spatial mean, and thus is recognizable by the off-axis viewer.

Embodiments of the present invention provide an improved image processing method for privacy, and a display in the form of using the inherent angular visibility dependencies of the LCD panel and image data manipulation to enable multiple viewers to view multiple images. A multiview display is provided that allows maintaining high image quality on an axis without the introduction of unwanted image artifacts in the manipulation process.

In the first embodiment of the present invention, at least one of the plurality of image data sets input to the apparatus is processed with a filter to detect specific known image features that result in image artifacts when multiview image processing is performed. Thereafter, certain picture features are operated according to their shape in a manner that prevents artifacts from being introduced while preserving as much detail as possible of the original picture.

In the second embodiment of the present invention, the localized pixel group of the output image resulting from the multiview image process is compared with the pixel group corresponding to the same area of at least one of the input images. The discrepancy is then detected and the multiview output picture is modified in such a way that the discrepancy is corrected or diffused over a wider area of the picture so that the discrepancy is less perceived.

1 is a schematic diagram of a standard layout of control electronics for a liquid crystal display.
2 is a diagram of the data process flow of a standard single view LCD.
3 is a diagram of a data process flow of a multiview LCD of known type in a personal or multiview mode.
4 is a schematic diagram showing output pixel data from a known multiview display process operating on a uniform area of main and side input image data.
5 is a schematic diagram showing output pixel data from a known multiview display process operating on the diagonal area of the main input picture and the uniform area of the side input picture.
6 is a schematic diagram of an image data process flow of a known multiview display apparatus with additional preprocessing filters added for main and side input images.
FIG. 7 shows pixels in the 4x1 pixel sampling window of the main input image when " high " (1) or " low " (0) in the first (classification) step of the first embodiment. It is a figure explaining the preferable method of classification.
FIG. 8 is a table showing different operations performed in the third step in ten different cases identified in the second step of the first embodiment.
9 shows a possible hardware implementation of the first embodiment.
FIG. 10 is a diagram for explaining calculation used in the third (operation) step in the second example of the first embodiment.
FIG. 11 shows the results of the calculation of FIG. 10 acting on two input data sets (a) and (b) with different “gamma” parameters used to convert data values back and forth to nonlinear.
FIG. 12 is a diagram showing an example of each of six different cases identified in the second step of the third example of the first embodiment.
FIG. 13 is a diagram showing possible operations performed on input main image data in a third step in each case identified in the embodiment of FIG. 12.
14 shows a possible hardware implementation of the first embodiment.
FIG. 15 is a diagram showing an image data process flow in the LCD display device of the second embodiment. FIG.
FIG. 16 is a diagram showing calculations used in the second embodiment. FIG.
FIG. 17 illustrates the possibilities for the calculation of FIG. 16 to produce various results depending on the position of the kernel in the output multiview image, even if the input main image is uniform.

In a preferred embodiment, the display consists of a standard LCD display with modified control electronics. LCD displays typically consist of several component parts:

1. Backlighting unit that supplies flat wide-angle illumination to the panel.

2. Control electronics that output analog signal voltages for each pixel as well as timing pulses and common voltages for the counter electrodes of all pixels and receive digital image data. A schematic diagram of a standard layout of LCD control electronics is shown in FIG. 1 (Ernst Lauder, Liquid Crystal Display, Willy and Suns LTI, 2001).

3. An LC panel displaying an image by spatial light modulation consisting of two opposing glass substrates on which one of the arrays of pixel electrodes is disposed, and an active matrix array that directs the electronic signals received from the control electronics to the pixel electrodes. . A common common electrode and color filter array film is usually placed over the other substrate. Liquid crystal layers, usually 2-6 μm in thickness, which can be aligned between the glass substrates by the presence of an alignment layer on the inner surface of the glass substrate, are included. Glass substrates are generally disposed between crossed polarizing films and other optical compensation films to induce an electrically induced alignment within each pixel region of the LC layer and produce the desired light modulation of light from the backlight unit and the surroundings, thereby producing an image. .

An embodiment of the invention disclosed in GB patent application 0804022.2 is schematically represented in FIG. 2, operating in public display mode. In general, LCD control electronics (also referred to herein as control electronics) (1) are mainly composed of the electro-optical properties of the LC panel (2) and viewed from a direction perpendicular to the display surface (axial). A signal voltage dependent on the input image data is output in such a manner as to optimize the recognition quality of the displayed image for the viewer 3, that is, resolution, contrast, brightness, response time and the like. The relationship (gamma curve) between the input image data value for a given pixel and the observed emission from the display is determined by the combined effect of the data value on the signal voltage mapping of the display driver and the signal voltage on the luminous response of the LC panel. Is determined.

The LC panel 2 is generally composed of multiple LC domains and / or passive optical compensation films per pixel to preserve the display gamma curve as close to the axial response as possible for all viewing angles, thereby substantially equaling the wide field of view 5. Provide high quality images. However, the electro-optic response depends on the angle and the off-axis gamma curve is different from the on-axis gamma curve, which is inherent in liquid crystal displays. As long as this does not result in contrast reversal or large color shift or contrast reduction, this generally does not result in a clearly perceived error in the picture observed for the off axis viewer 4.

When the display is operating in public mode, a set of main image data 6 consisting of a single picture is typically input to the control electronics 1 in each frame period in the form of a series of bit streams. The control electronics then outputs a signal data voltage set to the LC panel 2. Each of these signal voltages is directed to the corresponding pixel electrode by the active matrix array of the LC panel and the final batch electrooptic reaction of the pixels in the LC layer produces an image. Substantially the same image is then recognized by the on-axis viewer 3 and off-axis viewer 4, and the display may be said to be operated in the wide field mode. This situation is shown in Fig. 2 and can be referred to as a standard method of operation for LCD.

In the multi-view display device of the type described in GB patent application 0804022.2 and schematically shown in FIG. 3, in the personal mode, two image data sets are input to the control electronics 1 every frame period, The main image data 7 constitutes a main image and the sub image data 8 constitutes a sub image 8. The control electronics then outputs a set of signal data voltages, one data voltage for each pixel in the LC panel as before. However, the control electronics (display controller) now output data voltages for each pixel of the LC panel constituting the expanded lookup table (LUT) and the combined image in both the main image (7) and the sub-picture (8). It depends on the data value for the corresponding pixel (in terms of spatial position of the image). The output data voltage for each pixel may depend on the third parameter determined by the spatial location of the pixel within the display.

The output voltage from the control electronics 1 then causes the LC panel 2 to display a combined image which is the main image when viewed by the main viewer 3. However, due to the different gamma curve characteristics of the LC panel for the off-axis viewer 4, this off-axis viewer perceives the sub-images very remarkably, which blurs and / or degrades the main image, and on display normal 9 Locks the main image information to the viewer within a limited cone of angles centered at. This situation is shown in FIG. Further details can be found with reference to GB Patent Application No. 0804022.2.

As described in the multiview process of GB patent application 0804022.2, the image data compression and combination process and parameters can clearly provide the on-axis viewer 3 with small color artifacts. This is specifically the case for the area of the input main image composed of a single pixel width diagonal. This is due to the fact that this process often results in an output image in which the alternating color subpixels are set to black, so that a single pixel wide black diagonal line superimposed on this pattern produces only one or two color subpixels along all lengths. Lines of pixels on either side can be left on. In this case, colored lines can be seen by the eye.

This problem is illustrated in FIGS. 4 and 5. FIG. 4 shows the results of a multiview image process that acts on the emission values of 2x2 groups of pixels from uniform main and sub-pictures in a typical RGB stripe display. From this, it can be seen that the multiview process outputs image data that is changed to the subpixel level, but maintains the overall light emission and color balance for the 2x2 group. The set of output pixels is then shown to the viewer at a distance sufficient to see only the average pixel emission for that group, which is the same as the input main image.

Figure 5 shows the same multiview process acting on a 2x2 group of pixels from a uniform input sub-picture, but the input main image area consists of dark diagonal lines. It can now be seen that the multiview process results in significantly different output image data in both total emission and color balance. This discrepancy can be seen by the on-axis observer, in which case the result is a strong green artifact appearance.

Depending on the position of the diagonal with respect to the pattern of the light and dark subpixels, the appearance of the image artifact may change. For example, the dark diagonal lines in the input main image of the opposite slope to those shown in FIG. 5 bring the overall magenta appearance to the output pixel subset. If the multiview image process adds luminescence to all color subpixels in some mixed white pixels, and is patterned on the subpixel level, rather than correspondingly from all color components of adjacent mixed white pixels as described in GB patent application 0804022.2. If the light is subtracted, the diagonal may disappear as a whole.

GB patent application No. 0804022.2 adds preprocessing steps 10 and 11, respectively, to act on input main and sub-picture data sets in a multi-view image combining process, which is composed of image blurring filters to prevent such image artifacts from occurring. Explain. This replacement for the display process is described in FIG. 6 of the accompanying drawings (corresponding to FIG. 17 of GB patent application 0804022.2).

In a preferred embodiment of the present invention, if an improved image preprocessing method is provided, this is achieved in the output image with knowledge of image data manipulation performed by the multiview image processing method, minimal changes to the input image, and effective use of computing resources. Use the image artifacts they generate to alter the input image data set in a manner that protects the appearance of the artifacts (e.g., processing requires less gate, less memory for buffered data, less latency, and less power consumption). One or more).

In a preferred manner, the input image pixel data is buffered by a plurality of pixels as received by the additional image processing apparatus 10 or 11, and a subset of adjacent pixel data values is displayed in the "window" or "kernel" in the image. To be sampled together. Each such window or kernel is considered to contain a subset of pixel data, where each subset contains the same number of pixel groups and each pixel group contains at least one pixel. For example, each pixel group may include a single pixel or may be a mixed color pixel group including red, green, and blue pixels. For simplicity, pixel groups are often referred to herein simply as pixels.

Next, a three step process is applied to each subset.

The first (classification) step is used to represent pixels in the subset as pixels of "high" or "low" value. This sorting step may also be regarded as the step of assigning a pixel to one of a predetermined set of levels based on luminance (in this case there are two, ie “high” and “low” levels). If the pixel value relates to a single monochrome or monochrome pixel, the pixel value for the adjacent pixel is processed. However, in the case of mixed color pixel groups, each of them includes red, green, and blue pixels, and this process is applied to each of the color component pixel values of the adjacent mixed group in turn.

In the second (comparison or case detection) step, the final high / low pattern is compared with a known set of patterns.

In the case of a match that is found, a third (operation) step is performed on at least one pixel of the subset and the sampling window then moves within the pixel to sample a new set of pixels. If no matching occurs, the pixel value does not change and the sampling window continues to move (if no match is found, it is assumed that a filter is simply applied that takes existing image pixel data and uses it as new pixel data, at least conceptually, without any change). Can be). Once the window has scanned the entire main input image and made any necessary changes to its image data values, the adjusted image data is output from the preprocessing circuitry to the multiview circuitry.

In an exemplary implementation, the sampling window or "kernel" is a pixel of a 4x1 horizontal block. The sorting step averages the data values of all the pixels, and averages the data values defined by those pixels and the lower values in the kernel with data values higher than the average, such as having a high value indicated by " 1 ", or " 0 ". Defining lower pixels. This process is described for the illustrated column of pixel values in FIG.

The 4x1 pattern of 0 and 1 values is then compared to a known set of patterns and an operation corresponding to the current pattern is performed on the second pixel from the left of the kernel P _x . The 4x1 window is then changed to the right by one pixel in the image, and the high / low classification, pattern comparison and operation are repeated on the second pixel. This process continues until the kernel scans the full width of every row in the image. The modified data value from the previous steps can be used in subsequent steps or the original data value can be used. For example, in the first step, the pixel 2 of the 4x1 window can be changed (second pixel from the left), and then the pixel becomes the pixel 1 in the 4x1 window in the second step. In the first embodiment the new value for the pixel 1 is used for calculation in the second step, but in the second embodiment the original value for the pixel 1 is used. In a video display, this process occurs in real time, filtering each frame of the video sequence within frame time and outputting the correct picture to the display.

As described above, when each pixel group is composed of individual color component pixels, this process is repeated for each color component individually. For example, the description of FIG. 7 may be considered to indicate each value, green value or blue value of the red color component of four adjacent pixels of a 4 × 1 window. In other embodiments, the window may take into account more than one color component and process the values therefrom, thus occupying more complex color patterns.

In an exemplary embodiment, of the 16 possible kernel patterns, 10 have found that an action is performed on the second pixel from the left to correct image characteristics that would otherwise result in artifacts in the multiview image. This pattern and corresponding operation is shown in FIG. 8. The pixel value used to calculate P _x ', which is the new value of the second pixel from the left side of the kernel, is the original data value of pixel P _x and its neighboring P _{x -1} and P _{x +1} , which is the first step in this process. Note that it is not a 0 or 1 value used to indicate a high / low value of, nor is the newly corrected pixel value from the previous step with the kernel at different locations. Embodiments of the present invention can distinguish each pattern from its inverse, for example, pattern 1 is distinguishable from pattern 2, which is the inverse of pattern 1, and pattern 3 is distinguishable from pattern 4, which is the inverse of pattern 3. You should know from the table. Different filters or actions are applied to each pattern and its corresponding inverse. The embodiment of the present invention can do this by looking at the pattern of the pixel data rather than the pattern of the absolute difference derived from the pixel data (as in the case of GB-A-2445982). However, not every pattern is required to be distinguishable from its inverse pattern, and there may be cases where the same filter or operation is applied to a particular pattern and its inverse, so it is not necessary to make a random distinction between the two in that situation. . It is also required to distinguish one particular pattern from the inverse (e.g., from 0010 to 1101), but the filter can only be applied to one of the two (e.g., 1101) and the other of the two ( For example, 0010 does not apply a filter, in which case it is only required to detect the presence of one of the patterns, not both. In this situation, the inverse pattern of pixel data can be considered to be combined to produce a substantially uniform image, where the bright is the inverse of the dark and vice versa.

The parameters used when calculating P _x '(BN, WN, BL, WL, BE, WE, WC and BC) can be assigned any value between 0 and 1, and these are the values of P _x when a specific action is taken. The amount of data shifted toward its neighbor is determined. In order to prevent any image artifacts appearing in the multiview image while minimizing the amount of change applied to the image, these parameters need to be set accurately according to the optical characteristics of the display (luminance, contrast, gamma curve, etc.). In order to maximize the efficiency of the filtering process, a parameter that is an integer fraction of 2 ⁿ , i.e. m / 16 or m / 32, is preferred, where n and m are positive integers, which allows the partitioning step to achieve simple bit-shift To be.

Advantages of this embodiment over prior art image blurring filters are that the pixels being operated are part of a dark or light single pixel wide line, or adjacent pixels are dark or light single pixel wide lines, the left edge of two pixel widths The multiple case detection steps allow the determination of whether is a dark or bright line or part of the two pixel pitch is a dark-bright checkerboard pattern. These are the main image features that cause artifacts to arise from the multiview image process, and each requires different actions to be performed on each case for optimal correction of the artifacts that appear in different cases. Specifically, white features may be treated differently than black features, and pixels adjacent to thin features may be subject to different amounts of variation in the pixels making up the feature itself. This allows the same filter to enhance the appearance of both white and black lines, and the ability to significantly increase the perceived quality of the image after the multiview process. For each pattern matching case, this is accomplished using different behaviors, each with individually tailored parameters.

Although this process has been described above as having a kernel run from left to right across the image on the left, it can be formed in a different way, ie run vertically, and a corresponding change to the pixel in the kernel having its data value It should be noted that at each step, the performance of the entire process is not changed.

Although it has been shown that the classification step gives effective results as described above, there are many other possible ways to assign the input image pixel data values to the 0/1 state for comparison in a particular case pattern that may be equally or more effective. It should also be noted that there is. Simple threshold data values may be applied, which classify pixels as low when their data values are below the threshold and classify pixels as high when their data values will exceed the threshold. Thresholds with bands can be applied, with data values in the bands that assign pixels to them neither high nor low. Pixels in the subset may be assigned higher or lower values according to their relative data values than their adjacent pixels. This method can also be treated as "extended" as the pixels are classified as neither high nor low if their difference from their adjacent pixels is less than a certain amount. Such methods, combinations of these methods, and other methods that perform the same classification function will also be included within the scope of the present invention.

For example, it is also possible to eliminate the separate classification step so that the case detection step operates directly on the image data, and the case detection step is, for example, the predetermined case detection having image data (rather than the preprocessed classification value) as an input. The algorithm determines what changes are necessary based on the image data itself (if necessary). The first and second steps can be considered when combined, with classification and case detection being performed as one, and in practice this step is to classify the pixel data into one of several categories, each with an associated filter. It is also possible to combine all three steps into one, and all that is needed is that in some way new pixel data is derived based on the luminance pattern of the pixels in the subset, which matters how many steps are used to achieve this. none. Another possibility is that the first step calculates one or more "feature indicators" from the pixel subset data, which feature indicators reflect one or more corresponding respective measurements associated with any known picture feature, for example such The feature indicator reflects how "checkered" pixels are in the appearance, while the pixels reflect high checkered appearance and low values reflect non-checkered appearance. This "feature indicator" is then used to determine which picture feature actually exists (or which feature is most dominant) in the second stage and which filter is most suitable for use in the third stage.

It is apparent to those skilled in the art that the image processing method including the embodiment described above can be implemented in software for operating a processing unit for providing data for a display device. The processor may also be implemented in application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other electronic circuits integrated into existing display control electronics. One advantage of the low computing resource implementation of the embodiment is that the required circuitry can be placed on the glass substrate of the display together with the pixel transistor device, saving money and making the best possible adoption for each display to be implemented.

An example of a circuit design for performing the required process is given in FIG. In FIG. 9, it can be seen that the process described above is implemented in the following manner. Pixel data values of an image corresponding to the current position of the kernel are stored in the buffer memory. This value is input to the high / low detector and to the average calculator for the first (classification) step and also to the parameter multiplier process for the third (operation) step. The output of the high / low detector is generated by a pattern matching selector to select any of the available parameters input from the register block, to apply to the pixel values from the buffer to calculate the correct output value in the third (operational) step. 2 (case detection) is used. The available parameters are shown in the diagram as WNO / 1, BNO / 1, WLO / 1, BLO / 1, WEO / 1, BEO / 1, WCO / 1 and BCO / 1, and the 0/1 symbol indicates that it is currently changed. It depends on whether the parameter is an application for the pixel P _x or if one of them is immediately adjacent (eg, to use the value given in FIG. 8, WNl is equal to 1-WNO).

Although FIG. 9 shows input image data as having a bit depth of 6 bits per color channel (ie, 18 bits per pixel), it should be noted that the functional process shown in the figure may be applied to an image of any bit depth. .

The implementation as shown also allows a simple threshold approach to be used in the first (classification) phase, with the possibility to enter BT and WT values to determine data level thresholds for high and low states. This value can be used in conjunction with the kernel mean value to define the band above and below the kernel mean value, and must be external to classify the pixel data value as high or low.

It should also be noted that, as described above, the process causes a significant increase in perceived image quality due to the multiview process and requires very moderate computing resources, while allowing for changes in the process that allow for outstanding performance with compensation of elevated computing resources. . As previously described, the diagonal of the input image is a feature that typically causes the most notable image artifacts in a multiview image. Therefore, it is desirable to use a processing method that can distinguish horizontal and vertical lines and diagonal lines, and thus apply the required correction while leaving the non-diagonal lines relatively unchanged to maintain large overall image clarity. This is not possible with a single line kernel, but if the area for the kernel to cover two or more rows is increased, and the number of specific patterns used for comparison and determining which operations are applied, and the number of available operations, then increase, Maintaining as described above basically this process is preferentially operating on the diagonal and thus can provide excellent picture quality in the output multiview picture. Thus, the shape of the image scan kernel utilized may be 4x2, 5x2, 3x3 pixels or any shape found to provide optimum performance within the scope of the present invention.

In a second exemplary implementation of the embodiment, a 4x3 pixel window is used to sample each pixel of the main image and the surrounding pixels of the first (classification) step. Again, several possible cases are compared with pixel placement in the window in the second (case detection) phase, and the 3x3 pixel-based blurring mechanism is nonlinear to the "pixel appearance" transform phase, such as the "gamma" power-law transform. Regarding the data value, it is utilized in the third (operation) step. Again, in the case of a mixed color pixel group, each of which includes red, green and blue pixels, the process is applied to each color component pixel value of the image in turn. If the pixel value is a single monochrome or associated with a monochrome pixel, the pixel value for eight adjacent pixels is processed. In this example, because the 4x3 window should provide a more reliable way to distinguish white text from black text, the 4x3 window is used in the first step instead of the 3x3 window, for example as used in the third step. However, it is understood that any size window can be used, both smaller being better for memory purposes.

The first (classification) step of this implementation is the same as the method previously described, but can be extended for use in a 4x3 pixel window, that is, the pixel of which the average pixel data value of the subset is calculated and has the above data value. Is displayed as a "1" pixel and a pixel having a pixel value below the average is displayed as a "0" pixel.

In the second (case detection) step, the larger function of the 3x3 pixel kernel needs to be detected and in this case specific to automatically apply a greater degree of change to the diagonal feature than to the horizontal or vertical feature in fewer individual cases. Operation is applied. It has been found that only three general cases require the detection and application of 3x3 pixel-based filter operation specific to each case. These can be shown with a light thin feature on a dark background, a dark thin feature on a light background, and a single pixel pitch checkered pattern.

A relatively simple method for determining two precisely shaped cases involves summing the number of "1" shaped pixels in a 4x3 window. If the twelve pixel window comprises six or more "1" shaped pixels, this case may be defined as a dark feature on a light background. If the twelve pixel window comprises a " 1 " type of less than six, this case can be defined as a bright feature on a dark background.

The case of the existing single pixel pitch checkered pattern is a 4x3 window with alternating pixel classification types, eg, three or more pixels above or below them with alternate alternating patterns, e.g., 010 or 101, respectively. Three of the four pixels of the central horizontal line of may be determined by the left hand.

In a third (operational) step, the new data value is based on the original data value and the eight pixels surrounding it (i.e. the leftmost nine of the 12 pixels of the 4x3 subset used in the second step) of the 4x3 pixel subset. It can be written to the second pixel from the left in the center column. The 3x3 pixel operation used may be as described in the publication of GB Patent Application No. 0701325.3. Briefly, the overall measurement of the appearance of each pixel is calculated based on the weighted sum of the data values of the main image and its eight immediately adjacent ones.

The second overall appearance value is then calculated for the same subset of pixels, but in the approximation of the effect of the multiview process of the pixel subset, the immediate neighbors are set to zero. The weighted sums are normalized and set to be equal to each other by calculation of new values for the central pixel data value, P ' _{(x, y)} , if necessary. This equivalent is shown in FIG. 10. The equation P ' _{(x, y)} that provides a new data value for the center pixel is as follows.

식1

Equation 1

Where ω ₁ , ω ₂ and ω ₃ are the weighting parameters used and P _{(x, y)} and the like are the data values of the pixel at each coordinate in the main image. As in the previous example, once the new value for the center pixel of the 3x3 window is calculated, this value is written to the output image ready for input into the multiview image process, and the entire three-step process is performed by scanning the entire image by the window. It is repeated for the next pixel of the input main image until it is processed.

This method can be optimized by the selection of the pixel weight parameter to automatically apply to the image a greater degree of change to the diagonal feature than the horizontal and vertical line features. This reduces the number of specific cases that need to be considered in the second (case detection) step, and the improved image appearance, as previously described, has three characteristics: dark background, dark features of white background, and single pixel pitch checkerboard pattern. It has been found that this is achieved by applying only three different 3x3 pixels based on the operation of the case.

In three cases, the weighting parameter of the equation 1 process assumes a 2-D Gaussian profile for the apparent luminance of the pixels within the 3x3 subset as a function of distance from its center, as described in GB patent application number 0701325.3. It can be determined by measuring the influence of the surrounding pixels on the center pixel based on this. For example, assuming a 2D Gaussian with a standard deviation of 0.5 pixel width, the weighting parameters ω ₁ = 0.0240, ω ₂ = 0.1070 and ω ₃ = 0.4759 are given.

This same weight parameter can be used in the third (operational) step, in both cases of the light features of the dark background and of the dark features of the white background. However, the degree of change, and the overall trend of the process to enhance the thin (single pixel) dark feature or the thin bright feature or the appearance of both, depends on the gamma square law function, e.g., L = (D / 255) ^γ 2) can be controlled by first converting the main image data into an apparent emission value, where L = apparent emission, D = data value (0-255) and γ (gamma) result in a 3x3 pixel based operation. The square law coefficient before processing by.

If the gamma value (conversion parameter) used for this conversion produces a data value for the luminescence relationship that accurately describes the data value for the luminescence brightness of the pixel on the display itself, the 3x3 pixel-based operation is coined as observed on the display. It preserves the overall luminescence of the dark and bright features of the image and creates a soft effect sufficient for fine image features to prevent the introduction of color artifacts by the multiview process.

If the gamma value used is lower than describing the data value in the light emission response of the display itself, the overall trend of the 3x3 pixel-based operation described in Equation 1 is to bold the appearance of dark features in the main image. This is due to dark pixels with bright neighbors that remain dark after processing with this process, but bright pixels with dark neighbors are substantially dark.

If the gamma value used is higher than describing the data value in the luminescence response of the display itself, the overall trend of the 3x3 pixel-based operation described in equation 1 is to bold the appearance of bright features in the main image. This is due to bright pixels with dark neighbors that remain bright after processing with this process, but dark pixels with bright neighbors become substantially bright.

If the second (case detection) step indicates that the 3x3 window consists of an image region containing a bright feature of a dark background, a high gamma value can be used for the filter operation, but the second (case detection) step is If the 3x3 window indicates that it consists of an image area containing thin dark features with a light background, low gamma values can be used for the filter operation, and if the appearance of both thin and bright features is bold by the 3x3 pixel-based filter process, The overall appearance of the main image after the view process is applied may be an improved case, achieving the effect required for both cases.

After the 3x3 pixel-based filter process is applied to the gamma-converted pixel data values, the final output values can be converted back to equivalent new data values using the inverse of Equation 3 below. In other words,

D '= (L') ^{(1 / γ)} x 255

Where D 'is the new equivalent data value, L' is the new apparent emission value after processing by the 3x3 pixel-based filter, and gamma is the square law coefficient as used in equation (2).

As described by Equation 1 in two historical main image regions including (a) a black diagonal line and (b) a white diagonal line, the effect of the 3x3 pixel-based filter operation is shown in FIG. It has a before and after conversion of pixel data values, each having a gamma value of 1 and 2.4. How the selection of gamma values affects how bright and dark lines change differently, and how to choose the gamma values used in the calculations according to the results of the previous case detection step to bold the appearance of the thin features of both bright and dark images. It can be seen from the figure why it is preferable.

In the case of the result of the second (case detection) step in which a single pixel pitch checker pattern is present in the region of the main image being sampled within that step, a simpler 3x3 pixel, as described by the following scheme: The base weighting operation may be performed by the uniform weighting assigned to the center pixel changed in the third (operation) step and the sum of the four immediately adjacent pixels.

식4

Equation 4

Again, the pixel data values P _{(x, y)} and the like can be converted by a nonlinear method such as the gamma dependent process described above before applying the above operation.

Thereby the second example removes the means of removing any unwanted color artifacts from the image resulting from the multiview process in a more accurate manner than the first example, which requires to buffer 4x3 blocks of pixel data values rather than 4x1 blocks. This is because both the memory to be processed and the processing power to perform the required pixel data changing operations reduce the amount of unnecessary changes to the input main image at the cost of an increasing demand for display computing resources.

In a third exemplary implementation of the first embodiment, the 4x3 pixel window can be used again for the 3x3 pixel window that is used again for the first (classification) and second (case detection) steps and the third operation. However, the number of cases detected in the second stage can be increased from 3 to 6 for the previous example, which greatly simplifies the operations performed in the third stage and reduces the need for display computing resources. This is to maintain comparable performance in terms of removing color artifacts from the final multiview image with minimal blurring effect on the image. The increase in the number of detected cases simplifies the blurring operation on all six cases to a standard normalized weighted sum of the 3x3 pixel subset, with all pixel weights being integer fractions of 32, at any point It does not require nonlinear conversion of pixel data values to apparent emission values.

In the first (classification) step, the implementation may be the same as the previously described example, that is, the average pixel data value for the subset is calculated and the pixel with the above data value is represented by " 1 " Pixels having the following data values are displayed as " 0 " pixels.

In the second (case detection) step, the three main cases of the previous example (checker) by the same method described earlier (comparison of the sum of the number of pixels of the form "1" in another case with the known pattern for the checkered pattern case) Pattern patterns, dark thin features on a light background, and bright thin features on a dark background), the classification of the center pixel of the 3x3 subset used in the third (operational) step is considered, and the number of specific cases is 3 to 6 To double. For example, the checkered pattern having the center pixel of the form "0" becomes a different case, and different specific operation steps are applied to the checkered pattern having the center pixel of the form "1". An example of six detected cases is shown in FIG. 12 (in the pixel classification "?" Indicates that it is not important when the pixel shape of this position of this subset is determined).

In the third (operation) step, the normalized weighted sum operation is used to calculate a new value for the center pixel of the 3x3 subset, and different weighting parameters are applied to each of six different cases identified by the second (case detection) step. Used. In each operation, the weighting parameter can be formed with an integer fraction of 32, which simplifies the calculation since no linear conversion of pixel data values is required. This process is shown in FIG. 13 with weighting parameters found to produce good results in the final multiview image for each case. As can be seen in the figure, each of the six operations is a correct normalized weighted sum, except in the case of " around white text " where negative weighting parameters are used for the corner pixels. This causes a large degree of change to be imposed on the case center pixel next to the diagonal, and it is required to remove color artifacts from this area in the final multiview image.

An example of circuit design for performing the required process is given in FIG. 14. In this figure, the above-described process is shown to be operated in the same manner as shown in Fig. 9, except that the pixel buffer and the operation step process are enlarged to incorporate 4x3 and 3x3 pixel windows, respectively. The parameters input from the register block are labeled A to F in the figure for the six different cases detected in the second step, and the suffixes 0, 1 and 2 indicate that this parameter is the central pixel, lateral in the third (operational) step. Indicates whether it is applicable to one of the perimeters or one of the diagonal perimeters. For example, in the case detailed in Fig. 13, this case is around the white text denoted by " F ", followed by F0 = 4, F1 = 9 and F2 = -2.

There is an image area that can cause any of the processes described above to in this case produce new pixel values that are outside the normal range (e.g. 0-255), and the new values outside this range are on the edge of this range. The fraction can simply be discarded to make it easier.

As an alternative method comprising negative weighted parameter settings, if it is not desirable for a particular implementation of this process, it is operated in addition to other methods of applying particularly many changes to the input image data of a particular case, e. The addition or reduction of a certain amount of data to the pixel value can be used, and the main pixel data for the 4x3 kernel can be used as a parameter in the third (operational) step, as calculated in the first (classification) step of the process, Allow the degree of change to be adjusted according to this average.

Although FIG. 14 shows input image data having a bit depth of 6 bits per color channel (ie, 18 bits per pixel), this figure is merely an example implementation, and the functional process shown in the figure is applied to an image of any bit depth. It should be understood that it can be applied.

The main purpose of the second (case detection) step is to distinguish between an image area composed of bright features on a dark background and the other area, and these different areas are applied to the third (operation) step to obtain an optimal result. The different parameters are known to be required, which is clear from the above example. The type of LCD used may have information available on the important examples of these cases, i.e. if the image contains black text on a white background, the information of this case may be of a "flag" contained in the input image data. It may be entered into the LCD as part or as part of a user-created display mode selection. In this case, this information can be entered into the case detection process to allow automatic selection of correction parameters and avoidance of the first classification step.

In the second embodiment of the present invention, the multiview process is performed on the main image 7 and the sub-image 8 dataset of the LCD control electronics 1 without making any changes to the input image. Next, any undesirable image artifacts from this process are compressed by the first step of the multiview process 12 in the additional image processing apparatus 13 and then the emission values of the output multiview images and the input main The image emission value is detected by comparison. The further image processing apparatus 13 performs some correction on the multiview image to remove the visibility of the artifact before outputting the corrected multiview image dataset to the display. The process for this operation is shown in FIG.

In the exemplary method of the second embodiment, as the multiview image and the original image 7 data are input to the additional image processing apparatus 12, a 3x3 kernel is sampled in each pixel and all eight immediately adjacent of both images. Used. If the pixel value relates to a single monochrome or monochrome pixel, the pixel value for eight adjacent pixels is processed. In the case of a mixed color pixel group, each of them includes red, green, and blue pixels, and this process is applied to each of the color component pixel values of the image in turn.

The normalized weighted sum "A" of the pixel data values in the kernel of each image is then calculated with a weight applied to each pixel of the kernel according to this position in a process similar to the process shown in FIG. This type of arrangement is described in FIG. 16 using three different weighting factors ω ₁ , ω ₂ , ω ₃ . If the "pixel appearance values" of these normally weighted, corresponding groups of pixels of each image are compared, and inconsistencies are found, the center pixel of the kernel in the multiview image is required to equalize the weighted sum result from each pixel kernel. It is changed by the amount. If A _mv is the normalized weighted kernel sum from the multiview picture and A _ic is the normalized weighted kernel sum from the compressed input main picture, then the new value for the center pixel of the kernel in the multiview picture, P _{(x, y ) Is} given by

식5

Equation 5

Where ω ₃ is the weighting factor for the center pixel of the kernel.

Once the new center pixel value is written to the multiview image, the kernel is shifted by one pixel, this calculation is repeated, and the new pixel value from the previous step is provided to the adjacent pixel value of the new kernel. In this way, corrections to the multiview image pixel values taking into account subsequent calculations and images are not overcorrected. The process is repeated until all the center pixels of the kernel have scanned all the pixels of the image.

In order to minimize changes to the output multiview image resulting from this process, it is desirable to minimize image artifacts in the output image, while not making any changes to the uniform area of the main image. Also for this reason, it is preferable that the weighted kernel sum is not changed stepwise for a uniform main image area due to the patterning imposed on the multiview image by the multiview process. An example of the uniform area of the output multiview image and the uniform area of the compressed main image emission value as a result of the process described in GB patent application number 0804022.2 is shown in FIG. 17. From this figure, it can be seen that the 3x3 kernel in step (n) 14 may have a different weighting in step (n + 1) 15 even if the original image is uniform. Accordingly, it is desirable to take a weighted sum for the pixels in the kernel that produces the same weighted kernel sum for the area of the multiview image corresponding to a uniform area in the input main image. Examples of such weights are ω ₁ = 1, ω ₂ = 2 and ω ₃ = 4.

Image processing that minimizes the degree of change, in particular the blurring of fine features imposed on the input image, while improving the perceived quality of the image displayed on the multiview display device by the removal of color artifacts from other case multiview processes. A method is provided. It is described as a scope of the embodiments to provide various options to balance the overall perceived quality of the output multiview image for the computing resources required to execute the process. It is to be understood that the embodiments provided herein include specific examples of processes in which producing good results for different resource specifications for a particular display device is found. The optimal process for any particular available resource specification will vary from application to application. Accordingly, the exact method of pixel classification used in the first step, the number and shape of the specific pixel patterns detected in the second step, or the exact calculation performed to generate the image data change in the third step of the first embodiment, Certain aspects of the present embodiments vary from embodiment to embodiment, and such changes are apparent to those skilled in the art when detailed specifications for specific applications are known and such changes are contemplated to be within the scope of the present invention.

Embodiments of the present invention provide image processing filters that are particularly useful for removing color artifacts from processes as described in GB-A-2428152 and GB Patent Application No. 0804022.2. GB-A-2445982 and its non-patent GB counterpart describe blur filters for use in dual view displays for processing color artifacts. However, embodiments of the present invention provide a first blurring filter specifically designed to correspond to color artifacts from the software privacy process, and consequently reduce the amount of blurring introduced into the main image.

However, the present invention is not intended to be limited to software privacy but is applicable to correction for artifacts introduced by any image processing step in which received pixel data is processed to provide a multiview effect to the viewer. For example, the present invention is applicable to the scenario described in GB-A-2445982, in which the processing step in consideration of the technique according to the embodiment of the present invention interleaves a set of received pixel data representing each different image. To provide different images to viewers of different respective viewing positions.

Two main embodiments are described above, the first of which is the main image being blurred only by itself before the multiview process is performed thereon, the second of which is multiviewed. The image is compared with the input main image and corrected according to the comparison.

Within the first embodiment, three examples are provided. All examples include three steps, namely classification of input data, pattern matching and then application of appropriate blurring operations. All three changes have a different balance between performance and implementation / resource requirements.

There is a difference between the method of implementing the present invention and the method described in GB-A-2445982 and its non-patent GB counterpart.

The case detection filter in GB-A-2445982 and its non-patent GB counterpart is unidirectional, that is, the blur only uses the pixel to its right side in accordance with the pixel value of what is being processed. Filters according to embodiments of the present invention are used to the right, left or both sides to determine new values.

The filter according to the embodiment of the present invention can handle all black lines, white lines, edges, and checker plates differently. GB-A-2445982 and its non-patent GB counterpart have only one specific case.

 GB-A-2445982 and its non-patent GB counterparts use a differential method of classifying the kernel pixels according to whether they are similar or very different from the pixels being processed. This is not sufficient for the embodiment of the present invention, since it is important that the peripheral pixels have a higher or lower value than the important pixel for the operation to be performed, rather than whether the peripheral pixels are just similar or different. In this regard, in the embodiment of the present invention, new pixel data is derived for the pixel in a distinguishable manner between at least one pattern of pixel data and its respective inverse pattern, but the technique of GB-A-2445982 does not. .

There are some differences between the second embodiment described herein and the second method of GB-A-2445982 and its non-patent GB counterpart.

This second embodiment uses the actual output multiview image for the comparison of the pixel appearance sum, not the assumed parallax barrier arrangement. The pixel is not always set to zero, other pixels have twice that value, and the pattern of high / low pixels is reversed with each movement of the kernel.

This second embodiment writes corrections to an image, performs a kernel, and then calculates a new weighted sum using the corrected pixel values from the previous step. In this way, correction is " done " as described in GB-A-2445982 and its non-patent GB counterpart.

It is understood that the operation of one or more of the aforementioned components can be controlled by a device or a program running on the device. Such an operating program may be stored on a computer readable medium or embodied in a signal such as, for example, a downloadable data signal provided from an Internet website. The appended claims should be construed as including the operating program by themselves or in the form of records, signals or any other form on a carrier.

For a complete understanding of the features and advantages of the present invention, reference should be made to the detailed description taken in conjunction with the accompanying drawings.

Thus, it is clear that the invention described can be modified in many ways in the same manner. Such changes are not to be regarded as a departure from the spirit and scope of the present invention, and all such changes which are apparent to those skilled in the art are intended to be included within the scope of the following claims.

Claims (35)

An image data processing method for display by a display panel of a display device,
Receiving image pixel data representing an image;
In a first processing step, processing the pixel data to generate a multiview effect for the viewer;
In a second processing step, for each of a plurality of subsets of the pixel data, each subset includes the same number of pixel groups, and each pixel group includes at least one pixel-pixels in the pixel groups of the subset Inducing new pixel data for at least one of the subset of pixel groups in accordance with a pattern of data;
Deriving the new pixel data is done in a manner that can distinguish at least one such pattern from its inverse. The method of claim 1, wherein the first processing step is performed by a local averaging through spatial averaging so as not to be recognized by the on-axis viewer, and by the off-axis viewer because of local aberration through spatial averaging. An image data processing method performed in consideration of the characteristics of a display panel in order to introduce a change in deaxial emission. An image data processing method for display by a display panel of a display device,
Receiving image pixel data representing an image;
In the first processing step, the change in the on-axis emission that is not locally recognized through the spatial averaging by the axial viewer and the out-of-axis emission that can be perceived by the off-axis viewer due to the local averaging through spatial averaging Processing the pixel data considering the characteristics of the display panel to introduce a change,
In a second processing step, for each of a plurality of subsets of the pixel data, each subset includes the same number of pixel groups, and each pixel group includes at least one pixel-pixels in the pixel groups of the subset Inducing new pixel data for at least one of the subset of pixel groups in accordance with a pattern of data. The method according to claim 2 or 3, wherein in the first processing step, any increase in the light emission introduced on the axis for one of the pair of pixel groups recognized by the viewer when having a single light emission through spatial averaging is And substantially matched by a substantially equivalent reduction in light emission for the other of said pair of pixel groups. The image data processing method according to claim 4, wherein the final light emission of one of the pair of pixel groups is close to the maximum light emission or the final light emission of the pair of other pixel groups is close to the minimum light emission. The method according to claim 1 or 3,
Deriving the new pixel data may include selecting a filter according to a pattern of pixel data;
Applying a selected filter to at least a portion of the subset of pixel groups to derive the new pixel data. The method according to claim 1 or 3,
Comparing the pattern of the pixel data with a plurality of predetermined patterns,
Deriving the new pixel data in accordance with a result of the comparing step. 8. The method of claim 7, wherein deriving new data comprises selecting a filter according to a pattern of pixel data, and applying the selected filter to at least a portion of the pixel group of the subset to derive the new pixel data. Including steps
Each predetermined pattern is associated with a corresponding respective filter,
Determining a matching pattern in said comparing step, and selecting the associated filter for use in deriving said new pixel data. 8. The method of claim 7, wherein the comparing step includes determining, for at least one pixel group or each, for which new pixel data is derived, the luminance measurement of the pixel group for at least one immediately adjacent pixel group. Is an image data processing method for distinguishing between high luminance and low luminance. The method of claim 7, wherein
For each pixel group of the subset, assigning the pixel group to one of a predetermined set of levels in accordance with the pixel data;
And in the comparing step, comparing the predetermined level pattern with a predetermined pattern. 11. The method of claim 10 including calculating a measurement based on pixel data of at least a portion of the pixel group of the subset,
Allocating the group of pixels to one of the predetermined levels is performed according to the calculated measurement. An image data processing method for display by a display panel of a display device,
Receiving image pixel data representing an image;
In a first processing step, processing the pixel data to generate a multiview effect for the viewer;
In a second processing step, for each of the plurality of subsets of pixel data, each subset includes an equal number of pixel groups, each pixel group including at least one pixel-at least a portion of the pixel groups of the subset Calculating a measurement based on pixel data of
For each pixel group of the subset assigning the pixel group to one of a predetermined set of levels depending on the pixel data and depending on the calculated measurement;
Comparing the pattern of the allocated level with a plurality of predetermined patterns;
Deriving new pixel data for at least one of the subset of pixel groups in accordance with the result of the comparing step. The image data processing method according to claim 12, wherein the measurement value is an average value. 13. The image data processing method according to claim 12, comprising assigning the pixel group to one of a predetermined set of levels based on a comparison between the pixel data and the calculated measurement. 13. The method of any one of claims 1, 3 and 12, comprising converting the pixel data into apparent luminance values for use in the derivation step based on a gamma square law function. An image data processing method. 16. The image data processing method according to claim 15, comprising using at least one conversion parameter determined according to a pattern of the pixel group of the subset. 13. The image data processing method according to any one of claims 1, 3, and 12, wherein said pixel group of said subset extends adjacent and substantially one-dimensional. 13. The image data processing method according to any one of claims 1, 3, and 12, wherein the subset of pixel groups includes an adjacent two-dimensional array. 13. The image data processing method according to any one of claims 1, 3, and 12, wherein each of the plurality of subsets is spaced apart from one another by one pixel group. 13. The image data processing method according to any one of claims 1, 3, and 12, wherein the step of deriving the new pixel data uses knowledge of the processing performed in the first processing step. The pixel data according to any one of claims 1, 3, and 12, wherein the new pixel data includes a pattern of pixel data representing an image characteristic that is likely to cause artifacts as a result of the processing performed in the first processing step. Image data processing method, which is derived in a manner to be considered. 13. The artefact as claimed in any one of claims 1, 3 and 12, wherein the derivation step comprises at least one pixel group in which new pixel data is derived or each of the artifacts as a result of the processing performed in the first processing step. Determining whether to form a part of a pattern of pixel data representing an image characteristic that is liable to cause an error. 13. A pattern according to any one of claims 1, 3 and 12, wherein the pattern of pixel data exhibiting bright features on a relatively dark background is characterized by the pattern of pixel data exhibiting dark features on a relatively bright background in the derivation step. An image data processing method which is processed differently. The method according to any one of claims 1, 3 and 12,
The deriving step includes determining from the pattern of pixel data whether at least one group of pixels from which new pixel data is derived or each of which forms part of at least one of the following picture features,
The following image features include a dark or tread line with a width of a single pixel group, a pixel group adjacent to a dark or light line with a width of a single pixel group, a left edge of a dark or light line with a width of two pixel groups, A dark-bright checkerboard pattern with a pitch of one or two pixel groups and diagonal lines. 13. The image data processing method according to any one of claims 1, 3, and 12, wherein the second processing step is performed before the first processing step. 13. The image data processing method according to any one of claims 1, 3, and 12, wherein the second processing step is performed after the first processing step. 13. The image data processing method according to any one of claims 1, 3 and 12, wherein the new pixel data is derived for a single pixel group of the subset of pixel groups. 13. The image data processing according to any one of claims 1, 3 and 12, wherein each pixel group includes a mixed color pixel group of color component pixels, and the image data processing method is applied to each of the color component pixels in turn. Way. 29. The image data processing method according to claim 28, wherein the mixed color pixel group includes red, green, and blue color component pixels. 13. The set of received image data according to any one of claims 1, 3 and 12, wherein the first processing step represents different respective images for presenting different images to the viewer at different respective visible positions. The image data processing method is performed so as to interleave. 13. An apparatus configured to carry out the method as claimed in any one of claims 1, 3 and 12. A display device comprising the device as claimed in claim 31. A computer-readable storage medium storing a program for controlling an apparatus for performing the method as claimed in any one of claims 1, 3 and 12.


delete delete

Images