WO2004017286A2

WO2004017286A2 - Video circuit

Info

Publication number: WO2004017286A2
Application number: PCT/IB2003/003312
Authority: WO
Inventors: Rob A. Beuker; Petrus M. De Greef; Robert Koffrie
Original assignee: Koninklijke Philips Electronics N.V.
Priority date: 2002-08-19
Filing date: 2003-07-24
Publication date: 2004-02-26
Also published as: WO2004017286A3; AU2003249530A8; AU2003249530A1; EP1563479A2

Abstract

The invention relates to a video circuit (1) for processing video signals which show images on a display panel with linear light transition, comprising a gamma correction circuit (3), a quantizer (11) and a sub-field generator circuit (8). To save computing time, a memory (7) replaces the quantizer (11).

Description

Video circuit

The invention relates to a video circuit for processing video signals which show images on a display panel with linear luminance transmission, comprising a gamma correction circuit, a quantizer and a sub-field generator circuit.

From US-PS 6,097,368 is known that video signals for showing images on a display panel of a television set comprise a red, a green and a blue signal, which is 3 times 8 bits of video data. Plasma display panel or PDP for short have a linear luminance transmission. Therefore, the video signals subjected to a gamma function are to be corrected and the video signals converted into luminance signals. Plasma display panels are limited as regards the number of luminance levels that can be displayed, a quantization process therefore reduces the number of levels. For generating red, green or blue light for a pixel, sub-fields are addressed which make a red, green or blue light source of the pixel light up for a definite period. This technique is also referred to as sub-field generation or SFG for short. The conversion of video signals into luminance signals, the quantization and the sub-fields generation require processing power. Therefore it is an object of the invention to improve the picture quality as well as to save processing power and electronics hardware.

This object is achieved in accordance with the characteristic features of the coordinated claims 1-3.

In a first embodiment, in a first memory a coarse adjustment of the quantization is effected and in a second memory a fine adjustment. Time is saved with the quantization effected in a memory. This means, that a look-up table is stored in the memory which appoint to input signal a quantised signal at the output. The splitting up into two parts provides that the necessary memory size for a 12-bit input is significantly reduced.

In a second embodiment most significant bits, MSB for short, are quantized in a first memory and least significant bits, LSB for short, are quantized in a second memory. Time is saved with the quantization effected in a memory. This means, that the memories comprise look-up tables. In a first look-up table a coarse adjustment of the quantization is effected and in a second look-up table a fine adjustment. The splitting-up into two parts significantly reduces the necessary memory size for a 12-bit input. In a third embodiment a memory replaces the quantizer. Digital data signals are applied to a memory as addresses and associated values are issued from an output. A lookup table in the memory is used to generate the quantization-data and quantization-error. This saves time compared to a computer that carries out calculations in a plurality of steps. The memory advantageously replaces a dequantizer. The formerly quantized signal is reconverted and a comparison with the input values can be made. Quantizers and dequantizers are realized in a memory. A quantization error can be detected and is used for an error diffusion method by means of a feedback loop and a filter.

The memory advantageously replaces a gamma correction circuit. If a gamma correction function and a quantization are carried out in a single memory, the computing time is also saved. Based on the linear light transition the video signals subjected to a gamma function are to be corrected and video data are converted into luminance data. A corresponding gamma correction function reads x = y" with n = 2.4. In order to achieve a sufficiently high resolution for dark areas, at least 3 times 12 bits are to be used. A look-up table, LUT for short, can be used to implement a gamma correction circuit. The LUT is stored in the memory. If a gamma correction function and a quantization are carried out in a single LUT, less memory is required. Plasma display panels are limited as to the number of luminance stages that can be displayed which are typically 32 (= 2⁵) to 256 (= 2⁸) discrete levels. A quantization process reduces the number of levels. An error distribution method reduces the occurring quantization noise. The quantization process and the error diffusion method are also referred to as dithering.

An inverse gamma correction circuit is advantageously included downstream of the dequantizer. An inverse gamma correction circuit is required to close the feedback- loop in the right domain. The quantizer, the dequantizer, the gamma correction circuit and the inverse gamma correction circuit are then realized in a single look-up table, which is stored in a memory.

The memory advantageously replaces a sub-field generator. If the gamma quantizer and sub-field generator are collectively realized in a single memory, computer time is also saved. A single look-up table is used to implement a gamma quantizer and a subfield generator.

Advantageously the memory supplies the filter with values. The filter is used for the error diffusion method.

A gamma function, a quantization, a sub-field generation circuit and a partial line doubling are advantageously achieved by means of two memories, with other words by means of two look-up tables. Least significant bits of sub-fields of two neighboring lines are identical and addressing time is saved then. The sub-field generator circuit of a first memory outputs a bit sample with which a plasma display panel can be driven directly and furthermore outputs data via a converter, a quantizer and a filter to the input signal of the neighboring line and via a second converter, a second dequantizer also to the input signal. With sub-fields that are spread non-equidistantly by the bit sample of the least significant bits is output directly to the quantizer of the second memory.

Advantageously the memory is a read-only memory, ROM in short. For changing settings like brightness and luminance two electrically erasable programmable read- only memories, EEPROM in short, are arranged parallel. Also two random access memories, RAM for short, can be used, initialized by a processor.

Advantageously coefficients of the filter are changeable. If a quantisation in combination with spatial error diffusion is done, spatial error diffusion patterns are visible. A problems occurs if the quantisation steps are noticeable and the filtered error is correlated with the data being quantised. In the latter case quantisation patterns will appear, especially in case repetitive data like static images. Changing the coefficients of the filter different error diffusion pattern will appear for each frame. Due to this effect, the visibility of the error diffusion patterns will reduce.

Advantageously a circuit for processing video performs the quantization of video signals in the luminance domain, with an error diffusion. A LUT is used for the quantization of video signals and the generation of related quantization errors for the feedback loop.

Advantageously a circuit for processing video performs the quantization of video signals in the luminance domain, with an error diffusion. Two adders and two LUTs are used for the quantization of video signals and the generation of related quantization errors for the feedback loop. One LUT is used for the course quantization, the other one for the fine quantization. In this way less memory is required to implement the circuit.

Advantageously a circuit for processing video performs the quantization of video signals in the luminance domain with 2 dimensional error diffusion. Two adders and two LUTs are used for the quantization of video signals and the generation of related quantization errors for the feedback loop. One LUT is used for the quantization of the input signal MSBs and the generation of related quantization errors to be added to the input signal LSBs. The other LUT is used for the quantization of these signals and the generation of related quantization errors for the feedback loop. In this way less memory is required to implement the circuit.

Advantageously in the circuit the LUT implements the gamma correction, to convert signals from the video domain to the luminance domain. Advantageously the circuit for processing video performs the gamma correction and quantization of video signals, with an error diffusion in the luminance domain. A LUT is used for the gamma correction, quantization and the generation of related quantization errors for the feedback loop. In this way less memory is required to implement the circuit. Advantageously a method for processing video performs the gamma correction and quantization of video signals with 2 dimensional error diffusion in the video domain. A memory is used to store pre-calculated values of a LUT. While there are no arithmetic functions performed on the luminance-domain signals, this circuit also supports the generation of non-equidistant values for the quantized output signal. Advantageously the circuit for processing video performs the gamma correction and quantization of video signals, with 2 dimensional error diffusion in the video domain. A LUT is used for the gamma correction, quantization and the generation of related quantization errors for the feedback loop. In this way less memory is required to implement the circuit Advantageously a LUT implements the subfield generation and converts video signals from the luminance domain to the subfield domain.

Advantageously a method for processing video performs the gamma correction, quantization and subfield generation of video signals, with 2 dimensional error diffusion in the video domain. A memory is used to store pre-calculated values of the LUT. While there are no arithmetic functions performed on the subfield-domain signals, this circuit also supports the generation of non-equidistant values for the quantized output signal.

Advantageously a circuit for processing video performs the gamma correction, quantization and subfield generation of video signals, with 2 dimensional error diffusion in the video domain. A LUT is used for the gamma correction, quantization, subfield generation and the generation of related quantization errors for the feedback loop. In this way less memory is required to implement the circuit.

Advantageously a method for processing video, performing the gamma correction, quantization, equidistant subfield generation with partial line doubling, PLD for short, and error diffusion in the luminance domain. PLD requires processing of two pixels at a time. For each of these two pixels a LUT is used for the gamma correction, quantization, subfield generation and the generation of related quantization errors for the feedback loops. The primary pixel LUT is used to determine the MSG subfield value as well as the LSG subfield data and corresponding luminance of both pixels, and generates the related quantisation error for the 2D error diffusion feedback filter. The secondary pixel LUT is used to determine the MSG subfield value taking into account the LSG luminance of both pixels and generates the related quantisation error for the ID error diffusion feedback filter.

Advantageously a circuit for processing video performs the gamma correction, quantization, equidistant subfield generation with partial line doubling and error diffusion in the luminance domain. PLD requires processing of two pixels at a time. For each of these two pixels a LUT is used for the gamma correction, quantization, subfield generation and the generation of related quantization errors for the feedback loops. The primary pixel LUT is used to determine the MSG subfield value as well as the LSG subfield data and corresponding luminance of both pixels, and generates the related quantisation error for the 2D error diffusion feedback filter. The secondary pixel LUT is used to determine the MSG subfield value taking into account the LSG luminance of both pixels and generates the related quantisation error for the ID error diffusion feedback filter.

Advantageously a method for processing video performs the gamma correction, quantization, non-equidistant subfield generation with partial line doubling and error diffusion in the video domain. PLD requires processing of two pixels at a time. For each of these two pixels a LUT is used for the gamma correction, quantization, subfield generation and the generation of related quantization errors for the feedback loops. The primary pixel LUT is used to determine the MSG subfield value as well as the LSG subfield data of both pixels and generates the related quantisation error for the 2D error diffusion feedback filter. The secondary pixel LUT is used to determine the MSG subfield value taking into account the LSG subfield data of both pixels and generates the related quantisation error for the ID error diffusion feedback filter.

Advantageously a circuit for processing video performs the gamma correction, quantization, non-equidistant subfield generation with partial line doubling and error diffusion in the video domain. PLD requires processing of two pixels at a time. For each of these two pixels a LUT is used for the gamma correction, quantization, subfield generation and the generation of related quantization errors for the feedback loops. The primary pixel LUT is used to determine the MSG subfield value as well as the LSG subfield data of both pixels and generates the related quantisation error for the 2D error diffusion feedback filter. The secondary pixel LUT is used to determine the MSG subfield value taking into account the LSG subfield data of both pixels and generates the related quantisation error for the ID error diffusion feedback filter.

Advantageously the look-up tables are implemented as random access memories. In this way the LUTs of the circuit for video processing can be initialized during startup with their application specific settings. Like setting for gamma function, setting for quantisation factor, setting for non-equidistant luminance values and settings for subfield codes and partitioning of dual LUT circuits. New tables may be loaded.

Advantageously initial contents of the lookup tables is calculated off-line by a program which can process the input variables: setting for gamma value, setting for quantisation factor, setting for non-equidistant luminance values and settings for subfield codes and partitioning of dual LUT circuits.

Advantageously lookup tables are implemented as random access memory, the LUTs of the circuit for video processing can be dynamically modified during their operation. In the time between processed video frames, the RAMs can be reloaded with new settings. Like setting for gamma function, setting for quantisation factor, setting for non-equidistant luminance values (table) and settings for Subfield codes (table). The settings can also be user defined like e.g. contrast and brightness settings. They can also be adaptive and depend on video content like e.g. image load. Advantageously dynamic contents of the lookup tables is calculated by an embedded adaptive, controlling software program which can process the input variables, as listed in previous claims: setting for gamma value, setting for quantisation factor, setting for non-equidistant luminance values and settings for subfield codes and partitioning of dual LUT circuits. Control is based upon algorithms, driven by actually measured parameters like image-load, system-temperature, user-preferences, etc.

If no quantisation error is made, then the method and circuit will not affect the video or the audio data. For video this means that the sharpness of the image will not be affected.

The amplitude with which the filter coefficients need to be varied does not depend on the quantisation step size and does therefore not depend on the chosen quantizer or the configuration of the chosen quantizer.

The noise power which is measured and perceived per frame does not increase.

This method and circuit can also be used for audio signals. The invention will now be described in more detail with reference to exemplifying embodiments thereof illustrated in the accompanying drawings, in which

Fig. 1 shows a block circuit diagram comprising a memory for a quantization process and subsequent dequantization process,

Fig. 2 shows a display panel section with neighboring pixels,

Fig. 3 shows the display panel section with error diffusion filter coefficients for the neighboring pixels, Fig. 4 shows a feedback filter with delay elements,

Fig. 5 shows a circuit diagram with two memories with two look-up tables for a coarse and a fine adjustment of a quantization process,

Fig. 6 shows a circuit diagram with two memories with two look-up tables for most significant and least significant bits of a quantization process, Fig. 7 shows a circuit diagram with a memory with a look-up table which contains a gamma correction function and a quantization process for equidistant values, with a feedback loop in a luminance domain,

Fig. 8 shows a block circuit diagram with a memory with a look-up table, which contains a gamma correction function and a feedback loop in a video domain, Fig. 9 shows a diagram for the representation of perceived quantization noise with a classification in equidistant values,

Fig. 10 shows a diagram for the representation of perceived quantization noise for a classification in non-equidistant values,

Fig. 11 shows a circuit diagram with a memory with a look-up table, which replaces a gamma correction function, a quantizer and a sub-field generator,

Fig. 12 shows a block circuit diagram for a partial line doubling,

Fig. 13 shows a block circuit diagram with two memories with two look-up tables for processing pixel values of a primary and a secondary line for equidistant sub-field coding, when partial line doubling is applied, Fig. 14 shows a block circuit diagram with two memories with two look-up tables for processing pixel values of a primary and a secondary line for non-equidistant sub- field coding, when partial line doubling is applied,

Fig. 15 shows a timing diagram with sub-fields for the operation of a plasma display panel, Fig. 16 shows a second timing diagram with sub-fields for the operation of a plasma display panel in which a partial line doubling technique is used,

Fig. 17 shows a block circuit diagram with two memories replacing a gamma correction function, a quantization process and a sub-field generator and generating a quantization error,

Fig. 18 shows an error detection circuit,

Fig. 19 shows a filter with five branches and

Fig. 20 shows a filter and a noise generator changing values of the filter.

Fig. 1 shows a video circuit 1 having an input 2, a gamma correction circuit 3, an adder 4, a register 5, a trunk circuit 6, a memory 7, a sub-field generator 8 also called subfield generation or SFG circuit for short, an output 9 and a feedback filter 10. In the memory 7 a look-up-table is stored, which replaces a quantization function 11 of a quantization multiplier and a quantization rounding 12 of a rounding circuit, a dequantization function 13 of a dequantization multiplier and an addition 14 of a dequantization adder. The look-up table, LUT for short is an allocation table. In the memory 7 an incoming digital video signal, representing either a red, a green or a blue pixel value, is appointed to two digital video output signals. A first digital output signal is supplied to the SFG circuit 8, a second digital output signal is supplied to the filter 10. The first appointment replaces the quantization function 11 and the rounding function 12, the second appointment replaces the quantization function 11, the rounding function 12, the quantization function 13 and the addition 14. The appointment functions are stored as an allocation or a look-up table in the memory 7, LUT memory 7 for short. This means that for 2^m addresses, data is stored for n outputs to the SFG circuit 8 and m+l-n outputs to the filter 10, hence a 2^m * (m+1) bit memory is required.

Fig. 2 shows a display section with neighboring pixels x-1, y-1 and x, y-1 and x+1, x-1 and y-1, y and x, y. Then x is a substitute for the number of the column and y is a substitute for the number of the line.

Fig. 3 shows filter coefficients by which the quantization errors QE which occurred at the neighbouring pixels of the pixel x,y are multiplied, for the generation of a value to be displayed in the pixel x,y. The quantization error, QE for short, is also referred to as quantization noise.

Fig. 4 shows the feedback filter 10 comprising delay elements 15-18 and multiplier elements 19-22 and adders 23, 24, 25. The elements 15, 17 and 18 each delay by one pixel and have a memory location for the value of one pixel, the delay element 16 delays by the number of pixels of one line minus two pixels and accordingly has many memory locations.

The function of the video circuit 1 can be described as follows: in the gamma correction circuit 3 a red, green or blue signal is converted into a red, green or blue luminance signal under the influence of a gamma function. A typical gamma function is nonlinear and reads as follows x = y" with n = 2.4. In order to achieve a sufficiently high resolution for dark areas, at least 3 times 12 bits are to be used. The converted red, green or blue luminance signal is applied to the adder 4 over a parallel data line comprising m or 12, respectively, lines. A value Vi from the register 5 and a further value which is the sum of filtered quantization noise from previous pixels, are added to the luminance signal in the adder 4. With the constant value Vi added to the video signal the trunk circuit 6 acts as a rounding function. The feedback value of the quantization noise of preceding pixels is formed in a filter 10, as described in Fig. 4. The luminance value of a pixel to be displayed is thus calculated as the sum of a current pixel value X(_X,y), which is present at input 2, and of the quantization noise value of neighboring pixels calculated in filter 10. The following equation is valid after the trunk circuit 12:

pixel value to be displayed = trunk( 1/S* tnmk(V₍χ_)y) ⁿ+l/2 + l/16QE_(x-ι_>y-ι₎+5/16QE₍χ,y-i)+3/16QE₍χ_{+ y-}i)+7/16QE(χ.ι,y)))

with the current pixel value V(_X>y), gamma coefficient n with the value 1/2 from the register 5 and with the value l/16QE(_x-ι_>y-i₎+5/16QE₍χ_>y-i)+3/16QE(χ₊ι_ιy-i)+7/16QE₍χ-ι_ιy) as a total sum from filter 10.

The quantizer is defined by the following function F (v) = round(v / S) where S is the quantization factor that is calculated as follows

S = number of input levels / number of output levels (e.g. 1024 / 256 = 4)

The dequantization function is predefined by

F (v) = v * S The influence on the current pixel value N_(X,_y) by the feedback filter values is also known as the Floyd-Steinberg algorithm.

Fig. 5 shows a second video circuit 31 having an input 2, the gamma correction circuit 3, the adder 4, the register 5, the trunk circuit 6, the SFG circuit 8, the output 9, the filter 10 and a circuit 32. This circuit 32 contains a coarse-value LUT 33, a fine- value LUT 34 and two adders 35 and 36. The coarse-value LUT 33 performs a coarse adjustment in the quantization for the output and the fine- value LUT 34 a fine adjustment in the quantization for the output and a feedback loop 37. As a result, a look-up table is split up into two LUTS 33 and 34 and the required memory size is significantly reduced by 0.8 kbyte for a 12-bit input.

Fig. 6 shows a further video circuit 41 having the input 2, the gamma correction circuit 3, the adder 4, the memory 5, the trunk circuit 6, the SFG circuit 8, the output 9, the filter 10 and a circuit 42. This circuit 42 has an MSB LUT 43, an LSB LUT 44 and two adders 45 and 46. MSB is the abbreviation of most significant bits, thus high-order bits, LSB stands for least significant bits, low-order bits. The input data stream of the parallel data is divided into two parts, where m-k data, flow as MSB into the first LUT memory 43 and 2^m"k addresses are used. The LUT provides MSB output data and' a MSB quantization error. The other part, of k bit input data flows, after the MSB quantization error has been added, into LUT 44 and 2^k+1 addresses are used. The LUT provides LSB output which is added to the MSB output and forward to the subfield generation. The remaining quantization error is forwarded to the loop-back filter 10.

The function of the circuit 42 is explained for simplicity with values from the decimal system and a quantisation factor S=2 and is as follows: The input of quantization circuit 42 is value 83. The input value of the MSB

LUT 43 are issued in multiples often (8*10) and the LSB input value of the LUT 44 are issued in steps of one . If, however, the MSB LUT 43 cannot supply the value (8*10/2=40, but only 39), a quantization error QE (1*2=2) occurs on the output of the MSB LUT 43, which flows into the LSB LUT 44 via the adder 46. This quantization error is added to the value of LSB (2+3=5) and is the input for the LSB LUT. The quantized LSB outρut(5/2=2) is added to the quantized MSB ouput in adder 45 (39+2=41). If a quantization error QE occurs on the output of the LSB LUT 43 (l),it flows into the error diffusion feedback filter 10.

In this architecture the LUT memory is divided into two parts, one part generates an MSB quantization and an associated quantization error on a first output and the other part generates the LSB quantization and an associated quantization error on a second output. If the two output signals are added together, this will lead to the new quantized value. The size of the MSB random-access memory is 2^m/2 *(m+l) The size of the LSB random-access memory is 2^m/2+1) *(m+l) Fig. 7 shows a video processing circuit 51 having an input 2, comprising a memory 52 with a LUT in which a gamma correction function 53 and an error function 54 are combined. The error function 54 comprises a quantization function 11 of a quantization multiplier and a quantization rounding 12 of a rounding function, a dequantization function 13 of a dequantization multiplier and an addition 14 of a dequantization adder. The quantized output is passed, via adder 55 to the SFG circuit 8 and output 9. Adder 4 and filter 10 take care of the error diffusion feedback loop.

The quantization error from memory 52 is accumulated in adder 4. When an overflow occurs, the quantized output from memory 52 is incremented by adder 55, which is supplied via line 56. The rest is passed to the error diffusion feedback loop filter 10. This means that for 2^m"4 addresses, data is stored for n outputs to the SFG circuit 8 and m+l-n outputs to the filter 10, hence a 2^m * (m+1) bit memory is required.

Fig. 8 shows a video processing circuit 61 comprising a LUT 62 for values converted with equidistant or non-equidistant values. LUT 62 replaces a gamma correction circuit 63, a quantizer 64, a dequantizer 65, an inverse gamma correction circuit 66 and an adder 67. Since the input video values are converted in the gamma correction circuit 63, an inverse gamma correction circuit 66 is required in the feedback loop 68. A rounding circuit 69 is inserted between the filter 10 and the adder 4.

Fig. 9 shows a gamma curve 71 which is converted with 256 equidistant values into curve 72. The result is a quantization noise curve 73 in the perception domain, with a high quantization error in a dark area between the absolute values 1 and 22. Especially in dark areas the perception by the human eye is more sensitive than in bright areas. The high quantization error is thus perceived by a viewer. This provides a discrepancy between sampling and perception.

Fig. 10 shows a gamma curve 81 which is sampled with only 34 non- equidistant values. The non-equidistant values are shown as curve 82. The result is a quantization noise curve 83 in the perception domain. The first value in the dark area has a smaller quantization error. In bright areas the quantization noise is increased, however it is still less then the perceived quantization noise in the dark areas and therefore not significant. The small set of quantized luminance values (subfields) allows an improved motion portrayal for PDP displays.

Fig. 11 shows a video circuit 101 comprising a LUT 102 which replaces the gamma correction circuit 63, the quantizer 64, the dequantizer 65, the inverse gamma correction circuit 66, the addition 67 and the sub-field generator 8. The size of LUT 102 is only 2^m addresses, data is stored for k outputs of the SFG function and m+l-n outputs to the filter 10, hence a 2^m"4 * (m+l-n+k bit memory is required, which gives a significant reduction of memory. Equidistant as well as non-equidistant values are supported.

Fig. 12 shows a circuit 111 comprising a line delay 112, a min/max detection circuit 113, a first exchange circuit 114, a partial line doubling circuit 115, a second exchange circuit 116, another line delay 117 and a multiplexer 118. The line of a television picture is delayed by one line in the delay circuit 112. Then values of two pixels lying above each other in one column are compared in the detection circuit 113. The respective larger value is defined and applied to a primary input, the smaller value to the secondary input of the doubling circuit 115, which processes two pixels at a time. When the pixels where exchanged in the exchange circuit 114, they are swapped back in the second exchange circuit 116. Another line delay and multiplexer reconstruct the one pixel at a time nature of the video stream.

Fig. 13 shows a first partial line doubling circuit 120 for equidistant sub-field codings which circuit can be used for the partial line doubling circuit 115, comprising a first gamma correction circuit 121, an adder 122, an inverse gamma correction circuit 123, a LUT 124, a 2D filter 125, a further gamma correction circuit 126, an adder 127, an inverse gamma correction circuit 128, a further LUT 129 and a one-dimensional filter 130. The LUT 124 replaces a gamma correction circuit 131, a quantizer 132, an SFG and a PLD circuit 133, a converter 134, a dequantizer 135, a second converter 136, a second dequantizer 137 and an adder 138. The SFG and PLD circuit 133 includes an MSG function 139, an LSG function 140, an LSG light function 141 and a QE function 142. The LUT 129 replaces a gamma correction circuit 143, a quantizer 144, an SFG and PLD circuit 145, a further converter 146, a dequantizer 147 and an adder 148. The SFG and PLD circuit 145 includes an MSG function 149 and a QE function 150. Video signals are present on inputs 151 and 152 and subfield output signals are output via the outputs 153/154 and 155/156. This implementation of a partial line doubling circuit 120 can be used as partial line doubling circuit 115.

Fig. 14 shows another partial line doubling circuit 159, for non-equidistant sub-field codings. It can be used as partial line doubling circuit 115, with an adder 160, a LUT 161, a 2D filter 162, a further adder 163, a further LUT 164 and a one-dimensional filter 165. The LUT 161 replaces a gamma correction circuit 166, a quantizer 167, an SFG and PLD circuit 168, a converter 169, a dequantizer 170, an inverse gamma correction circuit 171 and an adder 172. The SFG and PLD circuit 168 includes an MSG function 173, an LSG function 174 and a QE function 175. The LUT 164 replaces a gamma correction circuit 176, a quantizer 177, an SFG and PLD circuit 178, a converter 179, a dequantizer 180, an inverse gamma correction circuit 181 and an adder 182. The SFG and PLD circuit 178 includes an MSG function 183 and a QE function 184. Signals are present on video inputs 185 and 186 and subfield output signals are output via outputs 187/188 and 189/190. Fig. 15 shows a timing diagram of the driving of a PDP display with eight sub- fields 201 to 208. While the display is driven according the Address and Displays Separated methode (ADS), each sub-field has an erasing time 209, an addressing time 210 and a sustaining time 211. The eight sub-fields cover a frame period 212. The sub-fields 201 to 204 represent a least significant group or LSG for short. The sub-fields 205 to 208 represent a most significant group or MSG.

If the LSG for two pixels in successive lines are identical, there can be a time- saving 213 as shown in Fig. 16. The line-doubling of parts of the addresing is called partial line doubling, PLD for short. Only during the sustain period light pulses can be emitted from the red, green or blue light sources of a pixel, hence the saved time 213 can be used to generate more light

The function of the circuit 120 is as follows: a pixel value is present on the input of the memory 124 and is converted in the gamma correction circuit 131 into the luminance domain, therefore, an 8-bit data word becomes an 12-bit data word to achieve a sufficiently high resolution in dark areas. In the subsequent quantizer the 12-bit data word is quantized to a resolution, which matches with the sub-field generation. The latter data word is applied to the SFG and PLD circuit 133 and associated sub-field data is generated in this circuit.

A light value, corresponding with the generated subfield data is generated 142 and provided to a converter 134 which converts the light value signal into a luminance signal. From the converter 134 the signal is conveyed to the dequantizer 135, which cancels the quantization. The value generated now is compared with the signal from the gamma correction circuit 131 and the actual quantization error is determined in the adder 138. The quantization error is applied to the 2D filter 125 and filtered in accordance with the Floyd- Steinberg algorithm. Since the filtered quantization error is situated in the luminance area and is to be applied to the input signal, the input signal is transformed into the luminance area by the gamma correction circuit 121. This transformation is cancelled in the inverse gamma correction circuit 123.

Since the filtered quantization error also influences pixel values of pixels of the neighboring line, the filter 125 is connected to the adder 127. Thus an output signal of the 2D filter is added for further processing to an input signal, which represents a pixel value of the neighboring line.

The light value of the LSG is transported from the LSG function 141 via the converter 136, the dequantizer 137 to the adder 127 and thus the input signal 152 is corrected for the luminance which is generated by the LSG subfields.

The video signals in the gamma correction circuit 143 in LUT 129 are transformed from the video area into the luminance area, then quantized in the quantizer 144 and conveyed to an SFG and PLD circuit 145 which generates the MSG part of the sub-field data of the PDP. A light value signal is generated which is equal to the subfield data, and converted 146 into a luminance signal. The luminance signal is dequantized in the dequantizer 147 and applied to the adder 148. In the adder is determined an actual quantization error and applied to the filter 130. The quantization error has no effect on neighboring lines, so that only a one-dimensional filter 130 is used. Since here too the processing takes place in the luminance area, the adder 127 is surrounded by a gamma correction circuit 126 and an inverse gamma correction circuit 128 which convert the values of the current pixel of the second neighboring line.

The function of the circuit 159 can be described as follows: a pixel value is present on video input 185 of the circuit 159 and is converted into the luminance area in the gamma correction circuit 166 and for this purpose an 8-bit data word becomes a 12-bit data word to achieve a sufficiently high resolution in dark areas. In the subsequent quantizer 167 the 12-bit data word is adapted to a data word that is necessary for the sub-field generation. This data word is applied to the SFG and PLD circuit 168 and in this circuit the associated sub-field data is generated, which is output via the outputs 187 and 188. A light value, corresponding with the generated subfield data is generated 175 and provided to a converter 169, which converts the light value signal into a luminance signal. From the converter 169 the signal is conveyed to the dequantizer 170, which cancels the quantization. The downstream inverse gamma correction circuit 171 converts luminance data into video data. The input signal is subtracted from this value . The quantization error is filtered in the filter 162 and added 160 to the input values of the primary line 185. Since the filtered quantization error also influences pixel values of pixels of the neighboring line, the filter 162 is connected to the adder 163. In the LUT 164 the video signals in the gamma correction circuit 176 are transformed from the video area to the luminance area, then quantized in the quantizer 177 and conveyed to an SFG and PLD circuit 178 which generates the sub-fields for the PDP. Only the MSG needs to be generated. In the converter circuit 179 a light value signal, corresponding to the MSG subfield 190, is reconverted into a luminance signal. The luminance signal is dequantized 180 and applied to the inverse gamma correction circuit 181. Subtractor 182 is determines the quantization error and which is applied to the filter 165. The quantization error has no effect on neighboring lines so that only a one-dimensional filter 165 is used.

The LSB on the output 188 connected to an input of LUT 164 and are applied to the quantizer 177 and are thus taken into account when the most significant bits of the neighboring line available are formed on output 189.

Figure 17 shows a video circuit 221 comprising two memories 222 and 223, a filter 10, an adder 224, an input 225 and an output 226. The first memory 222 replaces a gamma correction circuit 227, a quantizer 228 and a sub-field generator 229. The second memory 223 replaces a gamma correction circuit 230, a quantizer 231, a dequantizer 232, an inverse gamma correction circuit 233 and an adder 234. The first memory 222 is used to generate sub-fields to address directly a display-panel. The second memory 223 is used to generate a quantisation error. In the filter 10 the quantisation error is filtered and added to input signals via the adder 224. Advantageously errors generated by quantisation for a defined sub-field generator, are minimized in a feedback loop comprising the memory 223 and the filter 10.

Figur 18 shows an error detection circuit 240 with an input 241, a quantizer 242, a dequantizer 243, an adder 244, a filter 10, an second adder 245 and an output 246. The dequantizer 243, the adder 244 and the filter 10 are arranged in a feedback loop 247. The function of this circuit 240 is as follows:

Input signals reaches the quantizer 242 and are quantized. The quantized signals support the output 246 and the dequantizer 243. In the dequantizer the quantized signals are dequantized. In the adder 244 the dequantized signals are subtracted from the input signals, that means compared with the input signal and an error is detected. The detected error is filtered in the filter 10. In the adder 244 the filtered error is added to neighboring signals. Figure 19 shows a second implementation of the filter 10 comprising two switches 250 and 251, five filter branches 252-256 and a clock generator 257. The filter branches 252-256 comprise the delay elements 15-18 and the adders 23-25. The filter branch 252 comprises the four multiplier elements 19-22 with the coefficients 7/16, 3/16, 5/16, and 1/16. The filter branch 253 comprises the four multiplier elements 258-261 with coefficients 2/16, 4/16, 5/16 and 5/16, the filter branch 254 comprises the three multiplier elements 262- 264 with coefficients 5/16, 3/16, 8/16, the filter branch 255 comprises the four multiplier elements 265-268 with coefficients 1/16, 5/16, 3/16, and 7/16 and the filter branch 256 comprises the four multiplier elements 269-272 with coefficients 3/16, 5/16, 4/16, and 4/16. The idea behind this sequence is that each time the sum of coefficients equals 1, which means that the direct current error equals zero. Since the Floyd-Steinberg filter coefficients are proved to be optimal, the coefficients of each filter branch deviate minimal around its original value, which means that on average each filter coefficient remains approximately the same as the coefficient originally determined by Floyd-Steinberg. The filter branch 254 comprises a fourth coefficient with an amount of 0/16, therefore a multiplier element is not realized.

Figure 20 shows a video circuit 301 with an input 302, a gamma correction circuit 3, an adder 303, a memory 304. a sub-field generator 8, an output 305, a filter 10, a noise-generator 306 and a multiplier element 307. The memory 304 comprises a quantizer 308, a dequantizer 309 and an adder 310. In the adder 310 input data of the memory 304 are subtracted from the dequantised data.

The noise-generator 306 works in different operating modes. A maximun deviation and the number of different noise levels determine these operating modes. For switching between two deviation modes an amplitude of 10 % is sufficient. It is possible to use a new random value for each frame, for each pixel or for each sub-pixel. Output values of the filter 10 are multiplied with the random factor that has an average value of 1 and a maximum deviation of 0.1. Two bits determine a deviation of 0 %, 5 % or 10 % and one bit is a sign bit. Therefore only three bits of random data are needed to obtain good results. With these filters the visibility of spatial error diffusion patterns in quantisation is reduced, while not affecting the quality of the video data.

REFERENCE LIST

1 video circuit 32 circuit input 33 coarse- value memory gamma correction circuit 34 fine-value memory adder 35 adder memory 36 adder rounding circuit 37 feedback loop memory 38 sub-field generator circuit 39 output 40 filter 41 video circuit quantization function 42 circuit quantization rounding 43 MSB memory dequantization function 44 LSB memory addition 45 adder delay element 46 adder delay element 47 delay element 48 delay element 49 multiplier element 50 multiplier element 51 video circuit multiplier element 52 memory multiplier element 53 gamma correction function adder 54 error function adder 55 adder adder 56 line

57

58

59

60

61 video circuit video circuit 62 memory gamma correction circuit 94 quantizer 95 dequantizer 96 inverse gamma correction circuit 97 adder 98 feedback loop 99 rounding circuit 100

101 video circuit gamma curve 102 memory luminance curve 103 sub-field generation perceived quantisation error curve 104

105

106

107

108

109

110

111 circuit gamma curve 112 line delay luminance curve 113 Min/Max detection circuit perceived quantisation error curve 114 substitution circuit

115 partial line doubling circuit

116 second substitution circuit

117 line delay memory

118 multiplexer

119

120 partial line doubling circuit

121 gamma correction circuit

122 adder

123 inverse gamma correction circuit

124 LUT memory 125 2D filter 157

126 gamma correction circuit 158

127 adder 159 second partial line doubling circuit

128 inverse gamma correction circuit 160 adder

129 LUT memory 161 memory

130 one-dimensional filter 162 2D filter

131 gamma correction circuit 163 adder

132 quantizer 164 memory

133 SFG and PLD circuit 165 one-dimensional filter

134 converter 166 gamma correction circuit

135 dequantizer 167 quantizer

136 converter 168 SFG and PLD circuit

137 second dequantizer 169 converter

138 adder 170 dequantizer

139 MSG circuit 171 inverse gamma correction circuit

140 LSG circuit 172 adder

141 LSG light circuit 173 MSG circuit

142 QE circuit 174 LSG circuit

143 gamma correction circuit 175 QE circuit

144 quantizer 176 gamma correction circuit

145 SFG and PLD circuit 177 quantizer

146 inverse luminance circuit 178 SFG and PLD circuit

147 dequantizer 179 converter

148 adder 180 dequantizer 149 181 inverse gamma correction circuit

150 182 adder

151 input 183 MSG circuit

152 input 184 QE circuit

153 output 185 input

154 output 186 input

155 output 187 output

156 output 188 output 189 output 240 error detection circuit

190 output 241 input 191 242 quantizer

192 243 dequantizer 193 244 adder 194 245 adder 195 246 output 196 247 feedback loop 197 198 199 200

201 sub-field

202 sub-field 250 switch

203 sub-field 251 switch

204 sub-field 252 filter branch

205 sub-field 253 filter branch

206 sub-field 254 filter branch

207 sub-field 255 filter branch

208 sub-field 256 filter branch

209 erasing time 257 clock generator

210 addressing time 258 multiplier element

211 sustaining time 259 multiplier element

212 duration of image 260 multiplier element

213 time saving 261 multiplier element 214 262 multiplier element

215 263 multiplier element 216 264 multiplier element 217 265 multiplier element 218 266 multiplier element 219 267 multiplier element 220 268 multiplier element 221 video circuit 269 multiplier element 222 first memory 270 multiplier element

223 second memory 271 multiplier element

224 adder 272 multiplier element

225 input

226 output 301 video circuit

227 gamma correction circuit 302 input

228 quantizer 303 adder

229 sub-field generator 304 memory

230 gamma correction circuit 305 output

231 quantizer 306 noise generator

232 dequantizer 307 multiplier element

233 inverse gamma correction circuit 308 quantizer

234 adder 309 dequantizer

310 adder

Claims

CLAIMS:

1. A video circuit for processing video signals which show images on a display panel with linear luminance transmission, comprising a gamma correction circuit, a quantizer and a sub-field generator circuit, characterized in that a coarse adjustment of the quantization is done in a first memory and a fine adjustment of the quantization is done in a second memory.

2. A video circuit for processing video signals which display images on a display panel with linear luminance transmission, comprising a gamma correction circuit, a quantizer and a sub-field generation circuit, characterized in that most significant bits are quantized in a first memory and least significant bits are quantized in a second memory.

3. A video circuit for processing video signals which show images on a display panel with linear luminance transmission, comprising a gamma correction circuit, a quantizer and a sub-field generation circuit, characterized in that a memory replaces the quantizer.

4. A video circuit as claimed in claim 3, characterized in that the memory replaces a dequantizer.

5. A video circuit as claimed in claim 3 and/or 4, characterized in that the memory replaces the gamma correction circuit.

6. A video circuit as claimed in one or more of the preceding claims 4-5, characterized in that an inverse gamma correction circuit is arranged downstream of the dequantizer.

7. A video circuit as claimed in one or more of the preceding claims 3-6, characterized in that the memory replaces the sub-field generation circuit.

8. A video circuit as claimed in one or more of the preceding claims 3-6, characterized in that the memory supplies a filter with values.

9. A video circuit as claimed in claim 7, characterized in that the sub-field generation circuit applies values to a filter via a converter and the dequantizer.

10. A video circuit as claimed in claim 9, characterized in that the filter applies values to an adder which is situated in an input area of a second signal which represents pixel values of a neighboring line.

11. A video circuit as claimed in one or more of the preceding claims 7-9 or 10, characterized in that the sub-field generation circuit applies values to the adder via a second converter and a second dequantizer.

12. A video circuit as claimed in claim 10 and/or 11, characterized in that pixel values of the neighboring line are quantized in a second quantizer in a second memory and in the second memory sub-fields are generated in a second sub-field generation circuit.

13. A video circuit as claimed in claim 12, characterized in that the sub-field generation circuit applies values to the quantizer of the second memory.

14. A video circuit as claimed in one or more of the preceding claims 1-13, characterized in that the memory is a read-only memory.

15. A video circuit as claimed in claims 8 or 9, characterized in that coefficients of the filter are changeable.

16. Method for a video circuit as claimed in claims 8 or 9, characterized in that coefficients of the filter are changeable.

17. A circuit for processing video, performing the quantization of video signals in the luminance domain, with an error diffusion, characterized in that a table is used for the quantization of video signals and the generation of related quantization errors for a feedback loop.

18. A circuit for processing video, performing the quantization of video signals in the luminance domain, with an error diffusion, characterized in that two adders and two tables are used for the quantization of video signals, one for a coarse and one for a fine quantization, and the generation of related quantization errors for a feedback loop.

19. A circuit for processing video, performing the quantization of video signals in the luminance domain, with an error diffusion, characterized in that two adders and two tables are used for the quantization of video signals, one for most significant and one for least significant bits, and the generation of related quantization errors for a feedback loop.

20. A circuit as claimed in one of the preceding claims 17-19, characterized in that the table implements the gamma correction.

21. A circuit for processing video, performing the gamma correction and quantization of video signals, with an error diffusion in the luminance domain, characterized in that a table is used for the gamma correction, quantization and the generation of related quantization errors for a feedback loop.

22. A method for processing video, performing the gamma correction and quantization of video signals, with an error diffusion in the video domain, characterized in that a memory is used to store pre-calculated values of a table.

23. A circuit for processing video, performing the gamma correction and quantization of video signals, with an error diffusion in the video domain, characterized in that a table is used for the gamma correction, quantization and the generation of related quantization errors for the feedback loop.

24. Circuit as claimed in one of the preceding claims 20, 21 or 23, characterized in that the table is combined with a table to implement the subfield generation.

25. A method for processing video, performing the gamma correction, quantization and subfield generation of video signals, with an error diffusion in the video domain, characterized in that a memory is used to store pre-calculated values of the table.

26. A circuit for processing video, performing the gamma correction, quantization and subfield generation of video signals, with an error diffusion in the video domain, characterized in that a table is used for the gamma correction, quantization, subfield generation and the generation of related quantization errors for the feedback loop.

27. A method for processing video, performing the gamma correction, quantization, equidistant subfield generation with partial line doubling and an error diffusion in the luminance domain.

28. A circuit for processing video, performing the gamma correction, quantization, equidistant subfield generation with partial line doubling and error diffusion in the luminance domain.

29. A method for processing video, performing the gamma correction, quantization, non-equidistant subfield generation with partial line doubling and error diffusion in the video domain.

30. A circuit for processing video, performing the gamma correction, quantization, non-equidistant subfield generation with partial line doubling and error diffusion in the video domain.

31. A circuit as claimed in one of the preceding claims 17 to 21, 23, 24, 26, 28 and 30 with lookup tables implemented as random access memory.

32. A circuit as claimed in one of the preceding claims 17 to 21, 23, 24, 26, 28, 30 and 31 in which the initial contents of the lookup tables is calculated off-line by a software program.

33. A circuit as claimed in claim 31, while the lookup tables are implemented as random access memory new settings may be reloaded.

34. A circuit as claimed in one of the preceding claims 17 to 21, 23, 24, 26, 28, 30,

31, 32 and 33 in which the dynamic contents of the lookup tables is calculated by an embedded adaptive, controlling software program.

35. Error diffusion circuit with a quantizer, a dequantizer and a filter in a feedback loop.

36. Error diffusion circuit as claimed in claim 35, characterized in that coefficients of the filter are changeable.

37. Error diffusion circuit as claimed in claim 35,characterized in that output signals of the filter are multiplied with different factors.

38. Method for an error diffusion circuit with a quantizer, a dequantizer and a filter in a feedback loop, characterized in that coefficients of the filter are changeable.

39. Method for an error diffusion circuit with a quantizer, a dequantizer and a filter in a feedback loop, characterized in that output signals of the filter are multiplied with different factors.