Pre-compensation of high frequency component in a video scaler

Abstract

A video scaling process system, wherein the high frequency components of the input video sequence are compensated for, prior to video scaling. The video scaling system includes a pre-compensator for compensation of the high frequency components of the input video sequence, and a video scaler. The pre-compensator includes a high frequency component extractor, a noise detector and a shoot suppressor. The high frequency component extractor extracts the high frequency compensation component from an input video sequence. The noise detector adjusts the gain of the compensation to avoid the enhancement of the noise component. The shoot suppressor further adjusts the gain of the compensation to avoid shoot artifacts. The adjusted high frequency compensated video is added back to the input video. The pre-compensated video sequence is then passed through a video scaler to obtain the scaled video. As such, the sharpness of the scaled video is maintained or even enhanced.

Description

FIELD OF THE INVENTION

The present invention relates generally to video processing, and more particularly to video scaling in video processing.

BACKGROUND OF THE INVENTION

In TV systems, video scaling is often required when the native resolution of an incoming video frame does not fit the display resolution. Without loss of generality, assume the scaling ratio is L/M and the greatest common divisor gcd(L,M)=1. A conventional scaling system 100 as shown in FIG. 1 is generally used to scale an input signal x(k) with an arbitrary rational scaling ratio. In that system, the input signal x(k) is up-sampled by L in an up-sampler 102, filtered by an interpolation filter 104 with frequency response H_L,M, and then down-sampled in the down-sampler 106 by M. Such process can be implemented using a polyphase structure. Using such a system, a two dimensional video frame can be first scaled in the horizontal direction with one scaling ratio and then scaled in the vertical direction with another scaling ratio, or first scaled in the vertical direction with one scaling ratio and then scaled in the horizontal direction with another scaling ratio.

In the video scaling system of FIG. 1, the up-sampling expands the time scale by a factor of L and inserts L-1 zeros between each sample of x(k), to obtain the new sequence u(k). In the frequency domain, the up-sampling introduces image components at frequency 2πl/L, 1≦l≦L−1. Therefore, in order to prevent artifacts, u(k) has to be filtered by a low-pass filter with cutoff frequency π/L. On the other hand, the down-sampling takes every Mth sample of v(k) to obtain the scaled sequence y(k). However, such down-sampling generally results in aliasing.

In order to prevent aliasing, the sequence v(k) should be filtered prior to down-sampling with a low-pass filter that has a cutoff frequency π/M . The cascade of two low-pass filters with cutoff frequencies π/L and π/M is equivalent to a single low-pass interpolation filter H_L,M with a cutoff frequency min(π/L,π/M). Therefore, the frequency characteristics of the interpolation filter H_L,M depend on the values of L and M.

Because the interpolation filter H_L,M is a low-pass filter, the scaled video lacks in high frequency components, whereby the scaled video is often blurred and not as sharp as the original input video sequence.

BRIEF SUMMARY OF THE INVENTION

In one embodiment, the present invention provides a video scaling process and system wherein the high frequency components of the input video sequence are compensated for prior to video scaling. The video scaling system comprises a high frequency component extractor, a noise detector and a shoot suppressor. The high frequency component extractor extracts the high frequency compensation component from an input video sequence. The noise detector adjusts the gain of the compensation to avoid the enhancement of the noise component. The shoot suppressor further adjusts the gain of the compensation to avoid shoot artifacts. The adjusted high frequency compensated video is added back to the input video. The pre-compensated video sequence is then passed through a video scaler to obtain the scaled video. As such, the sharpness or details of the scaled video is maintained or even enhanced.

Other embodiments, features and advantages of the present invention will be apparent from the following specification taken in conjunction with the following drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 shows a block diagram of a system illustrating a conventional scaling method.

FIG. 2 shows a block diagram of an example video scaling system according to an embodiment of the present invention, providing pre-compensation of the high frequency image components of the input video signal.

FIG. 3 shows a block diagram of an embodiment of the video pre-compensator of FIG. 2, for compensation of the high frequency component of the input video signal, according to an embodiment of the present invention.

FIG. 4 shows an example of the high frequency peaking in the pre-compensation of the high frequency component of the input signal, according to an embodiment of the present invention.

FIG. 5 shows an example of adjusting the high frequency component α(k), according to an embodiment of the present invention.

FIG. 6 shows an example of adjusting the signal p(k), according to an embodiment of the present invention.

FIGS. 7(a)-(d) show four examples of luminance pattern where shoot artifacts may occur at the boundary between flat area and increasing or decreasing area during the pre-compensation of the high frequency component.

FIG. 8 shows an example of determining the intensity factor r(k) from the high frequency component α(k), according to an embodiment of the present invention.

FIG. 9 shows an example of a parallel pre-compensation system to compensate the high frequency component prior to the video scaler of FIG. 1, according to another embodiment of the present invention.

DETAILED DESCRIPTION OF THE INVENTION

In one embodiment, the present invention provides a video scaling process and system wherein the high frequency components of the input video sequence are compensated for prior to video scaling. An example video scaling system according to the present invention includes a pre-compensator for compensation of the high frequency components of the input video sequence, and a video scaler.

The pre-compensator comprises a high frequency component extractor, a noise detector and a shoot suppressor. The high frequency component extractor extracts the high frequency compensation component from an input video sequence. The noise detector adjusts the gain of the compensation to avoid the enhancement of the noise component. The shoot suppressor further adjusts the gain of the compensation to avoid shoot artifacts. The adjusted high frequency compensated video is added back to the input video. The pre-compensated video sequence is then passed through a video scaler to obtain the scaled video. As such, the sharpness of the scaled video is maintained or even enhanced.

FIG. 2 shows a block diagram of an embodiment of a video scaling system 200 with the pre-compensation of the high frequency components of the input signal x(k), according to an embodiment of the present invention. The system 200 includes a pre-compensator 202 for compensation of the high frequency components of the input video sequence, and a video scaler 204. The pre-compensator 202 first pre-compensates the high frequency component in the input signal x(k), and then the obtained signal x′(k) is passed through the scaler 204 to obtain the scaled signal y(k). An arbitrary scaler can be used instead of the scaler shown in FIG. 1. In addition, two one dimensional scalers 204 can be cascaded to scale a two dimensional video frame in horizontal and vertical directions, with an arbitrary order to obtain a scaled video frame. For ease of understanding in this description, the input signal x(k) can be regarded as a set of one dimensional pixel values in either horizontal or vertical direction.

FIG. 3 shows a block diagram of an embodiment of the video pre-compensator 202 in the system of FIG. 2, for compensation of the high frequency component of the input video signal, according to an embodiment of the present invention. The pre-compensator 202 comprises a high frequency component extractor 300, a shoot suppressor 302 and a noise detector 304. The high frequency component extractor 300 extracts the high frequency compensation component d(k) from the input signal x(k). The shoot suppressor 302 computes a suppression factor g₂(k) to avoid the shoot artifacts. The noise detector 304 computes a gain adjustment parameter g₁(k) based on the noise level.

In the pre-compensator 202, a pre-compensation gain g₀ is multiplied by the gain adjustment parameter g₁(k) at a multiplier 306 to generate the product (g₀·g₁(k)). The product (g₀·g₁(k)) is further multiplied by the suppression factor g₂(k) at a multiplier 308 to generate the product (g₀·g₁(k)·g₂(k)). Then, the product (g₀·g₁(k)·g₂(k)) is multiplied by the high frequency compensation component d(k) at the multiplier 310 to generate the product (g₀·g₁(k)·g₂(k)·d(k)). Then, the product (g₀·g₁(k)·g₂(k)·d(k)) is added to the input signal x(k) at the adder 312 to generate the pre-compensated signal x′(k) as: x′(k)=x(k)+g₀·g₁(k)·g₂(k)·d(k).

In order to obtain the high frequency compensation component d(k) above, the input signal x(k) is first passed through a high-pass filter in the high frequency component extractor 300 to obtain a high frequency component α(k). For example, a high pass filter [−1,2,−1] can be used to extract α(k). Other high pass filters can also be used. FIG. 4 shows an example graph 400 of adjusting the high frequency component α(k) to output the signal β(k), where t₁, t₂ and t₃ are thresholds. Basically, if 0≦|α(k)|≦t₁, the amplitude of α(k) is reduced to avoid boosting up the noise, i.e. $\beta \left(k\right)=\alpha \left(k\right)·\frac{\alpha \left(k\right)}{t_1}.$ If t₁<|α(k)|≦t₂, the signal α(k) is bypassed since the compensation will not be significant in small high frequency area. If t₂<|α(k)|≦t₃, then the amplitude of α(k) is enlarged to obtain a significant compensation, i.e. β(k)=sign(α(k))·(t₂+√{square root over ((|α(k)|−t₂)(t₃−t₂))}), where sign(.) denotes the sign function. If |α(k)>t₃, then the signal α(k) is bypassed to avoid the overcompensation. The thresholds t₁, t₂ and t₃ satisfy 0≦t₁≦t₂≦t₃, which can be either manually preset or dynamically calculated.

An example of dynamically calculating the thresholds t₁, t₂ and t₃ is now described. A signal p(k) is calculated as p(k)=|α(k)− x(k)|, where x(k) is the local mean of the signal x(k). FIG. 5 shows an example graph 500 of adjusting the signal p(k) to output the signal q(k), where P_max, R_min and R_max are pre-defined values. Basically, if p(k)=0, then let q(k)=R_min. If p(k)≧P_max, then let q(k)=R_max. And, if 0<p(k)<P_max, then q(k) is linearly interpolated. The thresholds t₁ and t₂ are calculated as t₁=t₂=min(T,q(k)), where T is a constant value that can be manually set. The threshold t₃ is calculated as t₃=q(k).

Once the signal β(k) is obtained, the high frequency compensation component d(k) thus can be calculated as: d(k)=β(k)−α(k).

Referring back to FIG. 3, the noise detector 304 is used to scale down the high frequency compensation where local standard deviation (or variance) is detected as low. Low local standard deviation often indicates a relatively flat area. Without scaling down the compensation in such area, noise may be boosted up.

In a two dimensional video frame, the local standard deviation σ(k) is often calculated based on a rectangular neighboring window centered at the current pixel. Other methods can also be used to calculate the local standard deviation. The gain adjustment parameter g₁(k) is a monotonically increasing function of the local standard deviation. FIG. 6 shows an example monotonically increasing function 600 wherein if the local standard deviation σ(k) is less than a threshold τ₁, then the input signal x(k) is in the flat area and the adjustment parameter is set as g₁(k)=0. If σ(k) is greater than a threshold τ₂ (τ₂≧τ₁), then the input signal x(k) contains useful edge information and the adjustment parameter is set as g₁(k)=1 to fully compensate the high frequency component. If τ₁≦σ(k)≦τ₂, then the adjustment parameter g₁(k) is linearly interpolated in a soft-switching manner as shown by example in FIG. 6. Specifically, if we consider linear soft-switching for instance, it can be $g\left(k\right)=\frac{\sigma \left(k\right)-{\tau }_1}{{\tau }_2-{\tau }_1},\mathrm{for}{\tau }_1\le \sigma \left(k\right)\le {\tau }_2.$ The thresholds τ₁ and τ₂ can be set to constant values or dynamically adjusted based on noise estimation value. When the noise estimation value is high and the video sequence is considered noisy, then τ₁ and τ₂ should also be set to relatively large values. When the noise estimation value is low and the video sequence is considered clean, then τ₁ and τ₂ could be set to relatively small values so that the fine details in the video frame can be better pre-compensated. The relation between noise detection thresholds τ₁, τ₂ and the noise estimation value σ_e can be expressed as: τ₁=c₁*σ_e, τ₂=c₂*σ_e, where c₁ and c₂ are two pre-defined values satisfying c₂≧c₁≧0. The noise estimation value σ_e can be obtained by the method described in commonly assigned patent application Ser. No. 10/991,265, titled “Methods to estimate noise variance from a video sequence”, filed on Nov. 17, 2004, which is incorporated herein by reference. The inventors have further noted that the pre-compensation in horizontal and vertical directions can share the same noise detector to save the computational cost.

In the pre-compensation of the high frequency components, shoot artifacts usually occur around edge areas. More specifically, shoot artifacts usually occur around a sharp turning point of pixel luminance that is transitioning from a relatively smooth area. FIGS. 7(a)-(d) show four example luminance patterns 702, 704, 706 and 708, respectively, where shoot artifacts may occur during the pre-compensation of the high frequency component. In each pattern, eight pixels 700 (•) are shown for illustration purposes. The number of pixels 700 actually involved can vary depending on the number of filter taps of the high pass filter in high frequency component extractor 300 (FIG. 3). It can be seen that one commonality among the four patterns in FIGS. 7(a)-(d) is that they all contain a transitioning area and a large flat area in terms of pixel luminance level.

Based on the above analysis, the shoot suppressor 302 (FIG. 3) further implements symmetry checking and intensity checking. For symmetry checking, the shoot suppressor 302 checks the luminance variation of pixels on both sides of the current pixel within the filtering range. Based on the checking result, different patterns of luminance variation in the neighborhood of the current pixel can be categorized. Therefore, higher suppression can be assigned to those patterns that are more likely to cause overshoot/undershoot. For intensity checking, the shoot suppressor 302 checks the magnitude of the high frequency component α(k). If α(k) is small, it is unlikely for obvious shoot artifacts to appear at that position. Therefore, suppression can be associated with the intensity of α(k) at each pixel position.

The goal of the symmetry checking is to detect the patterns shown in FIGS. 7(a)-(d). Within the filtering range, if the pixel luminance curve is close to any pattern in FIGS. 7(a)-(d), then the pre-compensation at that position should be suppressed. Let N be the value such that (2*N+1) is the length of the high pass filter in the high frequency component extractor 300. In order to check the symmetry of pixel luminance curve with reference to the position of x(k) and calculate a symmetric factor s(k), the two values M_l and M_r are first calculated as follows: M_l=max(d_0,−N,d_0,−N+1, . . . ,d_0,−l), M_r=max(d_N,0,d_N−1,0, . . . ,d_1,0),

where d_i,j represents the absolute difference between x(k+i) and x(k+j) as d_i,j=|x(k+i)−x(k+j)|

The value M_l is the maximum absolute difference between x(k) and any pixel on its left side within the filtering range. Likewise, the value M_r is the maximum absolute difference between x(k) and any pixel on its right side within the filtering range. If both values M_l and M_r are equal to zero, then x(k) is in a flat area. In this case, the symmetry factor s(k) is set to 1 indicating no suppression is necessary because the high frequency compensation component d(k) would be zero. If one and only one of M_l and M_r is equal to zero, then the pixel luminance pattern is close to one of those shown in FIGS. 7(a)-(d). Therefore, it is very likely for shoot artifacts to appear at such position. In this case, the symmetry factor s(k) is set to 0 to fully suppress the gain of the pre-compensation. If neither of M_l and M_r is equal to zero, then two more values, A_l and A_r, are calculated as: $A_l=\frac{1}{N}\sum _{i=0}^{N-1}{d}_{0,-N+i},A_r=\frac{1}{N}\sum _{i=0}^{N-1}{d}_{N-i,0},$

where A_l is the average of the absolute difference values on the left side of x(k), and A_r is the average of the absolute difference values on the right side of x(k). The symmetry factor s(k) can then be calculated as: $s\left(k\right)=\{\begin{array}{cc}{A}_l/M_r& \mathrm{If}{M}_r\ge M_l\\ A_r/M_l& \mathrm{Otherwise}\end{array}.$

Similarly, a more strict condition can also be used to calculate the symmetry factor s(k) as: $s\left(k\right)=\{\begin{array}{cc}{m}_l/M_r& \mathrm{If}{M}_r\ge M_l\\ m_r/M_l& \mathrm{Otherwise}\end{array},$

where m_l is the minimum of the absolute difference values on the left side of x(k), and m_r is the minimum of the absolute difference values on the right side of x(k), wherein: m_l=min(d_0,−N,d_0,−N+1, . . . d_0,−l) m_r=min(d_N,0,d_N−1,0, . . . d_1,0).

In intensity checking, an intensity factor r(k) is computed as a monotonically increasing function of the absolute value of the high frequency component α(k). FIG. 8 shows an example monotonically increasing function 800 where if |α(k)| is less than a threshold ξ₁, then the intensity r(k) is set to 0. If |α(k)| is greater than a threshold ξ₂, then the intensity factor r(k) is set to 1. And, if ξ₁≦|α(k)≦ξ₂₆, then the intensity factor r(k) is linearly interpolated in a soft-switching manner. The thresholds ξ₁ and ξ₂ are pre-defined values satisfying 0≦ξ₁≦ξ₂.

With both the symmetry factor s(k) and the intensity factor r(k) obtained, the suppression factor g₂(k) can be calculated as: g₂(k)=1−(1−s(k))*r(k).

When r(k)=1, the high frequency component α(k) is strong, wherein g₂(k)=s(k). Therefore, shoot suppression is performed solely based on the symmetry factor s(k). When r(k)=0, the high pass filter output α(k) is considered very weak and the suppression factor g₂(k) takes a value of 1, which means that no suppression is applied.

The pre-compensation of the high frequency component in a video frame can also be implemented in a pre-compensation parallel structure 900 as shown in FIG. 9, where x(i,j) denotes the input two dimensional video frame signal and x′(i,j) denotes the pre-compensated video frame signal and (i,j) denotes the geometric position. The system 900 comprises a high frequency component extractor 902 for the horizontal direction, a shoot suppressor 904 for the horizontal direction, a high frequency component extractor 906 for the vertical direction, a shoot suppressor 908 for the vertical direction, and a noise detector 910.

First, the system 900 extracts the high frequency component in the horizontal direction (using exactor 902) and in the vertical direction (using exactor 906). The gain of each component (i.e., horizontal and vertical high frequency components) is then adjusted by the corresponding shoot suppressor in the either horizontal direction (using suppressor 904) or the vertical direction (using suppressor 908). Specifically, for the horizontal direction, the suppression factor g_2h(k) output from the suppressor 904 is multiplied by gain g_h, and the product (g_h·g_2h(k)) is further multiplied by the compensation component d_h(k) output from the extractor 902 to generate the product (g_h·d_h(k)·g_2h(k)).

For the vertical direction, the suppression factor g_2v(k) output from the suppressor 908 is multiplied by gain g_v, and the product (g_v·g_2v(k)) is further multiplied by the compensation component d_v(k) output from the extractor 906 to generate the product (g_v·d_v(k)·g_2v(k)). The products (g_h·d_h(k)·g_2h(k)) and (g_v·d_v(k)·g_2v, (k)) are added and further adjusted by the gain adjustment parameter g₁(k) from the noise detector 910 to obtain a compensation component [g₁(k)·(g_h·d_h(k)·g_2h(k))+(g_v·d_v(k)·g_2v(k))]. The obtained compensation component is then added to the original video frame x(i,j) to obtain a pre-compensated video frame x′(i,j) as: x′(i,j)=[g₁(k)·(g_v·d_v(k)·g₂v(k))+(g_v·d_v(k)·g₂v(k))]+x(i,j).

The pre-compensated video frame x′(i,j) can be fed into a traditional video scaler or any other video scaler to obtain the scaled video, as shown in FIG. 2.

The present invention has been described in considerable detail with reference to certain preferred versions thereof; however, other versions are possible. Therefore, the spirit and scope of the appended claims should not be limited to the description of the preferred versions contained herein.

Claims

1. In a video scaling method, the improvement comprising pre-compensation of the high frequency components of an input video signal before video scaling.

2. The method of claim 1 wherein the steps of pre-compensation include the steps of: performing high frequency component extraction to extract a high frequency compensation component from the input signal.

3. The method of claim 2 wherein the steps of pre-compensation include the steps of: determining a shoot suppression factor for adjusting the gain of the compensation to avoid shoot artifacts.

4. The method of claim 3 wherein the steps of pre-compensation include the steps of: determining a gain adjustment factor for adjusting the gain of the compensation to avoid enhancement of the noise component.

5. The method of claim 4 wherein the steps of pre-compensation include the steps of: applying the shoot suppression factor and the gain adjustment factor to the high frequency compensation component to generate a compensation factor.

6. The method of claim 5 further comprising the steps of combining the compensation factor with the input signal to generate a pre-compensated signal.

7. The method of claim 6 further comprising the steps of scaling the pre-compensated signal to obtain a scaled signal, whereby sharpness of the scaled signal is maintained.

8. The method of claim 2 wherein the step of performing high frequency component extraction to extract a high frequency compensation component from the input signal, includes the steps of: passing the input signal through a high-pass filter to obtain a high frequency component α(k); and determining the high frequency compensation component as a function of the high frequency component α(k).

9. The method of claim 8 wherein the step of determining the high frequency compensation component as a function of the high frequency component α(k), further includes the steps of adjusting high frequency component α(k) such that: if 0≦|α(k)|≦t1, then amplitude of α(k) is reduced to avoid boosting noise; if t1<|α(k)≦t2, then α(k) is bypassed; if t2<|α(k)|≦t3, then α(k) is increased; and if |α(k)|>t3, then α(k) is bypassed to avoid overcompensation; wherein the thresholds t1, t2 and t3 satisfy 0≦t1≦t2≦t3.

10. The method of claim 9, wherein reducing amplitude of α(k) includes the steps of multiplying α(k) by an adjustment factor to generate the signal β(k) as:

11. The method of claim 9 wherein the step of determining the high frequency compensation component d(k) further includes the steps of determining d(k)=β(k)−α(k), wherein β(k) is a function of α(k).

12. The method of claim 9, wherein increasing amplitude of α(k) includes the steps of applying an adjustment factor to α(k) generate the signal β(k) as:

13. The method of claim 12 wherein the step of determining the high frequency compensation component d(k) further includes the steps of determining d(k)=β(k)−α(k), wherein β(k) is a function of α(k).

14. The method of claim 4 wherein the gain adjustment factor is a monotonically increasing function of the local standard deviation.

15. The method of claim 8 further comprising the steps of performing shoot suppression.

16. The method of claim 15 wherein the steps of performing shoot suppression further includes the steps of: performing symmetry checking by checking the luminance variation of pixels on both sides of the current pixel within the filtering range; based on the checking result, performing higher suppression for patterns that are more likely to cause overshoot/undershoot.

17. The method of claim 15 wherein the steps of performing shoot suppression further includes the steps of: performing intensity checking by checking the magnitude of the high frequency component α(k), performing suppression as a function of the intensity of α(k) at each input signal pixel position.

18. The method of claim 1 wherein the input signal comprises a two dimensional array of pixels, such that the pre-compensation of the high frequency components of the input signal is performed in the horizontal direction to determine a horizontal compensation factor, and in the vertical direction to determine a vertical compensation factor.

19. The method of claim 18 further comprising the steps of combining the horizontal and vertical compensation factors with the input signal to generate a pre-compensated signal.

20. A video scaling system for scaling of an input video signal, comprising: a pre-compensator that pre-compensates for the high frequency components of the input video signal to generate a pre-compensated signal prior to video scaling.

21. The system of claim 20 wherein the pre-compensator comprises an extractor that performs high frequency component extraction to extract a high frequency compensation component from the input signal.

22. The system of claim 21 wherein the pre-compensator further comprises a shoot suppressor that determines a shoot suppression factor for adjusting the gain of the compensation to avoid shoot artifacts.

23. The system of claim 22 wherein the pre-compensator further comprises a noise detector that determines a gain adjustment factor for adjusting the gain of the compensation to avoid enhancement of the noise component.

24. The system of claim 23 wherein the pre-compensator further applies the shoot suppression factor and the gain adjustment factor to the high frequency compensation component to generate a compensation factor.

25. The system of claim 24 wherein the pre-compensator further combines the compensation factor with the input signal to generate the pre-compensated signal.

26. The system of claim 25 further comprising a scaler that scales the pre-compensated signal to obtain a scaled signal, whereby sharpness of the scaled signal is maintained.

27. The system of claim 21 wherein: the extractor performs high frequency component extraction to extract a high frequency compensation component from the input signal by passing the input signal through a high-pass filter to obtain a high frequency component α(k); and the pre-compensator further determines the high frequency compensation component as a function of the high frequency component α(k).

28. The system of claim 27 wherein the pre-compensator determines the high frequency compensation component as a function of the high frequency component α(k) by adjusting high frequency component α(k) such that: if 0≦|α(k)|≦t1, then amplitude of α(k) is reduced to avoid boosting noise; if t1<|α(k)|≦t2, then α(k) is bypassed; if t2<|α(k)|≦t3, then α(k) is increased; and if |α(k)|>t3, then α(k) is bypassed to avoid overcompensation; wherein the thresholds t1, t2 and t3 satisfy 0≦t1≦t2≦t3.

29. The system of claim 28 wherein the pre-compensator reduces the amplitude of α(k) by multiplying α(k) by an adjustment factor to generate the signal β(k) as:

30. The system of claim 28 wherein the pre-compensator determines the high frequency compensation component d(k) by determining d(k)=β(k)−α(k), wherein β(k) is a function of α(k).

31. The system of claim 28 wherein the pre-compensator increases the amplitude of α(k) by applying an adjustment factor to α(k) generate the signal β(k) as:

32. The system of claim 31 wherein the pre-compensator determines the high frequency compensation component d(k) by determining d(k)=β(k)−α(k), wherein β(k) is a function of α(k).

33. The system of claim 23 wherein the gain adjustment factor is a monotonically increasing function of the local standard deviation.

34. The system of claim 27 wherein the shoot suppressor further performs shoot suppression.

35. The system of claim 34 wherein the shoot suppressor performs shoot suppression further by: performing symmetry checking by checking the luminance variation of pixels on both sides of the current pixel within the filtering range; and based on the checking result, performing higher suppression for patterns that are more likely to cause overshoot/undershoot.

36. The system of claim 34 wherein the shoot suppressor performs shoot suppression further by: performing intensity checking by checking the magnitude of the high frequency component α(k), performing suppression as a function of the intensity of α(k) at each input signal pixel position.

37. The system of claim 20 wherein the input signal comprises a two dimensional array of pixels, such that the pre-compensation of the high frequency components of the input signal is performed in the horizontal direction to determine a horizontal compensation factor, and in the vertical direction to determine a vertical compensation factor.

38. The system of claim 37 wherein the pre-compensator further combines the horizontal and vertical compensation factors with the input signal to generate the pre-compensated signal.

39. The system of claim 20, wherein the pre-compensator performs cascade pre-compensation.

40. The system of claim 20, wherein the pre-compensator performs parallel pre-compensation.