Motion adaptive median filter for interlace to progressive scan conversion

[0002] Embodiments of the present invention relate to conversion of interlaced scan video signals to progressive scan video signals. More particularly, embodiments of the present invention relate to embodiments of apparatus and methods relating to a motion adaptive median filter for interlace to progressive scan conversion.

[0003] Known methods can convert an interlaced scan video signal to a progressive scan video signal. An interlaced scan video signal can include a succession of fields, each field including a plurality of scan lines. Two successive fields of an interlaced scan video signal can define a frame where the first field includes the odd scan lines (e.g., scanning lines 1, 3, 5, 7, etc.) and the second field includes the even scan lines (e.g., 2,4,6,8, etc.).

[0004] A de-interlacing method can generate an interpolated line of interpolated pixels between two successive lines of a field. Examples of known methods of de-interlacing an interlaced scan video signal include temporal interpolation and spatial interpolation. In an area (e.g., an area of a video picture generated on a video display by a video signal, etc.) with no or low motion, temporal interpolation can be preferred. In an area with high motion, spatial interpolation can be preferred. In low/no motion areas, spatial interpolation (e.g., a low pass filter) typically tends to make the picture softer while temporal interpolation can maintain picture details. In high motion areas, temporal interpolation typically tends to introduce motion artifacts, while spatial interpolation avoids these. In view of the foregoing, it can be appreciated that a substantial need exists for methods and apparatus which can advantageously provide motion adaptive median filtering for interlace to progressive scan conversion.

[0005] FIG. 1 shows pixels in accordance with an embodiment of the present invention.

[0006] FIG. 2 shows pixels in accordance with an embodiment of the present invention.

[0007] FIG. 3 shows an apparatus in accordance with an embodiment of the present invention.

[0008] FIG. 4 shows a more detailed illustration of a portion of the apparatus illustrated in FIG. 3.

[0009] FIG. 5 shows an illustration of a median filter in accordance with an embodiment of the present invention.

[0010] FIG. 6 shows an apparatus in accordance with an embodiment of the present invention.

[0011] FIG. 7 shows pixels in accordance with an embodiment of the present invention.

[0012] According to an embodiment of the present invention, pixel values from a current field, a previous field, and a next field of an interlaced scan video signal can be used to determine a motion estimate value, a spatial interpolation value, and a temporal interpolation value. An weighted average value can be output by a soft switch having the motion estimate value as a weighting factor input and the spatial interpolation value and the temporal interpolation value as inputs. A predicted pixel value of a progressive frame generated based at least in part on the interlaced scan video signal can be output by a median filter having as inputs the weighted average value and the pixel values from the current field, the previous field, and the next field.

[0013] With the emergence of digital television (DTV) and high definition television (HDTV), scan and format conversion between the Advanced Television Systems Committee (ATSC) input formats and other output formats can become necessary for a Moving Pictures Expert Group (MPEG) decoder. Interlaced video can have limitations that produce visible artifacts (e.g., line crawl, flicker, etc.) Artifacts can be particularly objectionable when an input is a standard definition (SD) sized interlaced video signal (i.e., 720×480, 130) and an output is high definition (HD) video (e.g., 1080I, 720P) or progressive SD (480P). Displaying 480I input with good quality on high quality SD or HD monitors can be important because much existing material (e.g., video material) is in interlaced format and interlaced television broadcast may be around for an extended period of time. Embodiments of the present invention can advantageously de-interlace interlaced video for upscaling to HD formats, displaying on progressive displays, etc.

[0014] FIG. 1 shows pixels in accordance with an embodiment of the present invention. Methods and apparatus in accordance with embodiments of the present invention can generate an output progressive frame including predicted O pixel 155 of a predicted scan line 122 based at least in part on pixel data from previous field 110, current field 120, and next field 130 of an interlaced scan video signal. Current field 120 can have an N pixel 152 of a scan line 121 and an S pixel 154 of a scan line 123. N pixel 152 can be the pixel above the O pixel 155 being predicted in the current field 120, and S pixel 154 can be the pixel below the pixel being predicted in the current field 120. W pixel 151 of scan line 22 of previous field 110 can be the pixel in the previous field (of opposite polarity) at the same spatial location as O pixel 155, and E pixel 153 of scan line 222 of next field 130 can be the pixel in the next field (of opposite polarity) at the same spatial location as O pixel 155.

[0015] FIG. 2 shows pixels in accordance with an embodiment of the present invention. An output progressive frame 200 can include predicted scan line 122 between scan line 121 and scan line 123. In an embodiment, W pixel 151 of previous field 110, predicted O pixel 155, and E pixel 153 of next field 130 are pixels in different fields that have the same spatial location. In accordance with an embodiment of the present invention, a de-interlacing method outputs a progressive frame by de-interlacing each current field of an input interlaced pixel sequence based at least in part on data from the current field, the previous field, and the next field. In an embodiment, when a top field is de-interlaced, the previous and next field are bottom fields, and when a bottom field is de-interlaced, the previous and next field are top fields. Half of the line in the output progressive frame can be identical to the lines in the current field, and the other half can be generated in accordance with embodiments of the present invention.

[0016] FIG. 3 shows an apparatus in accordance with an embodiment of the present invention. A predicted pixel O signal can be generated based on pixel sequence P_in. Pixel sequence P_incan include a sequence of pixel values and be input to field delay 310 and field delay 311 to generate pixel E, W, N, and S signals, where pixel N is the pixel above pixel O in the current field (i.e., pixel N and pixel O have the same horizontal position), pixel S is the pixel below pixel O in the current field (i.e., pixel S and pixel O have the same horizontal position), pixel W is the pixel in the previous field (of opposite polarity) at the same spatial location as O (i.e., pixel W and pixel O have the same vertical and horizontal position), and pixel E is the pixel in the next field (of opposite polarity) as the same spatial location as 0 (i.e., pixel W and pixel O have the same vertical and horizontal position). Averaging circuit 330 can be coupled to field delay 311 and generate a temporal average signal T_avbased on pixel E and pixel W signals. An averaging circuit can include an adder and a divider, an adder and a shift register, etc. The term “coupled” encompasses a direct connection, an indirect connection, etc. Averaging circuit 340 can generate a spatial average signal Say based on pixel N and pixel S signals.

[0017] Motion detect circuit 320 can generate a motion signal M based on pixel E, W, N, and S signals. In one embodiment, motion signal M is a motion estimation signal having a range from zero to one. Motion signal M can equal CLIP (K_s*(|E−W|/(|E−W|+|N−S|))). CLIP (n) is clip to 1 function where CLIP (n)=1 when (n>=1), and CLIP (n)=n, when (n<1). K_sis a slope control factor that can control how quickly motion signal M changes from 0 to 1 as motion (|E−W|) increases for a given amount of detail (|N−S|) in a picture. In one embodiment, a slope factor of 32 provided good results.

[0018] Soft switch 350 can output an interpolated pixel i signal based on receiving temporal average signal T_avand spatial average signal S_avas signal inputs and motion signal M as a weighting factor input. Interpolated pixel i signal can be a weighted average of T_avand S_avbased on the value of motion signal M. i can be equal to T_avwhen motion signal M is zero, and i can be equal to S_avwhen motion signal M is one. In one embodiment, interpolated pixel signal i equals (M*(N+S)/2)+((1−M)*(E+W)/2).

[0019] Median filter 360 can output predicted pixel O signal based at least in part on median filtering pixel i, E, W, N, and S signals. In one embodiment, median filter 360 is a 5-point median filter. In an embodiment of the present invention, a 5-point median filter having pixel i, E, W, N, and S signals as inputs provided advantageous results in low motion areas because the filter was partially biased toward using one of pixel E, W, and i signals, where pixel i signal tends toward T_av. In an embodiment of the present invention, a 5-point median filter having pixel i, E, W, N, and S signals as inputs provided advantageous results in high motion areas because the filter was partially biased toward using one of pixel N, S, and i signals, where pixel i signal tends toward S_av.

[0020] Embodiments of the present invention provide advantageously prediction of pixel values for de-interlacing interlaced scan video signals, generating progressive frames from interlaced scan video signals, etc. Embodiments of the present invention can be simpler than motion vector based de-interlacing schemes. Embodiments of the present invention can provide superior results than known median filter based schemes. Simple three-point median filters typically can not preserve high details. Weighted median filters can require large median filters and can be hard to characterize and predict. A Finite Impulse Response (FIR) median hybrid filter may not be motion adaptive. Whereas known de-interlacing methods introduce new pixel values that are computed as a weighted sum of surrounding (e.g., spatial, temporal, etc.) pixels, embodiments of the present invention can introduce new interpolated values only when the interpolated value is found to be a very good representation of the surrounding pixels.

[0021] Embodiments of the present invention can be motion adaptive, preserve high details in no/low motion areas, reduce interlace artifacts, and generally perform better than vertical averaging in high motion areas.

[0022] FIG. 4 shows a more detailed illustration of a portion of the apparatus illustrated in FIG. 3. In one embodiment of the present invention, a soft switch 450 can output an interpolated pixel i signal based on receiving temporal average signal T_avand spatial average signal S_avas signal inputs and motion signal M as a control input. In one embodiment, interpolated pixel i signal equals (M*(S_av−T_av))+T_avgenerated by subtractor 451, multiplier 452, and adder 453.

[0023] FIG. 5 shows an illustration of a median filter in accordance with an embodiment of the present invention. Embodiments of the present invention do not require a full 5-point median filter, and can be efficiently implemented in hardware using comparators (e.g., minimum comparators, maximum comparators). Interpolated pixel i signal, being a weighted sum of pixel N, S, E, and W signals, will be greater than the minimum (N, S, E, W) and less than maximum (N, S, E, W). Accordingly, in one embodiment, a five-input median filter 560 can be implemented as MEDIAN (i, min(max(n,s), max(e,w)), max(min(n,s), min(e,w))) and require only a 3-point median filter. MAX 561 can output the greater of N and S, and MAX 562 can output the greater of E and W. MIN 563 can output the lesser of N and S, and MIN 564 can output the lesser of E and W. MIN 565 can output the lesser of the output of MAX 561 and MAX 562. MAX 566 can output the greater of MIN 563 and MIN 564. 3-point median filter 567 can output the median of i, the output of MIN 565 (i.e., min(max(n,s), max(e,w) )), and the output of MAX 566 (i.e., max(min(n,s), min(e,w))).

[0024] Embodiments of the present invention can be advantageous motion adaptive filters. In accordance an embodiment of the present invention, at zero motion, pixel i equals pixel E (and pixel E equals pixel W), and the filter output pixel O equals pixel i (i.e., equals pixel E and pixel W). At low motion, pixel i approaches T_av, and the filter output pixel O is biased toward one of pixel E, W, or i. As motion increases, pixel i approaches S_av, and the filter output pixel O is biased toward one of pixel N, S, or i. In low motion areas when pixels are well co-related, pixels E, W, and i can fall between pixel N and S, and the filter output pixel O is the median of pixel E, W, or i. In low motion areas when pixels are not well co-related, pixels E, W, and i do not fall between pixel N and S, and the filter output pixel O is biased towards choosing the one of pixel E, W, or i that is closest to S_av. In high motion areas, pixel i approaches T_av, the filter output pixel O is biased towards choosing the one of pixel N, S, or i that is closest to T_av. The filter output is not always using one of i, S_av, or T_av, but can choose one of pixels N, S, E, and W (i.e., can allow pixel repeating), when one of pixels N, S, E, and W is better representative of the set of pixels N, S, E, W, and i. Other known median filter based de-interlacing schemes do not allow for pixel repeating. Embodiments of the present invention can advantageously reduce interlace artifacts and preserve details in stationary (e.g., no motion, low motion, etc.) areas.

[0025] FIG. 6 shows an apparatus in accordance with an embodiment of the present invention. Motion adaptive median filter 600 can advantageously further reduce artifacts in high motion areas. High motion areas of concern include those where MIN (EB,W)>>MAX (N,S) and MAX (E, W)<<MIN (N,S). Under such conditions, a median filter can pick the one of pixel N and S that is closer to T_avwhen T_avis considerably different from pixel N and S, and thereby introduce visible artifacts. Embodiments of the present invention can address such conditions. The embodiment illustrated in FIG. 3 can be modified as illustrated in FIG. 6 to reduce artifacts in high motion areas.

[0026] A pixel sequence P_incan be received by signal processing circuit 610 to generate temporal average signal T_av, spatial average signal S_av, and motion signal M. Signal processing circuit 610, in one embodiment, includes field delays 310 and 311, motion detect circuit 320, and averaging circuits 330 and 340 described in FIG. 3. In one embodiment, the motion range of the motion detect 320 can be doubled by halving the slope control by setting K_sto 16.

[0027] A second motion signal M2 can be generated by motion signal processing circuit 660. In one embodiment, multiplexer 664 outputs one of two signals generated based at least in part on motion signal M. Multiplier 663 can generate a signal of 2*M as an input to multiplexer 664, and subtractor 661 and multiplier 662 can generate a signal of 2*(M−0.5) as an input to multiplexer 664. Mulitplexer 664 can output second motion signal M2 as 2*M when motion signal M is <=0.5, and can output second motion signal M2 as 2*(M−0.5) when motion signal M is >0.5.

[0028] Motion adaptive median filter 600 (filter 600) can operate in a first mode when motion signal M<0.5 and in a second mode when motion signal M>=0.5. In the first mode of operation of filter 600 (e.g., during no motion, low motion, etc.), when M<0.5, multiplexer 610 outputs T_avto soft switch 650. S_avis also an input to soft switch 650, and soft switch 650 outputs pixel i signal (i.e., the weighted average of T_avand S_avwhen M<0.5) as second motion signal M2 varies from zero to one. Multiplexer 620 outputs pixel i signal to median filter 360, and median filter 360 outputs the median of pixel W, E, N, S, i signals, i.e., median signal MED. Multiplexer 630 outputs the median signal MED as the pixel O signal when M<0.5.

[0029] In the second mode of operation of filter 600 (e.g., during high motion, etc.), when M>=0.5, multiplexer 620 outputs S_avto median filter 360, and median filter 360 outputs the median of signals W, E, N, S, S_av, i.e., median signal MED. Multiplexer 610 outputs MED to soft switch 650. S_avis also an input to soft switch 650, and soft switch 650 outputs pixel i signal (i.e., the weighted average of MED and S_avwhen M>=0.5) as second motion signal M2 varies from zero to one. Multiplexer 630 outputs the pixel i signal as the pixel O signal when M>=0.5. In another embodiment, multiplexors 610, 620, 630, 664 can be a hard switch, a selector, etc.

[0030] FIG. 7 shows pixels in accordance with an embodiment of the present invention. In an embodiment of the present invention, S_avis generated by a four tap vertical filter. Rather than averaging pixel N and S (e.g., pixel N 152 and pixel S 154 of FIG. 1) to generate S_avas part of predicting pixel O 155 of scan line 722, a four tap vertical filter generates S_avbased on pixel N2750 of scan line 719, pixel N 752 of scan line 721, pixel S 754 of scan line 723, and pixel S2756 of scan line 725. In one embodiment, a four tap vertical filter generates S_avas equaling ((−{fraction (1/16)})*N2)+(({fraction (9/16)})*N)+(({fraction (9/16)})*S)+((−{fraction (1/16)})*S2). Embodiments of a four tap vertical filter can be implemented using shifts and adds, and embodiments of the present invention generating S_avfrom a four tap vertical filter can improve picture quality in high motion areas.

[0031] Embodiments of methods and apparatus to de-interlace interlaced scan video signals have been described. In the foregoing description, for purposes of explanation, numerous specific details are set forth to provide a thorough understanding of the present invention. It will be appreciated, however, by one skilled in the art that the present invention may be practiced without these specific details. In other instances, structures and devices are shown in block diagram form. Furthermore, one skilled in the art can readily appreciate that the specific sequences in which methods are presented and performed are illustrative and it is contemplated that the sequences can be varied and still remain within the spirit and scope of the present invention.

[0032] Embodiments of the present invention encompass systems incorporating a motion adaptive median filter including systems such as a video system of a computer, a television, an HDTV receiver/decoder, an SDTV receiver/decoder, an MPEG decoder, digital versatile disc (DVD) player, portable video player, home video system, digital video receiver, etc.

[0033] In the foregoing detailed description, apparatus and methods in accordance with embodiments of the present invention have been described with reference to specific exemplary embodiments. Accordingly, the present specification and figures are to be regarded as illustrative rather than restrictive.