CN1529500A - Three-dimensional video format conversion method based on motion adaption and marginal protection - Google Patents

Abstract

In the invention, video format conversion is completed based on noninterlace. In procedure of converting interlaced sequence to noninterlaced sequence, strategy of edge protection and motion self-adapting strategy are utilized to modify result of interpolation in field. Steps for converting video format are: non-interlaced treatment of interlaced sequence is carried out for video signal inputted in interlaced format first. In procedure of non-interlaced treatment, following operations is executed: summarizing 2D interpolation in field of edge protection and weighted Cubic 2D interpolation; then, summarizing the said above 2D summarized result and weighted 3D result of motion self-adapting interpolation. After non-interlaced treatment, conversion of frame rate is carried out. Finally, converting resolution of frame outputs video sequence in specific frame rate or field rate. The invention uses handy algorithm to achieve conversion task, providing full functions and wide applicable region.

Description

Based on Motion Adaptive and edge-protected 3 D video format conversion method

Technical field

What the present invention relates to is a kind of method of video image format conversion, specifically is a kind of based on Motion Adaptive and edge-protected three-dimensional heterogeneous interpolation video format conversion method.Belong to electronic information field.

Background technology

Video format conversion be guarantee the video of different resolution and frame per second form can be on the display of a certain fixedly display format the necessary means of normal play.Along with the popularization of digital video services such as Digital Television, the requirement of video format conversion technology is also being improved constantly.The video format conversion technology comprises the interlaced/progressive conversion, and the content of resolution conversion and three aspects of frame per second conversion can adopt different interpolating methods to realize.

Find that by literature search the Chinese patent name is called: " video format conversion apparatus and method ", application number is: 94105360.The movable information that the method that this patented technology proposes utilizes video decode to obtain is changed processing line by line to the interlacing list entries, thereby finishes the Presentation Function line by line of interlacing list entries.According to the algorithm that this patented technology proposes, the interpolation pixel in the current display frame is obtained by the motion compensation information of consecutive frame before and after it or field.The motion compensation information that this patent is used is the motion estimation result in the compressed encoding, and this movable information is more coarse, is the requirement that is difficult to reach format conversion generally speaking.The method that provides of this patent is only finished interlacing is in addition changeed function line by line, and other resolution conversions and frame per second conversion are not all related to, and can't finish the conversion of standard definition video to high definition video, so its format conversion function is incomplete.

Summary of the invention

The objective of the invention is to overcome deficiency of the prior art; provide a kind of based on Motion Adaptive and edge-protected three-dimensional heterogeneous interpolation video format conversion method; make it by adopting simple and practical algorithm, finish the interlaced/progressive conversion in functions such as interior resolution conversion and frame per second conversion.

The present invention is achieved by the following technical solutions; video format conversion of the present invention is to finish on basis line by line; in with the interlaced process that transfers sequence line by line to, adopted edge-protected strategy and Motion Adaptive strategy that the result of field interpolation is revised; the process of video format conversion is: the vision signal to the interlacing input is carried out the interlaced processing of change line by line earlier; when changing line by line; adopt two-dimensional interpolation and the weighted sum of Cubic two-dimensional interpolation in edge-protected; do three-dimensional weighted sum with it with the motion-adaptive interpolation result again, to improve the adaptability of interpolation result to non-motion component.Sequence after changing is line by line carried out the frame per second conversion, do the conversion of frame intrinsic resolution at last, the video sequence of a frame per second or a rate is specified in output.

Below the present invention is further illustrated, particular content is as follows:

1, interlaced change line by line

Do three kinds of processing to interlaced: an one dimension Cubic interpolation of vertical direction at first being doubles the vertical resolution of input video sequence; Secondly the interpolation pixel is done edge-protected interpolation, and with the weighted sum as a result of one dimension Cubic interpolation, obtain the result of field interpolation; At last the interpolation pixel in interlaced is carried out motion detection, the Motion Adaptive weight coefficient of determining according to motion detection obtains the pixel line by line through three-dimensional interpolation with the result and the weighted sum of motion reference pixel of field interpolation.Be implemented as follows:

(1) one dimension Cubic interpolation:

The Cubic interpolation kernel is a kind of interpolation kernel relatively more commonly used.Its pass band damping is little, and transition band is narrow, and sidelobe decay is big, so it is less to do the distortion that interpolation kernel introduces with Cubic.When carrying out interlaced change line by line, only need the image of input is carried out the interpolation of vertical direction, so the algorithm that the present invention relates to adopts the Cubic function interpolation filter of 4 taps, 128 phase places input signal to be carried out the one dimension interpolation of vertical direction.The coefficient of interpolation filter calculates by following formula:

The computing formula that input imagery is carried out the Cubic function interpolation is in vertical direction:

$I_{\mathrm{cubic}}^{*}(x+1,y,n)=\underset{i=-1}{\overset{2}{\mathrm{\Σ}}}I(x+2i,y,n)u_v\left(\frac{2i-1}{2}\right)------\left(2\right)$

Wherein

Be in the n field in the position (x+1 y) locates, and is the two-dimensional interpolation result of interpolation kernel with the Cubic function, x={1 wherein, 3,5, Λ } or { 0,2,4, Λ };

I (x, y, n) be in the n field in the position (x, the known pixel of y) locating;

I ∈ [1,2] is the vertical offset of the pixel of participation interpolation;

Be the coefficient of one dimension 4 taps 128 phase place Cubic function interpolation filters, subscript v represents the interpolation filter of vertical direction.

(2) edge-protected interpolation

The utilization of Cubic interpolation be interpolation point level with vertical field in information, so poor to the peripheral adjustment of visual medium dip.Introduced edge-protected interpolation algorithm among the present invention for this reason.

Edge-protected interpolation is made up of two parts: the edge related side is to judgement, the edge directional interpolation of being correlated with.At first the edge related side who judges interpolation point according to the known pixel interpolation difference around the interpolation point to (be 0 ° to the right with level, the related side to set be: { 45 °, 90 °, 135 ° }).Make progress the mean value of known pixel as edge-protected interpolation result with this related side then, and determine the weight coefficient of edge-protected interpolation with the related side to the difference of pixel.

Edge-protected interpolation and two-dimentional Cubic interpolation have constituted the result of an interior two-dimensional interpolation together.Promptly

$I_{\mathrm{infra}}^{*}(x+1,y,n)=(1-w_{\mathrm{edge}})I_{\mathrm{cubic}}^{*}(x+1,y,n)+w_{\mathrm{edge}}{I}_{\mathrm{edge}}^{*}(x+1,y,n)------\left(3\right)$

Wherein

Be to adopt the interpolation result that the two-dimensional interpolation algorithm obtains in the field in the n field of input, this two-dimensional interpolation comprises Cubic two-dimensional interpolation and edge-protected interpolation, x={1,3,5, Λ } or { 0,2,4, Λ };

It is the interpolation result that in the n field of input, adopts two-dimentional Cubic interpolation to obtain;

It is the interpolation result that in the n field of input, adopts edge-protected method to obtain;

w _EdgeBe that (x+1, edge-protected weight coefficient y) is by the make progress difference decision of known pixel of the related side of this point for interpolation picture element in the n field of input.

(3) motion-adaptive interpolation

Go the interlacing effect for what improve static content, the present invention has adopted the method for Motion Adaptive that the motion and standstill content is handled respectively.The basis of Motion Adaptive algorithm is motion detection and motion determination.Several determination methods that the present invention adopts are based on the algorithm of field.If, think that then the pixel when the front court moves, otherwise think that the pixel of working as the front court is static when the difference of certain pixel of front court and the interpolation pixel of reference field corresponding position surpasses certain thresholding.When the pixel in the front court was carried out detection of motion, the algorithm that patent of the present invention adopts gave the motion weight coefficient of this pixel.Therefore weighted sum obtains the interpolation pixel in progressive frame according to the motion weight coefficient by the corresponding pixel with reference field of field interpolation (comprising two-dimentional Cubic interpolation and edge-protected interpolation).Adopt the interpolation formula of Motion Adaptive technology to be:

$I_{\mathrm{infra}}^{*}(x+1,y,n)={(1-w_{\mathrm{edge}})I}_{\mathrm{cubic}}^{*}(x+1,y,n)+w_{\mathrm{wdge}}{I}_{\mathrm{edge}}^{*}(x+1,y,n)------\left(3\right)$

Wherein,

Be that the x+1 that obtains through the field value of interleaving in the n field is capable, y row pixel, x={1,3,5, Λ } or { 0,2,4, Λ };

Be that the x+1 that adopts field interpolation (comprising two-dimentional Cubic interpolation and edge-protected interpolation) to obtain in the n field is capable, y row pixel;

(x+1, y are that x+1 in the n-1 field is capable n-1) to I, y row pixel;

M_ratio is the motion weight coefficient, is judged as static pixel, then m_ratio=0 if work as the front court pixel through motion detection.m_ratio∈[0，1]。

2, frame per second conversion

For reducing hard-wired complexity, the method that the present invention adopts frame to repeat is finished the frame per second conversion, promptly with output frame rate output progressive frame, in case of necessity progressive frame is carried out repetition, to satisfy the requirement of bit rate output.

The input of frame per second conversion is a progressive scanning sequence; The method that repeats is adopted in the frame per second conversion nearby, the content that is output frame equals the most close with it in time input signal line by line, therefore the required frame memory of frame per second conversion only need store the progressive video signal of two frame single-definitions, and is lower to the requirement of memory.

3, resolution conversion

The input of resolution conversion be change line by line and frame per second for specifying the vision signal of output frame rate; Resolution conversion adopts two-dimentional Cubic leggy interpolation algorithm in the frame, can finish the resolution conversion of arbitrary proportion.

The present invention adopts the Cubic function two-dimensional interpolation filter under 4 * 4 windows that progressive-scan signal is carried out resolution conversion.Because two-dimentional Cubic interpolation kernel can be decomposed into the product of two one dimension interpolation kernel, so in order to reduce the memory space of heterogeneous two-dimensional interpolation filter coefficient, the present invention is decomposed into the cascade form of twice one dimension interpolation with a two-dimentional Cubic interpolation, and what the one dimension interpolation adopted is the Cubic interpolation kernel of 128 phase places, 4 taps.Two dimension Cubic function interpolation formula is:

$I_{\mathrm{scaing}}^{*}(x^{*},y^{*},n)=I_{\mathrm{scaing}}^{*}(x+\mathrm{\Δx},y+\mathrm{\Δy},n)=\underset{i=-1}{\overset{2}{\mathrm{\Σ}}}\underset{j=-1}{\overset{2}{\mathrm{\Σ}}}{I}_{\mathrm{inter}}^{*}(x+i,y+j,n)u_j\left(p_{\mathrm{mh}}\right)u_i\left(p_{\mathrm{nv}}\right)-----\left(5\right)$

Wherein $I_{\mathrm{scaling}}^{*}(x^{*},y^{*},n)=I_{\mathrm{scaling}}^{*}(x+\mathrm{\Δx},y+\mathrm{\Δy},n)$ Be that (x+ Δ x, y+ Δ y) locates to carry out with the interpolation filter of Cubic function the result of interpolation in the position in the image line by line at the n frame;

Be in adopting and between the n frame that obtains of interpolation algorithm participate in the pixel gray scale of interpolation line by line in the image in 4 * 4 windows, i, j=[-1,2] be the vertical and horizontal offset that participates in the interpolation pixel in 4 * 4Cubic interpolation window, x={1,2,3, Λ } expression changes each row in the image line by line through last joint, y={1,2,3, Λ } represent to change respectively being listed as in the image line by line through last joint;

u _i(p _Nv), u _j(p _Mh) to be illustrated respectively in vertical phase place be p _NvWith horizontal phase be p _MhThe time each interpolation point the Cubic interpolation coefficient, $p_{\mathrm{nv}}\≅\mathrm{\Δx}\·N_p,$ $p_{\mathrm{mh}}\≅\mathrm{\Δy}\·N_p,$ N _pNumber of phases for interpolation filter.

The present invention has substantive distinguishing features and marked improvement.The present invention is used to finish the video image format conversion that comprises resolution conversion and frame per second conversion, adopt in simple and between algorithm, the vision signal of interlacing input is changed line by line; Adopt the simple method that repeats that the standard definition video signal of changing is line by line carried out the frame per second conversion; Progressive signal is carried out in the frame two-dimentional Cubic function leggy filtering interpolation finish resolution conversion.The algorithm that the present invention relates to is simple and practical, and complete function is owing to adopted comparatively meticulous interlacing to change means line by line, for follow-up frame rate conversion and resolution conversion have been laid good basis; For data-storing require lowly, even output signal is a high definition video, also only need the form line by line of storage two frame standard definition video signals during the conversion of conducting frame rate, relatively be fit to hardware circuit and realize.The pictorial detail of exporting after the interpolation is stable, and image is fine and smooth, and the rapid movement image is shake and flicker not; Interpolation filter has adopted the leggy design, can finish multiple different video format conversion, so its range of application is wider, can be used in the control device of multiple digital displays such as LCOS, LCD, DLP.

Description of drawings

Fig. 1 is the The general frame that the present invention relates to algorithm

Fig. 2 is 4 * 4 window Cubic interpolation schematic diagrames

Fig. 3 is the motion detection schematic diagram

Fig. 4 is that the edge related side is to judging and the interpolation schematic diagram

Fig. 5 is a frame per second conversion schematic diagram

Fig. 6 is edge-protected interpolation weight coefficient curve

Embodiment

Provide following examples to the specific embodiment of the present invention in conjunction with the accompanying drawings:

As shown in the figure, Fig. 1 is the The general frame that the present invention relates to algorithm.The video format conversion process is as follows: the interleaved input imagery 16 of single-definition is carried out format conversion, and establishing its sequence number is n.Input imagery 16 links to each other with five passages, and wherein four is luminance channel, and one is the carrier chrominance signal passage; In four luminance channels, first connects a memory 1, so that import the motion detection reference signal of field as next; Second links to each other with motion detection and interpolating unit 2; Expanding (I) unit 3 with image for the 3rd links to each other; The 4th with the edge related side to judging and interpolating unit 4 links to each other.Carrier chrominance signal Cr _nAnd Cb _nLink to each other with one dimension interpolation device 11.The luminance signal of n-1 input field exists in the memory 1.The luminance component Y of n field input signal _nLuminance component Y with n-1 field input signal _N-1Be input in motion detection and the interpolating unit 2, this unit is according to Y _nAnd Y _N-1The difference of middle corresponding blocks judges that whether the interpolation picture element moves, and provides the coefficient w of motion-adaptive interpolation weighting _MotionWith motion-adaptive interpolation amount (1-w _Motion) I (x+1, y, n-1).The while input signal Y _nCarry out the image border expansion through image expansion I unit 3, so that the edge pixel of image is carried out interpolation.Carry out the stationary phase interpolation of vertical direction among the signal input one dimension interpolation device I (4) after the expansion, this one dimension interpolation device is 4 taps, 128 phase place Cubic interpolation filters, and the vertical resolution of the luminance signal 17 of 4 outputs is the twice of input signal vertical resolution.Enter the brightness signal Y of edge related side to judgement and interpolating unit 5 _nThe edge related side who in this unit, judges interpolation point to, and export the related side to weight coefficient and edge-protected interpolation weight coefficient w _Edge(24) and edge-protected interpolation weighted volumes (1-w _Edge) I ^* _Edge(x+1, y, n) (25), wherein I ^* _Edge(x+1, y, n) be the related side to interpolation result.One dimension Cubic interpolation result 17 multiplies each other in multiplier 12 with edge-protected weight coefficient 24, and its result and the addition in adder 13 of edge-protected interpolation weighted volumes 25 obtain signal 18.The pixel that obtains through interpolation in this moment 18 has obtained the present invention relates to the correction of the edge-protected strategy that algorithm proposes.Signal 18 and motion-adaptive interpolation weight coefficient w _Motion(22) in multiplier 14, multiply each other its result and motion-adaptive interpolation weighted volumes (1-w _Motion) (n-1) (23) addition in adder 15 obtains signal 19 to I for x+1, y.So far, the pixel that obtains through interpolation in 19 obtains the present invention relates to the correction of the motion-adaptive interpolation strategy that algorithm proposes, and becomes the luminance signal of changing line by line.The carrier chrominance signal Cr of input _nAnd Cb _nAfter visual expanding element II10 expands in vertical direction, enter among the one dimension interpolation device II11, finish the interlacing of vertical direction and change interpolation line by line, obtain carrier chrominance signal 20.Luminance signal 19 of Huaing and carrier chrominance signal 20 are synthesized a frame progressive signal again in frame per second converter unit 7 line by line, and by the speed output of the simple principle that repeats with output frame rate.Therefore the frame per second of the video sequence 21 of 7 outputs is identical with the output frame rate of appointment, but resolution not necessarily reaches the output requirement.Video sequence 21 through visual expanding element III8 expansion, is upwards expanded row and a delegation respectively left again, expands two row and two row to the right downwards, carries out in the two-dimensional interpolation device 9 again.The 9th, two-dimentional leggy Cubic interpolation filter, the level of this interpolation device and the tap number of vertical direction are 4, and level is 128 with vertical number of phases.9 interpolation interval delta x and Δ y with the interpolation pixel send in the phase generator 26, calculated level and vertical interpolation phase place p _NhAnd p _Mv, and from interpolation coefficient memory cell 27, extract the filter coefficient u of horizontal direction and vertical direction _j(p _Mh), u _i(p _Nv), carry out two-dimensional interpolation and calculate, obtain output image 28 at last.

Fig. 2 is two-dimentional Cubic interpolation schematic diagram.Be in one 4 * 4 known image region, to insert a pixel shown in the figure, this pixel apart from point (x, distance y) is respectively Δ x and Δ y, corresponding vertical and horizontal phase is respectively $p_{\mathrm{nv}}\≅\mathrm{\Δx}\·N_p,$ $p_{\mathrm{mh}}\≅\mathrm{\Δy}\·N_p.$ According to the vertical and horizontal interpolation phase place of determining, the two-dimensional interpolation filter extracts the coefficient of each tap correspondence from the interpolation coefficient memory cell, with the summation of multiplying each other of each known pixel point in the interpolation window, obtains the gray scale of interpolation pixel, as shown in Equation (5).

Fig. 3 is the motion detection schematic diagram.To interleaved video list entries, in the n field 4 * 4 with the n-1 field in 3 * 4 spatially be corresponding.When carrying out motion detection, earlier the picture block in the n field is done average interpolation in the ranks, as shown in Equation (6), make it to become one 3 * 4 interpolation block; Ask the mean absolute difference MotionDiff of pixel again with picture block corresponding in this interpolation block and the n-1 field _Pixel, as shown in Equation (7); At last according to MotionDiff _PixelDetermine Motion Adaptive weight coefficient w _MotionAdopt curve shown in Figure 4 to determine different Mo tionDiff in the present embodiment _PixelUnder w _Motion

$\mathrm{MotionBlock}(x+2i+1,y+j,n)=\frac{1}{2}[I(x+2i,y+j,n)+I(x+2(i+1),y+j,n)],i=-\mathrm{1,0,1},j=-\mathrm{1,0,1,2}----\left(6\right)$

Wherein MotionBlock (x+2i+1, y+j, the n) pixel in the interpolation block that obtains for one 4 * 4 picture block from the n field, it with the n-1 field in one 3 * 4 corresponding one by one on the locus; I (x+2i, y+j, n) and I (x+2 (i+1), y+j n) are pixel in these 4 * 4.Pixel (x, position y) as shown in Figure 3, x=0,2,4,6,8... or 1,3,5,7,9....

${\mathrm{MotionDiff}}_{\mathrm{pixel}}=\frac{1}{12}\underset{i=-1}{\overset{1}{\mathrm{\Σ}}}\underset{j=-1}{\overset{2}{\mathrm{\Σ}}}|\mathrm{MotionBlock}(x+2i+1,y+j,n)-I(x+2i+1,y+j,n-1)|------\left(7\right)$

Fig. 5 is that the edge related side is to judging and the interpolation schematic diagram.As shown in the figure, in input imagery, get one 2 * 3 piece.Calculate the absolute difference EdgeDiff of two pixels on all directions according to formula (8) _j(j=-1,0,1), then { EdgDiff _jThe j of the minimum value correspondence related side that is the interpolation pixel to, and respective edges protection interpolation is the mean value of two known pixels that make progress the related side, promptly when the related side when being k, formula (9) establishment.

EdgeDiff _j＝|I(x，y+j，n)-I(x+2，y-j，n)|，j＝-1，0，1 (8)

$I_{\mathrm{Edge}}^{*}(x+1,y,n)=\frac{1}{2}[I(x,y+k,n)+I(x+2,y-k,n)],k=-\mathrm{1,0,1}------\left(9\right)$

The weight coefficient w of edge-protected interpolation _EdgeBe the EdgeDiff that makes progress according to the related side _kDetermine, determine different EdgeDiff according to curve shown in Figure 6 in the present embodiment _kUnder w _Edge

Adopt the parameter among the embodiment that one 10 frame interlacing list entries are expanded 2.6 times in the horizontal direction, 1.8 times of vertical direction expansions, and become progressive signal demonstration, the format conversion processing that frame per second doubles.Having calculated 10 frame average peak Y-PSNRs is 24.58db, and the image after the conversion outputs to through the HD-STORE graphics workstation and watched effect on the television set, shows that pictorial detail is stable, clear.

Claims (7)

1; a kind of based on Motion Adaptive and edge-protected 3 D video format conversion method; it is characterized in that; video format conversion is to finish on basis line by line; in with the interlaced process that transfers sequence line by line to, adopted edge-protected strategy and Motion Adaptive strategy that the result of field interpolation is revised; the process of video format conversion is: the vision signal to the interlacing input is carried out the interlaced processing of change line by line earlier; when changing line by line; adopt two-dimensional interpolation and the weighted sum of Cubic two-dimensional interpolation in edge-protected; do three-dimensional weighted sum with it with the motion-adaptive interpolation result again; sequence after changing is line by line carried out the frame per second conversion; do the conversion of frame intrinsic resolution at last, the video sequence of a frame per second or a rate is specified in output.

2, according to claim 1ly it is characterized in that based on Motion Adaptive and edge-protected 3 D video format conversion method described interlaced change line by line is specially:

Do three kinds of processing to interlaced: an one dimension Cubic interpolation of vertical direction at first being; the vertical resolution of input video sequence is doubled; secondly the interpolation pixel is done edge-protected interpolation; and with the weighted sum as a result of one dimension Cubic interpolation; obtain the result of field interpolation; at last the interpolation pixel in interlaced is carried out motion detection; the Motion Adaptive weight coefficient of determining according to motion detection obtains the pixel line by line through three-dimensional interpolation with the result and the weighted sum of motion reference pixel of field interpolation.

3, according to claim 2ly it is characterized in that based on Motion Adaptive and edge-protected 3 D video format conversion method, described one dimension Cubic interpolation, specific as follows:

When carrying out interlaced change line by line, only need the image of input is carried out the interpolation of vertical direction, the Cubic function interpolation filter that adopts 4 taps, 128 phase places carries out the one dimension interpolation of vertical direction to input signal, and the coefficient of interpolation filter calculates by following formula:

The computing formula that input imagery is carried out the Cubic function interpolation is in vertical direction:

$I_{\mathrm{cubic}}^{*}(x+1,y,n)=\underset{i=-1}{\overset{2}{\mathrm{\Σ}}}I(x+2i,y,n)u_v\left(\frac{2i-1}{2}\right)-------\left(2\right)$

I wherein _Cubic ^*(x+1, y, n) be in the n field in the position (x+1 y) locates, and is the two-dimensional interpolation result of interpolation kernel with the Cubic function, x={1 wherein, 3,5, Λ } or { 0,2,4, Λ },

I (x, y, n) be in the n field in the position (x, the known pixel of y) locating,

I ∈ [1,2] is the vertical offset of the pixel of participation interpolation,

Be the coefficient of one dimension 4 taps 128 phase place Cubic function interpolation filters, subscript v represents the interpolation filter of vertical direction.

4, according to claim 2ly it is characterized in that based on Motion Adaptive and edge-protected 3 D video format conversion method, described edge-protected interpolation, specific as follows:

Edge-protected interpolation is made up of two parts: the edge related side is to judgement; the edge directional interpolation of being correlated with; at first the edge related side who judges interpolation point according to the known pixel interpolation difference around the interpolation point to; be 0 ° to the right with level; the related side to set be: { 45 °; 90 °; 135 ° }; make progress the mean value of known pixel as edge-protected interpolation result with this related side then; and determine the weight coefficient of edge-protected interpolation to the difference of pixel with the related side; edge-protected interpolation and two-dimentional Cubic interpolation have constituted the result of an interior two-dimensional interpolation together, promptly

$I_{\mathrm{infra}}^{*}(x+1,y,n)=(1-w_{\mathrm{edge}})I_{\mathrm{cubic}}^{*}(x+1,y,n)+w_{\mathrm{edge}}{I}_{\mathrm{edge}}^{*}(x+1,y,n)------\left(3\right)$

I wherein _Infra ^*(x+1, y are to adopt the interpolation result that the two-dimensional interpolation algorithm obtains in the field in the n field of input n), and this two-dimensional interpolation comprises Cubic two-dimensional interpolation and edge-protected interpolation, x={1,3,5, Λ } or { 0,2,4, Λ },

I _Cubic ^*(x+1, y n) are the interpolation result that adopts two-dimentional Cubic interpolation to obtain in the n field of input,

I _Edge ^*(x+1, y n) are the interpolation result that adopts edge-protected method to obtain in the n field of input,

w _EdgeBe that (x+1, edge-protected weight coefficient y) is by the make progress difference decision of known pixel of the related side of this point for interpolation picture element in the n field of input.

5, according to claim 2ly it is characterized in that based on Motion Adaptive and edge-protected 3 D video format conversion method, described motion-adaptive interpolation, specific as follows:

Adopt the method for Motion Adaptive that the motion and standstill content is handled respectively, the basis of Motion Adaptive algorithm is motion detection and motion determination, employing is based on several determination methods of the algorithm of field, if when the difference of certain pixel of front court and the interpolation pixel of reference field corresponding position above certain thresholding, think that then the pixel when the front court moves, otherwise think that the pixel when the front court is static, when the pixel in the front court is carried out detection of motion, this algorithm gives the motion weight coefficient of this pixel, therefore weighted sum obtains the interpolation pixel in progressive frame according to the motion weight coefficient by the corresponding pixel with reference field of field interpolation, adopts the interpolation formula of Motion Adaptive technology to be:

$I_{\mathrm{inter}}^{*}(x+1,y,n)=m\_\mathrm{ratio}{I}_{\mathrm{intra}}^{*}(x+1,y,n)+(1-m\_\mathrm{ratio})\·I(x+1,y,n-1)------\left(4\right)$

Wherein, I _Inter ^*(x+1, y are that the x+1 that obtains through the field value of interleaving in the n field is capable n), y row pixel, x={1,3,5, Λ } or { 0,2,4, Λ };

I _Intra ^*(x+1, y are to adopt field interpolation in the n field n), comprise two-dimentional Cubic interpolation and edge-protected interpolation, and the x+1 that obtains is capable, y row pixel,

I (x+1, y are that x+1 in the n-1 field is capable n-1), y row pixel,

M_ratio is the motion weight coefficient, is judged as static pixel, then m_ratio=0 if work as the front court pixel through motion detection.m_ratio∈[0，1]。

6, according to claim 1ly it is characterized in that based on Motion Adaptive and edge-protected 3 D video format conversion method described frame per second conversion is specially:

The input of frame per second conversion is a progressive scanning sequence, the method that repeats is adopted in the frame per second conversion nearby, the content that is output frame equals the most close with it in time input signal line by line, and the required frame memory of frame per second conversion stores the progressive video signal of two frame single-definitions.

7, according to claim 1 based on Motion Adaptive and edge-protected 3 D video format conversion method, it is characterized in that described resolution conversion is specially:

The input of resolution conversion be change line by line and frame per second for specifying the vision signal of output frame rate, resolution conversion adopts two-dimentional Cubic leggy interpolation algorithm in the frame, finishes the resolution conversion of arbitrary proportion.

Images