CN101312490A - Pixel interpolation devices and methods therefor - Google Patents

Abstract

A pixel interpolation method comprises steps of: according to the pixel data of a first real pixel row and a second real pixel row, generating a plurality of first large blocks and second large blocks, comparing the first and second large blocks to find the minimum block difference value and the check direction corresponding to the target pixel, while the target pixel belongs to the first interpolation pixel row; according to the pixel data of the first and second real pixel rows and the check direction, interpolating the target pixel generating a first interpolation pixel row.

Description

Pixel interpolation devices and method thereof

Technical field

The relevant a kind of pixel interpolation devices of the present invention, and particularly relevant a kind of pixel interpolation devices that is applied to release of an interleave (De-Interlace) system.

Background technology

General film is taken and is play in staggered (Interlace) mode, is example with the TV, and the broadcast of an one picture (Frame) is to broadcast strange figure field (an Odd Field) earlier, broadcasts bigraph field (EvenField) again and finishes.Strange figure the odd number display line of picture and even number display line constitute for this reason respectively with the bigraph field.Because existing market is more and more high to the requirement of image quality, but not the image quality that interleave method shows is better than alternating expression image quality, therefore existing display packing produces gradual (Progressive) picture again and broadcasts for after the alternating expression picture is carried out release of an interleave (De-Interlace).

The tradition de-interlace method is to carry out interpolation (Interpolation) according to the pixel of actual pixels row in the input figure field, to find out the pixel data that is positioned at the object pixel between the actual pixels row, so to produce gradual picture.Yet traditional de-interlace method is not because of considering in the actual pixels row consistency in order to similarity degree between the pixel data that carries out interpolation and pixel data, so, make the gradual picture that obtains produce pixel data and the discontinuous pixel of surrounding pixel often, the gradual picture that causes obtaining has flicker, fuzzy and irregular colour and problem such as spares.

Summary of the invention

Relevant a kind of pixel interpolation devices of the present invention and method thereof, can overcome traditional de-interlace method effectively and not consider the problem that gradual picture that the consistency of similarity degree between the pixel data that carries out interpolation and its pixel data causes obtaining has flicker, fuzzy and irregular colour, and have the quality advantage of higher of the gradual image that obtains in fact.

Propose a kind of pixel interpolation devices according to the present invention, include first line buffer (LineBuffer), second line buffer, block comparison (Block Matching) unit and interpolation operation unit.First and second line buffer is used for storing the pixel data of first actual pixels row and second actual pixels row respectively.The block comparing unit is coupled to first and second line buffer, it lists to produce a plurality of first blocks (Macroblock) and list at second actual pixels in first actual pixels and produces corresponding a plurality of second largest blocks and carry out the block comparison, with the minimum block difference of finding out target and detection side to, object pixel belongs to first inter polated pixel row.The interpolation operation unit is coupled to block comparing unit and first, second line buffer, be used for according to the pixel data of first, second actual pixels row and detection side always interpolation produce the object pixel of first inter polated pixel row.

Propose a kind of pixel interpolation method according to the present invention, comprise following step.At first, produce a plurality of first blocks and second largest block respectively according to the pixel data of first actual pixels row and the pixel data of second actual pixels row.Then, to first and second big block carry out block comparison with the minimum block difference of finding out target and detection side to, object pixel belongs to first inter polated pixel and is listed as.Afterwards, according to the pixel data of first, second actual pixels row and detection side always interpolation produce the object pixel of first inter polated pixel row.

Description of drawings

For foregoing of the present invention can be become apparent, a preferred embodiment cited below particularly, and conjunction with figs. is elaborated, wherein:

Fig. 1 is the calcspar according to the pixel interpolation devices of first embodiment of the invention.

Fig. 2 is the actual pixels row of present embodiment and the schematic diagram of inter polated pixel row.

Fig. 3 is the detailed block diagram of the interpolation operation unit of Fig. 1.

Fig. 4 is the flow chart according to the interpolating method of first embodiment of the invention.

Fig. 5 is the calcspar according to the pixel interpolation devices of second embodiment of the invention.

Fig. 6 is the detailed block diagram of the interpolation operation unit of Fig. 5.

Fig. 7 is the flow chart according to the interpolating method of second embodiment of the invention.

Fig. 8 is the calcspar according to the pixel interpolation devices of third embodiment of the invention.

Fig. 9 is the detailed block diagram of the analytic unit of Fig. 8.

Figure 10 is the detailed block diagram of the comprehensive arithmetic element of Fig. 8.

Figure 11 is the flow chart according to the pixel interpolation method of third embodiment of the invention.

Figure 12 is the calcspar according to the interpolation operation unit of the pixel interpolation devices of fourth embodiment of the invention.

Figure 13 is the flow chart according to the pixel interpolation method of fourth embodiment of the invention.

Embodiment

The pixel interpolation devices of present embodiment carries out the block comparison via the actual pixels row to the figure field of staggered (Interlace) input, find out pairing minimum block difference of an object pixel in the inter polated pixel row and detection side to.The pixel interpolation devices of present embodiment also produces the pixel data of object pixel according to the detection side to the pixel data with actual pixels row, to reach the release of an interleave operation.

First embodiment

Please refer to Fig. 1 and Fig. 2, Fig. 1 is the calcspar according to the pixel interpolation devices of first embodiment of the invention, and Fig. 2 is the schematic diagram of actual pixels row of the present invention and inter polated pixel row.Pixel interpolation devices 10 is in order to receiving staggered input figure field data 100, and comes interpolation to go out the pixel data of each pixel among the inter polated pixel row L3 between actual pixels row L1 and L2 according to the pixel data of actual pixels row L1 and L2 wherein.

Wherein, actual pixels row L1, L2 and inter polated pixel row L3 comprise that respectively (j=1～n), Pb (j) and Pd (j), n is the natural number greater than 1 to pixel Pa (j).Next do explanation with pixel interpolation devices 10 according to the example that is operating as that the pixel data of the pixel Pa (j) of actual pixels row L1 and L2 and Pb (j) produces the pixel data So of object pixel Pd (x), wherein x be between 1 and n between natural number.

Pixel interpolation devices 10 comprises line buffer (Line Buffer) 12,14, block comparison (Block Matching) unit 16 and interpolation operation unit 18. Line buffer 12 and 14 is respectively in order to store the pixel data of actual pixels row L1 and L2.Block comparing unit 16 is coupled to line buffer 12 and 14, in order on actual pixels row L1 and L2, to produce first block (Macroblock) Ma and second largest block Mb respectively, and to its carry out block comparison with the minimum block difference Dmin (x) that finds out target Pd (x) and detection side to, the detection side is to representing to numerical value D (x) with the detection side.Interpolation operation unit 18 is coupled to line buffer 12,14 and block comparing unit 16, in order to the pixel data So that produces corresponding object pixel Pd (x) to numerical value D (x) according to these actual pixels row L1 and pixel data and the detection side of L2.

Suppose that block comparing unit 16 comprises a plurality of direction index m, and in pixel Pa (j), find out a first block Ma (m) according to one of them direction index, in pixel Pb (j), find out a corresponding second largest block Mb (m), the line of the center pixel of this first block and the center pixel of corresponding second largest block is the indicated direction of this direction index m, and object pixel Pd (x) is positioned on this line.In the present embodiment with direction index m for-2 ,-1,0,1,2 being that example illustrates, as shown in Figure 2, therefore can on these 5 direction index, find out paired first block and second largest block: Ma (2) and Mb (2), Ma (1) and Mb (1), Ma (0) and Mb (0), Ma (1) and Mb (1) and Ma (2) and Mb (2) respectively.

First block and second largest block all comprise k pixel, equaling 3 with k in this embodiment is the detailed step of piece comparison in area pellucida the example, as shown in Figure 2, first block Ma like this (2)～Ma (2) comprises pixel Pa (x-3)～Pa (x-1), Pa (x-2)～Pa (x), Pa (x-1)～Pa (x+1), Pa (x)～Pa (x+2), Pa (x+1)～Pa (x+3) respectively, and the pixel among second largest block Mb (2)～Mb (2) can be analogized according to above-mentioned rule and obtains.Described before center pixel be first or second largest block near the pixel of central point, for example first block Ma (2) comprises Pa (x-3), Pa (x-2) and Pa (x-1) in this example, then center pixel is Pa (x-2).

Block comparing unit 16 carries out the block comparison with first and second big block in correspondence with each other, to find out the block difference Diff (m) corresponding with all directions index.For example block comparing unit 16 subtracts each other the pixel value of corresponding pixel among first block Ma (2) and the second largest block Mb (2), and the summation after taking absolute value is as this block difference.The block difference Diff (2) that is direction index m=-2 can be expressed as:

Diff (2)=| Pa (x-3)-Pb (x+3) |+| Pa (x-2)-Pb (x+2) |+| Pa (x-1)-Pb (x+1) |, then block comparing unit 16 also calculates block difference Diff (the 1)～Diff (2) of all the other direction index, and the minimum value of finding out block difference Diff (2)～Diff (2) is as minimum block difference Dmin (x) and always export the detection side to numerical value D (x) with its pairing direction index as the detection side.For instance, if minimum value is Diff (1) among Diff (2)～Diff (2), then minimum block difference Dmin (x) equals Diff (1), and the detection side is 1 to numerical value D (x).

Please refer to Fig. 3, it is the detailed block diagram of the interpolation operation unit 18 of Fig. 1.Interpolation operation unit 18 for example comprises pixel selection unit 18a and comprehensive arithmetic element 18b.Pixel selection unit 18a is in order to receiving the pixel data of detection side's each pixel in numerical value D (x), actual pixels row L1 and L2, and finds out first corresponding reference pixel P1 and the pixel data of the second reference pixel P2 to numerical value D (x) according to the detection side in actual pixels row L1 and L2.For example, as above supposed, the detection side that block comparing unit 16 is exported is 1 to numerical value D (x), then pixel selection unit 18a promptly is chosen among actual pixels row L1 and the L2, is positioned at pixel data that direction index m equals pixel Pa (x+1) and Pb (x-1) on 1 and comes respectively pixel data as the first reference pixel P1 and the second reference pixel P2.

Comprehensive arithmetic element 18b receives the pixel data of first and second reference pixel P1 and P2, and it is carried out interpolation (Interpolation) calculate, and with the pixel data output of interpolative operation result as object pixel Pd (x).

The pixel interpolation devices 10 of present embodiment carries out block to each actual pixels earlier to be cut apart, carry out block comparison then finding out the detection side, and obtain pixel data So to the pixel data of the pairing first reference pixel P1 of numerical value D (x) and the second reference pixel P2 with the detection side to numerical value D (x).So the pixel interpolation devices 10 of present embodiment can be chosen and come computing to obtain pixel data So according to pixel data the similarity degree higher first reference pixel P1 and the second reference pixel P2, makes near the pixel of object pixel Pd (x) and its have close pixel data.So, can reduce the probability that gradual picture glimmers, blurs and irregular colour is even of generation effectively.

Please refer to Fig. 4, it is the flow chart according to the interpolating method of first embodiment of the invention.As step (a), block comparing unit 16 produces first block and second largest block (for example Ma (2)～Ma (2) and Mb (2)～Mb (2)) respectively according to the pixel among actual pixels row L1 and the L2 (for example Pa (x-3)～Pa (x+3) and pixel Pb (x-3)～Pb (x+3)).

Then, as step (b), 16 pairs of first blocks of block comparing unit (for example Ma (2)～Ma (2)) carry out the block comparison with corresponding second largest block (for example Mb (2)～Mb (2)), with the minimum block difference Dmin (x) that finds out object pixel Pd (x) and detection side to numerical value D (x).Afterwards, as step (c), interpolation operation unit 18 is found out the pixel data of the first reference pixel P1 and the second reference pixel P2 according to the detection side to numerical value D (x) in actual pixels row L1 and L2, and carries out the pixel data So that computing obtains object pixel Pd (x) according to it.

Though the example that is operating as that only is respectively the pixel interpolation devices 10 of 5 present embodiments during with 3 with m and k in the present embodiment is done explanation, yet, in the present embodiment number of m and k be not limited to 5 with 3, and also can be other numerical value.

Second embodiment

Please refer to Fig. 5 and Fig. 6, Fig. 5 is the calcspar according to the pixel interpolation devices of second embodiment of the invention, and Fig. 6 is the detailed block diagram of the interpolation operation unit of Fig. 5.Pixel interpolation devices 10 differences of the pixel interpolation devices 20 of present embodiment and first embodiment are that the block comparing unit 26 minimum block difference Dmin of output (x) are to interpolation operation unit 28, interpolation operation unit 28 produces the first weight coefficient pw1 according to minimum block difference Dmin (x), and the data that produced according to the first weight coefficient pw1, first and second operational data (not shown) are as the pixel data So ' of object pixel Pd (x).

First operational data for example is the data of interpolation operation unit 28 according to the pixel data computing gained of the first reference pixel P1 and the second reference pixel P2, and the second operand certificate is the data of interpolation operation unit 28 according to the pixel data computing gained of the 3rd reference pixel P3 and the 4th reference pixel P4 for example.P3 and P4 are respectively and are positioned at that actual pixels row L1 goes up with L2 and are corresponding pixel mutually.In the present embodiment P3 and P4 for example be respectively with the detection side to numerical value 0 corresponding pixel Pa (x) and pixel Pb (x).

Interpolation operation unit 18 differences of the interpolation operation unit 28 of present embodiment and first embodiment are that it also comprises coefficient generating unit 28c, in order to receive and to come to compare, to produce the first weight coefficient pw1 with one first critical numerical value th1 according to minimum block difference Dmin (x).For example, the first weight coefficient pw1 can set according to following rule:

$\mathrm{<figure-callout\; id="1"\; label="pw"\; filenames="A20071010407900171.png"\; state="\{\{state\}\}">pw</figure-callout>}1=\left\{\begin{array}{cc}1& D\min <\left(\mathrm{th}1\right)/4\\ 0.75& \left(\mathrm{th}1\right)/4<D\min <\left(\mathrm{th}1\right)/2\\ 0.5& \left(\mathrm{th}1\right)/2<D\min <\mathrm{<figure-callout\; id="1"\; label="th"\; filenames="A20071010407900171.png"\; state="\{\{state\}\}">th</figure-callout>}1\\ 0.25& \mathrm{<figure-callout\; id="1"\; label="th"\; filenames="A20071010407900171.png"\; state="\{\{state\}\}">th</figure-callout>}1<D\min \end{array}\right.$

Pixel selection unit 28a is except finding out the pixel data of the first reference pixel P1 and the second reference pixel P2 in actual pixels row L1 and L2 respectively according to the detection side to numerical value D (x), also in order to respectively at the pixel data of choosing the 3rd and the 4th reference pixel P3 and P4 among actual pixels row L1 and the L2, and output.Comprehensive arithmetic element 28b receives first to fourth reference pixel P1～P4, and produce the first operational data s1, and produce the first operational data s2 according to the pixel data of the 3rd and the 4th reference pixel P3 and P4 according to the pixel data of first and second reference pixel P1 and P2.In this embodiment, first and second operational data can be expressed from the next:

s1＝P1+P2

s2＝P3+P4

Comprehensive arithmetic element 28b is also in order to deciding the weight of first operational data and second operand certificate according to the first weight coefficient pw1, and with the numerical value that obtains as pixel data So '.For example comprehensive arithmetic element 28b can calculate So ' according to following equation:

${\mathrm{So}}^{\&prime;}=\frac{1}{2}[\mathrm{<figure-callout\; id="1"\; label="s"\; filenames="A20071010407900171.png"\; state="\{\{state\}\}">s</figure-callout>}1\&times;\mathrm{<figure-callout\; id="1"\; label="pw"\; filenames="A20071010407900171.png"\; state="\{\{state\}\}">pw</figure-callout>}1+\mathrm{<figure-callout\; id="2"\; label="s"\; filenames="A20071010407900171.png"\; state="\{\{state\}\}">s</figure-callout>}2\&times;(1-\mathrm{pw}1)]$

The first weight coefficient pw1 can be considered the reliability coefficient of first and second reference pixel of finding out via the block comparison in the present embodiment, and its degree of reliability system promotes along with dwindling of its minimum block difference Dmin (x).The pixel interpolation devices 20 of present embodiment can distribute the weight of first operational data and second operand certificate by the first weight coefficient pw1, makes the pixel data So ' more can be effectively near desired quantity.

Please refer to Fig. 7, it is the flow chart according to the interpolating method of second embodiment of the invention.The interpolating method difference of the interpolating method of present embodiment and first embodiment is that it also comprises step (d), coefficient generating unit 28c is in order to more minimum block difference Dmin (x) and the first critical numerical value th1, and produces the first weight coefficient pw1 according to comparative result.And also comprise in the step of present embodiment (c): comprehensive arithmetic element 28b produces the second operand certificate according to the pixel data of the 3rd reference pixel P3 and the 4th reference pixel P4, and decide the weight of first and second operational data, and then calculate the pixel data So ' of object pixel Pd (x) according to the first weight coefficient pw1.

The 3rd embodiment

Please refer to Fig. 8 and Fig. 9, Fig. 8 is the calcspar according to the pixel interpolation devices of third embodiment of the invention, and Fig. 9 is the detailed block diagram of the analytic unit of Fig. 8.The present embodiment and the second embodiment difference are that the pixel interpolation devices 30 of present embodiment also comprises analytic unit 37 and memory cell 39, analytic unit 37 in order to according to the detection side to numerical value D (x) produce interpolation direction numerical value D ' (x), indicating label wba (x) and wbb (x), and memory cell 39 is in order to store indicating label wba (x) and wbb (x).

See also Fig. 9.Analytic unit 37 comprises label setup unit 37a, and it is that the detection side who judges object pixel Pd (x) always finds out the affiliated direction classification of object pixel Pd (x).In this embodiment, direction index m divides into three major types, and the direction index of negative (1 ,-2 ...) be classified as category-A, the direction index of positive number (1,2 ...) be classified as category-B, the direction index is 0 then to be classified as the C class.For instance, the detection side that label setup unit 37a detects object pixel Pd (x) equals at 1 o'clock to numerical value D (x), and the direction category setting that is about to object pixel Pd (x) is B.In addition, can be paid one group of indicating label (wba (x), wbb (x)) according to different direction classification object pixel Pd (x).In the present embodiment, the indicating label of direction category-A be set to (wba, wbb)=(1,0), the indicating label of direction category-B be set to (wba, wbb)=(0,1), and the indicating label of direction C class be set to (wba, wbb)=(0,0).Therefore, when the direction classification of object pixel Pd (x) is category-B, the indicating label of object pixel Pd (x) (wba (x), wbb (x))=(0,1).

In addition, the direction classification that label setup unit 37a also detects near a plurality of pixels object pixel Pd (x) with indicating label wba (x) and wbb (x) be modified to wba ' (x) and wbb ' (x), its judgement formula is as follows:

if{(wba(x-s)＝1)&(wba(x-s+1)＝1)&...

&(wba(x)＝1)&...&(wba(x+s-1)＝1)&(wba(x+s)＝1)}

wba′(x)＝1

else

wba′(x)＝0

and

if{(wbb(x-s)＝1)&(wbb(x-s+1)＝1)&...&(wbb(x)＝1)&...

&(wbb(x+s-1)＝1)&(wbb(x+s)＝1)}

wbb′(x)＝1

else

wbb′(x)＝0

Wherein s represents the scope that detects.Can determine more that by above-mentioned judgment mechanism the detection side of object pixel Pd (x) is reliable to whether.For example the detection side of all pixels is to belonging to similar in the scope that detects, and the detection side who then represents object pixel Pd (x) is to for reliable; Yet in the scope that detects if the detection side that any pixel arranged to other inhomogeneities, then the detection side of object pixel Pd (x) is unreliable to being regarded as, and its indicating label be corrected for (wba ' (x), wbb ' is (x))=(0,0), that is its revised detection side be 0 to numerical value D (x).Revised indicating label (wba ' (x), wbb ' is (x)) store by memory cell 39.

Analytic unit 37 also comprises adjustment in direction unit 37b in addition, and it produces interpolation direction numerical value D ' (x) in order to the detection side according near a plurality of pixels object pixel Pd (x) to numerical value.In the present embodiment, adjustment in direction unit 37b be find out D (x-s), D (x-s+1) ..., D (x) ..., D (x+s-1), D (x+s) middle number be used as interpolation direction numerical value D ' (x), the scope that detects of s representative wherein.For instance, when s=2, if { D (x-2), D (x-1), D (x), D (x+1), D (x+2) }={ 4 ,-2 ,-6,-1 ,-2} will become after its ascending arrangement: [6 ,-4,-2 ,-2 ,-1], then the interpolation direction numerical value D ' of object pixel Pd (x) (x) is-2.

Please refer to Figure 10, it is the detailed block diagram of the interpolation operation unit 38 of Fig. 8.Interpolation operation unit 28 differences of the interpolation operation unit 38 of present embodiment and second embodiment are that it also comprises coefficient generating unit 38d, in order to receive and to produce the second weight coefficient pw2 according to indicating label.

(x) from selecting first and second reference pixel P1 and the pixel data of P2 and the pixel data of the 3rd and the 4th reference pixel P3 and P4 in the pixel data of actual pixels row L 1 and L2, the 3rd and the 4th reference pixel P3 and P4 are Pa (x) and Pb (x) to pixel selection unit 38a in this embodiment according to interpolation direction numerical value D '.

Coefficient generating unit 38d from memory cell 39, read each pixel in the inter polated pixel row indicating label (wba ' (x), wbb ' (x)), and according to the indicating label of object pixel Pd (x) surrounding pixel (wba ', wbb '), finished in the pixel column (being the L4 among Fig. 2) of interpolation with previous column, the indicating label of the pixel of corresponding this object pixel Pd (x) same vertical position (be Pd ' (x)) surrounding pixel (Pwba ', Pwbb ') calculates and produces the second weight coefficient pw2.Coefficient generating unit 38d is earlier according to equation:

$\left\{\begin{array}{c}\mathrm{sum}\_a\left(x\right)=\underset{i=x-s}{\overset{x+s}{\mathrm{\&Sigma;}}}[{\mathrm{wba}}^{\&prime;}\left(i\right)+\mathrm{Pwb}{a}^{\&prime;}\left(i\right)]\\ \mathrm{sum}\_b\left(x\right)=\underset{i=x-s}{\overset{x+s}{\mathrm{\&Sigma;}}}[w{\mathrm{bb}}^{\&prime;}\left(i\right)+\mathrm{Pw}{\mathrm{bb}}^{\&prime;}\left(i\right)]\end{array}\right.$

Coefficient sum_a (x) and sum_b (x), the wherein scope of s representative detection obtain adding up.And according to down with the judgement formula decide the second weight coefficient pw2:

if?wba′(x)＝1

$\mathrm{<figure-callout\; id="2"\; label="pw"\; filenames="A20071010407900171.png"\; state="\{\{state\}\}">pw</figure-callout>}2=\frac{|\mathrm{sum}\_a\left(x\right)-\mathrm{sum}\_b\left(x\right)|}{\mathrm{sum}\_a\left(x\right)}$

elseif?wbb′(x)＝1

$\mathrm{<figure-callout\; id="2"\; label="pw"\; filenames="A20071010407900171.png"\; state="\{\{state\}\}">pw</figure-callout>}2=\frac{|\mathrm{sum}\_a\left(x\right)-\mathrm{sum}\_b\left(x\right)|}{\mathrm{sum}\_b\left(x\right)}$

Comprehensive arithmetic element 38b distributes the weight of first operational data (s1) and second operand certificate (s2) according to the weight coefficient pw that is produced based on weight coefficient pw1 and pw2, and with the data that the obtain pixel data So as object pixel Pd (x) " export.Weight coefficient pw is the product of first, second weight coefficient pw1 and pw2 in the present embodiment.The pixel data So of object pixel Pd (x) " can obtain according to following formula:

${\mathrm{So}}^{\&prime;\&prime;}=\frac{1}{2}[\mathrm{<figure-callout\; id="1"\; label="s"\; filenames="A20071010407900171.png"\; state="\{\{state\}\}">s</figure-callout>}1\&times;\mathrm{pw}+s2\&times;(1-\mathrm{pw})]$

Please refer to Figure 11, it is the flow chart according to the pixel interpolation method of third embodiment of the invention.The pixel interpolation method difference of the pixel interpolation method of present embodiment and second embodiment is step (b) afterwards, the pixel interpolation method of present embodiment also comprise step (e), (f) with (g).As step (e), label setup unit 37a judges the direction classification of detection side under numerical value D (x), produce and store indicating label wba ' (x) and wbb ' (x), and memory cell 39 is in order to storing in the staggered input figure field data 100 indicating label of each pixel in each inter polated pixel row, and in order to output indicating label wba ', wbb ', Pwba ' and Pwbb ' to interpolation operation unit 38.

Then, as step (f), adjustment in direction unit 37b produces and exports interpolation direction numerical value D ' (x) to interpolation operation unit 38 according to the detection side of pixel Pd (x-s)～Pd (x+s) to numerical value.Then, as step (g), coefficient generating unit 38d according to the pixel Pd ' of pixel Pd (the x-s)～Pd (x+s) of inter polated pixel row L3 and last inter polated pixel row L4 (x-s)～Pd ' indicating label (x+s) produces the second weight coefficient pw2.S represents detection range.

And in step (c), comprehensive arithmetic element 38b distributes the weight of first operational data (s1) and second operand certificate (s2) according to the weight coefficient pw that is produced based on weight coefficient pw1 and pw2, and with the data of the generation pixel data So as object pixel Pd (x) ".

Also for example comprise step (e ') in step (e) and (f), label setup unit 37a for example analyzes the indicating label of pixel Pd (x-s)～Pd (x+s) and judges whether it belongs to identical direction classification.When pixel Pd (x-s)～Pd (x+s) did not belong to identical direction classification entirely, the detection side of modifying target pixel Pd (x) was to being respectively vertical direction and detection side to numerical value 0 with the detection side to numerical value D (x).

Though only comprise that with analytic unit 37 adjustment in direction unit 37b produces interpolation direction numerical value D ' (x) with the detection side according to pixel Pd (x-s)～Pd (x+s) to numerical value in the present embodiment, and pixel selection unit 38a (x) selects the example that is operating as of the pixel data of first and second reference pixel P1 and P2 to do explanation by interpolation direction numerical value D ', yet, the analytic unit 37 of present embodiment also can omit the setting of adjustment in direction unit 37b, and pixel selection unit 38a selects first and second reference pixel P1 and P2 according to the detection side to numerical value D (x).So, the pixel interpolation method of present embodiment also can be saved step (f), and can also the detection side replace interpolation direction numerical value D ' (x) to numerical value D (x) in the step (c).

Though only have coefficient generating unit 38c simultaneously in the present embodiment and 38d produces first and second weight coefficient pw1 and pw2 respectively with interpolation operation unit 38, and comprehensive arithmetic element 38b produces pixel data So according to weight coefficient pw " situation be that example is done explanation; yet; the interpolation operation unit 38 of present embodiment also can omit coefficient generating unit 38c and 38d one of them, and only be used as the weight that weight coefficient pw distributes first and second operational data with weight coefficient pw1 or pw2.So, the pixel interpolation method of present embodiment also can save step (d) with (g) one of them, and step (c) also can be saved the correlation step with reference to weight coefficient pw1 or pw2.

The 4th embodiment

Please refer to Figure 12, it is the calcspar according to the interpolation operation unit of the pixel interpolation devices of fourth embodiment of the invention.The pixel interpolation devices 40 of present embodiment and the pixel interpolation devices 30 of the 3rd embodiment on the circuit structure except interpolation operation unit 48 different, all the other are all identical, present embodiment is that interpolation operation unit 38 is replaced it with interpolation operation unit 48.Compare with interpolation operation unit 38, also comprise coefficient generating unit 48e in the interpolation operation unit 48, in order to produce weight coefficient pw3.

Coefficient generating unit 48e receives the pixel data of actual pixels row L1 and L2, and for example according to equation:

$\mathrm{sum}\_\mathrm{diff}=\underset{i=x-s}{\overset{x+s}{\mathrm{\&Sigma;}}}\left(|\mathrm{Pa}\left(i\right)-\mathrm{Pa}(i-1)\right|+|\mathrm{Pb}\left(i\right)-\mathrm{Pb}(i-1)|)$

Obtain accumulation factor sum_diff, wherein s is a detection range.| Pa (i)-Pa (i-1) | with | Pb (i)-Pb (i-1) | represent the absolute value of the pixel data difference of wantonly two adjacent actual pixels row L1 in detecting form respectively, with the absolute value of the pixel data difference of wantonly two adjacent actual pixels row L2 in detecting form.

Coefficient generating unit 48e comes accumulation factor sum_diff to compare with one second critical numerical value th2, to produce the 3rd weight coefficient pw3.For example, the 3rd weight coefficient pw3 can set according to following rule:

$\mathrm{<figure-callout\; id="3"\; label="pw"\; filenames="A20071010407900171.png"\; state="\{\{state\}\}">pw</figure-callout>}3=\left\{\begin{array}{cc}0.25& \mathrm{sum}\_\mathrm{diff}<\left(\mathrm{th}2\right)/4\\ 0.5& \left(\mathrm{th}2\right)/4<\mathrm{sum}\_\mathrm{diff}<\left(\mathrm{th}2\right)/2\\ 0.75& \left(\mathrm{th}2\right)/2<\mathrm{sum}\_\mathrm{diff}<\left(\mathrm{th}2\right)\\ 1& \left(\mathrm{th}2\right)<\mathrm{sum}\_\mathrm{diff}\end{array}\right.$

Comprehensive arithmetic element 48b distributes the weight of first and second operational data according to the weight coefficient Pw ' that is produced based on weight coefficient pw1～pw3, and with the data of the generation pixel data So as object pixel Pd (x) " '.Weight coefficient Pw ' equals the product of pw1～pw3 in the present embodiment.Comprehensive arithmetic element 48b is for example according to equation:

So″′＝s1×Pw′+s2×(1-Pw′)

Produce pixel data So " ', wherein s1 and s2 are respectively first and second operational data.

Because the pixel data drop between each pixel comes obviously manyly in the flat region in the non-flat forms zone of picture, therefore, list consideration in when near the smooth degree of the pixel object pixel Pd (x), then can lower the probability that the object pixel Pd (x) that interpolation is gone out becomes the assorted point of picture.Present embodiment promptly utilizes accumulation factor sum_diff to be used as the smooth coefficient of detection zone, and it detects the smooth situation of object pixel Pd (x) picture on every side, and reduces the weight of the 3rd weight coefficient pw3 when picture is comparatively smooth.So, reduce pixel data So via improving the 3rd and the 4th reference pixel P3 adjacent and the weight of P4 with object pixel Pd (x) " ' and the drop of the pixel data of surrounding pixel.So, the pixel interpolation devices 40 of present embodiment can make pixel data So by the 3rd weight coefficient pw3 effectively " ' near its ideal value, and reduce the probability that object pixel Pd (x) forms the assorted point of picture.

Please refer to Figure 13, it is the flow chart according to the pixel interpolation method of fourth embodiment of the invention.The pixel interpolation method difference of the pixel interpolation method of present embodiment and the 3rd embodiment is that it also comprises step (h), and the summation after the difference of coefficient generating unit 48e wantonly two adjacent pixels in detection range according to actual pixels row L1 and L2 takes absolute value produces weight coefficient pw3.And in the step (c), comprehensive arithmetic element 48b is that the product according to weight coefficient pw1, pw2 and pw3 obtains the weight that weight coefficient Pw ' decides first and second operational data, and with the data of the generation pixel data So as object pixel Pd (x) " '.

Though only have coefficient generating unit 48c simultaneously in the present embodiment with interpolation operation unit 48,48d and 48e produce weight coefficient pw1～pw3 respectively, and comprehensive arithmetic element 48b is simultaneously according to all weight coefficients, be pw1～pw3, produce pixel data So " ' situation be that example is done explanation; yet; the interpolation operation unit 48 of present embodiment also can omit coefficient generating unit 48c and 48d one of them; and only one of them comes to multiply each other with weight coefficient pw3 with weight coefficient pw1 and pw2; and distribute the weight of first and second operational data as weight coefficient Pw ' with its product; or interpolation operation unit 48 also can omit coefficient generating unit 48c and 48d simultaneously, and only is used as the weight that weight coefficient Pw ' distributes first and second operational data with weight coefficient pw3.So, the pixel interpolation method of present embodiment also can save step (g) with (d) one of them or all, and step (c) also can be saved the correlation step with reference to weight coefficient pw1 and pw2.

The pixel interpolation devices of present embodiment and method thereof also can be judged the smooth situation of picture according to the difference of the pixel data of wantonly two adjacent pixels among actual pixels row L1 and the L2, and produce the weight that the 3rd weight coefficient is adjusted first and second operational data according to this, and be used as the pixel data of object pixel with the data of its addition.So, the pixel data that the pixel interpolation devices of present embodiment also can reduce object pixel forms the probability of the assorted point in the picture, makes gradual picture quality higher in fact.

In sum, though the present invention with preferred embodiment announcement as above, yet it is not in order to limit the present invention.The persond having ordinary knowledge in the technical field of the present invention, without departing from the spirit and scope of the present invention, when doing various changes that are equal to and retouching.Therefore, protection scope of the present invention is when looking accompanying being as the criterion that the application's claim scope defined.

Claims (12)

1. pixel interpolation method comprises following step:

(a) produce a plurality of first blocks (Macroblock) and a plurality of second largest blocks respectively according to the pixel data of one first actual pixels row and the pixel data of one second actual pixels row;

(b) these first and second big block is carried out block comparison (Block Matching), with a minimum block difference of finding out a corresponding object pixel and a detection side to, this object pixel belongs to one first inter polated pixel and is listed as; And

(c) according to the pixel data of the pixel data of this first actual pixels row, these second actual pixels row and this detection side always interpolation produce this object pixels of this first inter polated pixel row.

2. pixel interpolation method according to claim 1 is characterized in that also comprising:

(d) produce one first weight coefficient according to this minimum block difference;

Wherein, step (c) also comes interpolation to produce this object pixel of these first inter polated pixel row according to this first weight coefficient.

3. pixel interpolation method according to claim 1 is characterized in that also comprising step:

(e) judge this detection side to producing a direction classification, and analyze this object pixel with and the direction classification of neighborhood pixels produce and store an indicating label; And

(f) analyze this object pixel with and the detection side of neighborhood pixels always produce an interpolation direction.

4. pixel interpolation method according to claim 3 is characterized in that also comprising:

(d) produce one first weight coefficient according to this minimum block difference; And

Wherein, step (c) also comes interpolation to produce this object pixel of these first inter polated pixel row according to this first weight coefficient and this interpolation direction.

5. pixel interpolation method according to claim 3 is characterized in that also comprising:

(g) produce one second weight coefficient according to the indicating label of this object pixel and neighborhood pixels thereof and the indicating label of pixel of having finished the one second inter polated pixel row of interpolation;

Wherein, step (c) also comes interpolation to produce this object pixel of these first inter polated pixel row according to this second weight coefficient and this interpolation direction.

6. pixel interpolation method according to claim 5 is characterized in that also comprising:

(d) produce one first weight coefficient according to this minimum block difference;

Wherein, step (c) also comes interpolation to produce this object pixel of these first inter polated pixel row according to this first weight coefficient.

7. pixel interpolation method according to claim 3 is characterized in that also comprising:

(h) produce one the 3rd weight coefficient according to the pixel data of these first actual pixels row and the pixel data of this second actual pixels row;

Wherein, step (c) also comes interpolation to produce this object pixel of these first inter polated pixel row according to this 3rd weight coefficient and this interpolation direction.

8. pixel interpolation method according to claim 7 is characterized in that also comprising:

(d) produce one first weight coefficient according to this minimum block difference;

Wherein, step (c) also comes interpolation to produce this object pixel of these first inter polated pixel row according to this first weight coefficient.

9. pixel interpolation method according to claim 7 is characterized in that also comprising:

(g) produce one second weight coefficient according to the indicating label of this object pixel and neighborhood pixels thereof and the indicating label of pixel of having finished the one second inter polated pixel row of interpolation;

Wherein, step (c) also comes interpolation to produce this object pixel of these first inter polated pixel row according to this second weight coefficient.

10. pixel interpolation method according to claim 9 is characterized in that also comprising:

(d) produce one first weight coefficient according to this minimum block difference;

Wherein, step (c) also comes interpolation to produce this object pixel of these first inter polated pixel row according to this first weight coefficient.

11. pixel interpolation method according to claim 3 is characterized in that also comprising:

(e ') before producing this interpolation direction, also analyze this object pixel with and the indicating label of neighborhood pixels determine whether that detection side with this object pixel is to being set at a specific direction.

12. pixel interpolation method according to claim 1 is characterized in that this first block and this second largest block include the pixel data of same number.

Images