CN100562129C - Time-domain filtering method for video code - Google Patents

Time-domain filtering method for video code Download PDF

Info

Publication number
CN100562129C
CN100562129C CN 200410088837 CN200410088837A CN100562129C CN 100562129 C CN100562129 C CN 100562129C CN 200410088837 CN200410088837 CN 200410088837 CN 200410088837 A CN200410088837 A CN 200410088837A CN 100562129 C CN100562129 C CN 100562129C
Authority
CN
Grant status
Grant
Patent type
Prior art keywords
macroblock
current
step
motion
luminance component
Prior art date
Application number
CN 200410088837
Other languages
Chinese (zh)
Other versions
CN1770880A (en )
Inventor
宁 王
黄晓红
Original Assignee
中兴通讯股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Grant date

Links

Abstract

本发明公开了一种视频编码的时域滤波方法,在同一宏块内部,对亮度分量和色度分量数据分别进行操作:对于亮度分量,首先计算当前帧当前宏块与前一帧对应宏块中亮度分量的绝对差值总和以初步将该宏块赋予一定的属性标记,根据所判定的宏块属性标记分别采取相应策略。 The present invention discloses a video encoding method for filtering a time domain, the same macroblock in the interior, the luminance and chrominance components, respectively, manipulate the data: the luminance component, first calculating the current macroblock in the current frame the previous frame with the corresponding macroblock in the sum of the absolute difference of the luminance component in the macroblock to impart the initial properties of certain markers, take the appropriate policy based on the determined macroblock attribute tags, respectively. 对于判定为可能静止并且其邻近宏块满足特定规则的宏块,根据其中各象素点与前一帧对应象素点亮度分量的绝对差值的统计信息,判断其运动属性,根据判定的宏块属性,分别采取相应的滤波方法,对于判定为静止宏块的宏块,根据其属性采取相应策略。 For determination may satisfy certain rules is stationary and its neighboring macroblocks, wherein each of the pixel dots in accordance with a prior statistical information corresponding to the absolute difference of the luminance component of the pixel dots, which determines the motion properties, determined according to the macro block properties were taken corresponding filtering method, it is determined that the stationary macroblock for a macroblock, according to the policy take its properties. 本发明克服了现有技术中复杂度较高、适应性较弱的缺点,在满足实时性要求的同时具有较强的鲁棒性。 The present invention overcomes the prior art high complexity, adaptive weaker disadvantages, to meet real-time requirements have robust.

Description

视频编码的时域滤波方法 The video encoding temporal filtering method

技术领域 FIELD

本发明属于计算机图像信息处理领域,具体涉及一种在视频编码中减少噪声的时域滤波方法。 The present invention belongs to the field of computer image processing information, particularly to the temporal filtering method of reducing noise in a video coding.

背景技术 Background technique

计算机图像信息处理技术随着社会各领域的需求而快速发展。 Computer image processing technology with the information needs of all areas of society and rapid development. 视频编码噪声的处理是其中的重要问题。 Processing video coding noise is one of the important issues. 将摄像头拍摄到的静止场景输送到电视屏幕时,往往会发现图形序列出现麻点噪声,对于紋理丰富的场景,这一现象尤为明显。 When the camera captured a still scene to transport the TV screen, often find graphic sequence pitting noise, for the rich textures scene, this phenomenon is particularly evident. 由于麻点噪声的影响,使得静止场景相同位置下的象素值 Pitting due to noise, so that the pixel values ​​at the same position of the stationary scene

在不同时刻出现非相关性。 Non-correlation appear at different times. 当采用诸如国际标准化組织(ISO)的MPEG 和国际无线通信联盟(ITU)推荐的11.263图像编码标准直接对这些序列进行编码时,由于相邻时刻图像在预测时的相关性比较弱,必将导致编码压缩比降低,进而影响静止序列下的图像效果。 When using MPEG and the International League of wireless communication, such as the International Standards Organization (ISO) is (ITU) recommendation 11.263 standard image encoding these sequences encoding directly, since the correlation of adjacent images in time when the prediction is relatively weak, will lead to coding compression ratio is lowered, thereby affecting the image quality in a stationary sequence.

为了抑制噪声,可以通过空域的、时域的或时空域的滤波器来实现, 其中空域滤波器只需要在某时刻一帧图像内完成,而时域滤波器则需要从两个或更多的连续的图像上获取数据,另外,时域滤波能够更好地消除序列间的不相关性。 To suppress the noise, can be achieved by time-domain or spatial filter airspace, wherein the spatial filter only needs to be completed within a certain time one frame of image, and the time-domain filter is required from two or more obtain data from the continuous image, additionally, the temporal filtering can better eliminate the correlation between the sequences. 对于时域滤波,近年来出现一些相关研究和方法,例如: 索尼公司申请的专利"视频数据处理设备和视频数据处理方法"(公开号为CN 1336765A),该专利方法中的时域滤波部分将输入一见频数据作为图像编码难易程度的函数,在编码时由自适应滤波器自适应滤波,方法包括场景变化检测、编码难易计算、运动补偿字段递归降噪、多路复用等一系列处理过程,其计算复杂度较高,对实时视频的编码不太适用;美国公开号 For the time-domain filtering, and appeared in recent years, a number of research methods, for example: Sony patent application "video data processing apparatus and video data processing method" (Publication No. CN 1336765A), the time-domain filter section of this patent will see input audio data as a function of a degree of difficulty of image encoding, adaptive filtering by the adaptive filter, comprising detecting a scene change, the encoding difficulty calculating the motion compensated field recursive noise reduction, and the like multiplexed in encoding a series of processing, high computational complexity, real-time video coding is not applicable; US Publication No.

6为6,281,942的"Spatial and temporal filtering mechanism for digital motion video signals",该专利方法首先以一定大小的块为单位判断其是否满足特定准则,然后在符合条件的块内根据其中各象素与其前帧对应象素之间的差值信息决定该象素的滤波强度,此方法对噪声小的静止序列能获得较好的处理效果,但难以区分噪声较小的小运动区域和噪声较大的静止区域, 适应性较弱。 6 of 6,281,942 "Spatial and temporal filtering mechanism for digital motion video signals", the patented process is first units of blocks of a certain size is determined whether certain criteria are met, then in accordance with its previous frame wherein each pixel in the block qualifying information determining filter strength difference between corresponding pixels of the pixels, this method still less noise sequence to obtain better treatment results, but it is difficult to distinguish the small movement area less noise and stationary noise is large area adaptability weak.

发明内容 SUMMARY

本发明所要解决的技术问题是提供一种在编码系统中减小编码序列噪声的时域滤波方法,使编码序列中的静止宏块在序列中不同图像之间具有更好的相关性,并且克服现有技术中复杂度较高、适应性较弱的缺点, 在满足实时性要求的同时具有较强的鲁棒性。 The present invention solves the technical problem is to provide a method for the temporal filtering to reduce the noise of a coding sequence in a coding system, that the coding sequence of the stationary macroblock has a better correlation between the different images in the sequence, and against prior art high complexity, adaptive weaker disadvantages, to meet real-time requirements have robust. 本发明所述的视频编码的时域滤波方法中,作为视频编码系统的一部分,当前帧当前宏块的滤波操作在其上一宏块的运动搜索之后、当前宏块的运动搜索之前进行,在同一宏块内部,对亮度分量和色度分量数据分别进行操作: Temporally filtered video coding method according to the present invention, as part of a video encoding system, the filtering operation of the current macroblock in the current frame after which a macroblock motion search, motion search performed prior to the current macroblock in inside the same macroblock, the luminance and chrominance components, respectively, operate on data:

(一)亮度分量的处理,包括以下步骤: (A) processing the luminance component, comprising the steps of:

第一步:计算当前帧(指未经时域滤波的当前图像,以下同)当前宏块与前一帧(指经过时域滤波的前帧图像,以下同)对应宏块中亮度分量的绝对差值总和(即SAD值)以初步将宏块赋予一定的属性标记(此时的属性标记自定义为运动宏块或可能静止宏块两种情况); The first step: calculation of the current frame (refer to the current image without temporal filtering, hereinafter the same) of the current macroblock and the previous frame (of the previous frame image is subjected to temporal filtering, hereinafter the same) of the luminance component corresponding to the absolute macroblock the sum of the difference (i.e., the SAD value) to the initial macroblock impart certain attribute tags (when labeled custom attributes as the stationary motion of macroblocks or macroblock may be two cases);

1 )计算当前帧的任一宏块与前一帧对应宏块的各亮度分量之差的绝对值总和; 1) calculates the current frame Renyi Hong block sum of absolute values ​​of each difference between the luminance component of the previous frame corresponding macro block;

2)才艮据该总和与门限的比较结杲,将宏块赋予一定的标记;其中门限值为经验值,通过大量实验确定。 2) It is only the sum of Burgundy threshold comparison result Gao, impart a certain macro block marker; wherein the threshold is an empirical value, determined by experiment.

第二步:才艮据第一步所判定的宏块属性标记,及其邻近宏块的标记及 Step two: before Gen macroblock data attribute tags determined in the first step, and mark neighboring macroblocks and

运动矢量等信息,分别采取相应策略:对于运动宏块或其邻近宏块不满足预定规则的可能静止宏块,不执行滤波操作,以当前宏块的亮度分量作为 Motion vector information, respectively, take the appropriate strategy: For the motion may still macroblock or macroblock neighboring macroblocks not satisfy the predetermined rules, filtering operation is not performed, the luminance components of the current macroblock as

处理结果,跳到第六步;否则继续执行第三步操作; The results, skip to step sixth; otherwise, proceed to the third step operation;

笫三步:对于由第二步判断为需要继续进行处理的宏块,根据其中各象素点与前一帧对应象素点亮度分量的绝对差值的统计信息,判断其运动属性(此时运动属性定义为运动或静止两种); Zi three steps: For the macro-block need to continue the processing by the second step that, according to the point in which each pixel of the previous frame statistics of pixel dots corresponding to the absolute difference of the luminance component, which determines the movement property (in this case attribute is defined as the movement of two kinds of moving or still);

1) 根据各对应象素点的亮度分量之间的绝对差值与设定门限的比较结果,将象素点赋予运动或静止标记; 1) based on the comparison with the set threshold absolute difference between the luminance component of each of corresponding pixel points, the pixel points given mark moving or still;

2) 分别统计该宏块的四个块中运动象素点的数目并计算其总和; 2) the number of pixels in the motion block of four points of the macroblock and statistics were calculated sum;

3 )由2 )的结果,依照预定判决规则及门限,最终确定该宏块的属性。 3) a 2) the result of decision rules in accordance with a predetermined threshold and, ultimately determining an attribute of the macroblock. 其中,预定判决规则以及门限值由大量实验结果确定。 Wherein the predetermined decision rules and thresholds are determined by a large number of experimental results.

第四步,根据第三步判定的宏块属性,分别采取相应的滤波方法:对于运动宏块,不执行滤波操作,跳到第六步;对于静止宏块,继续执行第五步操作; A fourth step, the third step according to the macro block attribute determination, the corresponding filter were taken: For macroblock motion, does not perform filtering operation jumps to a sixth step; the stationary macroblock, proceed to Step fifth operation;

第五步,对于在第三步判定为静止宏块的宏块,将其亮度分量SAD 值与设定门PM目比较判定其属性(此处的属i自定义为绝对静止或一般静止)。 A fifth step, the third step is determined in respect to the stationary macroblock in macroblock, which is the luminance component SAD value and the set gate PM mesh comparison judgment its properties (here i genus custom or general stationary stationary absolute). 然后根据其属性采取以下策略:如果纟皮判定为绝对静止宏块,则釆用前一帧对应宏块的亮度分量作为该宏块的亮度分量;反之,如果被判定为一般静止宏块,则将该宏块的各亮度分量与前一帧对应亮度分量采用预定权值逐一进行滤波,所产生的新亮度值为该宏块的处理结果。 Then take based on their properties the following strategy: If Si skin is determined absolute rest macroblock, preclude the use of the luminance component of the previous frame corresponding to the macro block as the luminance component of the macroblock; conversely, if it is determined that a general stationary macroblock, each of the luminance component corresponding to the macroblock luminance components one by one using a predetermined filter weights of the previous frame, generating new luminance values ​​of the macroblock processing result.

第六步,结束本宏块亮度分量的时域滤波处理。 A sixth step, the end of the macro-block time-domain filtering of the luminance component. (二)色度分量的处理,包括以下步骤: (B) processing the chrominance components, comprising the steps of:

第一步,计算当前帧当前宏块与前一帧对应宏块的色度分量的绝对差值总和(即SAD值)并判断其是否需要执行滤波操作; The first step, calculates the current frame in the current macroblock and the previous frame corresponding to the sum of the absolute chrominance difference components macroblock (i.e., the SAD value) and determines whether it needs to perform a filtering operation;

第二步,根据第一步的判断结果,分别采JM目应的处理方法:如判断为需要进行滤波,则将该宏块的色度分量与前一帧对应宏块色度分量采用 The second step, according to the judgment result of the first step, each entry should adopt a processing method JM: filtering as determined to be necessary, then the chrominance component corresponding macroblock of the previous frame macroblock chrominance components using

8预定权值逐一进行滤波,所产生的新象素值为色度分量的处理结果;否则, 不执行滤波操作,直接保存当前数据; 8 one by a predetermined weight value filtering processing result generated new pixel values ​​of chrominance components; otherwise, not perform the filtering operation directly save the current data;

笫三步,结束本宏块色度分量的时域滤波处理。 Zi three steps, the temporal filtering process of the present macroblock chrominance components ends.

采用本发明的方法,使编码序列中的静止宏块在序列中不同图像之间具有更强的相关性,并且克服现有技术中复杂度较高、适应性较弱的缺点, 在满足实时性要求的同时具有丰支强的鲁棒性。 Using the method of the invention, that the coding sequence of the stationary macroblock has a stronger correlation between the different images in the sequence, and to overcome the prior art high complexity, adaptive weaker disadvantages satisfying real time It required both a strong HSBC branch robustness.

附图说明 BRIEF DESCRIPTION

图l是本发明所述的时域滤波方法中亮度分量处理流程图; 图2是本发明所述的时域滤波方法中色度分量处理流程图; Figure l is a flowchart showing the temporal filtering method of the present invention, the luminance component processing; FIG. 2 is a flowchart illustrating the temporal filtering method of the present invention is processed chrominance component;

具体实施方式 Detailed ways

下面结合附图和具体实施方式对本发明所述方法作进一步描述。 Next, the method of the invention is further described in conjunction with the accompanying drawings and specific embodiments.

本发明的时域滤波方法以宏块为单位进行,作为3见频编码系统的一部分,当前帧当前宏块的滤波操作在其上一宏块的运动搜索之后、当前宏块的运动搜索之前进行,在同一宏块内部,对亮度分量和色度分量数据分别进行操作,两者的处理可以并行进行。 Temporal filtering method of the present invention in macro block units, as part of the 3 See frequency coding system, a current frame of the current macroblock filtering operation after which a macroblock motion search, prior to motion search current macroblock , inside the same macroblock, the luminance and chrominance components, respectively, manipulate the data, the processing may be performed in parallel both. 下面结合附图对本发明做进一步的详细描述。 The present invention will be described in further detail in conjunction with the accompanying drawings.

(一)亮度分量的具体实施方式 (A) the luminance component DETAILED DESCRIPTION

本发明的时域滤波方法对亮度分量的处理可具体分为宏块中运动宏块的检测与处理,可能静止并满足预定规则的宏块中绝对静止宏块与一般静止宏块的检测与滤波几部分(如图1所示)。 Temporal filtering method of the invention processing of the luminance component can be divided into specific macroblock macroblock motion detection and processing, and may still satisfies a predetermined rule absolute rest macroblock macroblock macroblock is generally stationary and filtering and detection parts (as shown in Figure 1).

第一步,计算当前帧当前宏块与前一帧对应宏块中亮度分量的绝对差值总和(即SAD值)以初步将该宏块赋予一定的属性标记,此时属性标记分别表示运动宏块、可能静止宏块。 The first step, calculates the current frame in the current macroblock and the previous frame corresponding to the sum of absolute luminance component in the macroblock difference (i.e. SAD value) of the initial macroblock to impart certain attributes tag, attribute flag case denote motion macro a block, a macroblock may be stationary. 具体包括以下步骤: It includes the following steps:

1) 计算当前宏块与前一帧对应宏块各亮度分量之差的绝对值总和(SAD) 5 The sum of absolute values ​​(SAD) 1) of the current macro block corresponding to the difference of the luminance component of each macroblock of the previous frame 5

具体实施如下: Specific embodiments are as follows:

对任意当前宏块,计算台; For any current macroblock, calculating table;

其中,"^为当前宏块与前一帧对应宏块亮度分量之间的SAD值;W 为宏块的亮度分量总数;"为当前宏块第z'个亮度分量的值;K为前一帧对应宏块第z'个亮度分量的值; Wherein "^ as SAD value between the luminance component of the macroblock corresponding to the current macroblock of the previous frame; W is the luminance component of the total number of macroblocks;" the 'value of the luminance component of the macroblock of the current z; K is a front frame corresponding macro block 'value of the luminance component of z;

2) 根据该SAD值与门限的比较结果,确定该宏块属性。 2) based on the comparison result of the SAD value with a threshold, the macroblock is determined attribute. 具体实施如下: Specific embodiments are as follows:

如果任一宏块的"^〉S,将该宏块标记为运动宏块;否则将该宏块标记为可能静止宏块。其中^为门限值,该数值通过大量的实验获得,这里选择的是一个偏大的数值,以尽量避免一些噪声较大的静止宏块被误判为运动宏块,例如,可取为1200。 If Renyi Hong block "^> S, the macro block labeled as the motion of macroblocks; otherwise, the mark may still macroblock macroblock where ^ is the threshold value, the value obtained by a lot of experiments, selected here. a value is too large, in order to avoid some of the larger stationary noise mistaken for motion macroblock is a macroblock, e.g., preferably 1200.

第二步,根据第一步判定的宏块属性标记,及其邻近宏块的标记及运动矢量等信息,分别采取相应策略: A second step, the first step is determined based on the information of the macro block attribute tags, and other markers and the motion vector of the neighboring macroblocks, respectively, take the appropriate strategy:

对于运动宏块或其邻近宏块不满足预定规则的可能静止宏块,则认为该宏块与前一帧对应位置宏块已不具备内容上的相关性,对其不执行滤波操作,以当前宏块的亮度分量作为处理结果,但不改变当前宏块属性标记, 跳到第六步;否则,继续执行第三步的操作。 For macro-block or a motion may still macroblock neighboring macroblocks does not satisfy the predetermined rules, it is considered that the position of the macroblock corresponding to the previous frame macroblock no longer have correlation in content, its filtering operation is not performed, the current luminance component processing macroblock as a result, but without changing the current attribute tag macroblock skip sixth step; otherwise, proceed to operation of the third step.

具体实施如下: Specific embodiments are as follows:

如果宏块被标记为可能静止宏块, 如果其邻近宏块不满足预定规则, If the macro block is marked as static macroblocks may, if it does not satisfy the predetermined rule neighboring macroblocks,

以当前宏块亮度分量作为处理结果,并跳到第六步; 否则继续进行第三步操作, Luminance component to the current macro block as a processing result, and skip the sixth step; otherwise, proceed to step three operations,

10否则以当前宏块亮度分量作为处理结果,并跳到第六步。 10 Otherwise, the current macro-block of the luminance component as a processing result, and skip the sixth step.

其中,这里提到的邻近宏块指的是以当前宏块为中心, 一定尺寸窗口内已经过处理(包括-见频编码中的前处理、运动搜索等操作)的一个或多 Wherein neighboring macroblocks is referred to herein means the center of the current macro block, a certain size of the window has been processed (including - see pretreatment frequency coding, motion search operations, etc.) or of a

个宏块;预定规则是指为增强方法鲁棒性所添加的附加规定,这样使得一些SAD值较小的非静止宏块在不进行后续判断的同时得到正确处理。 Macroblocks; refers to the predetermined rule is a method to enhance the robustness of a predetermined additional added, so that some of the smaller non-stationary macroblock SAD value correctly process without subsequent determination. 这些规定包括邻近宏块是否包含带运动标记的宏块(第一步的判断结果)、 带运动矢量的宏块,以及该宏块相同列方向的邻近宏块在水平方向一定范围内是否存在带运动矢量的宏块等情况,此规则由实验确定,可根据效果进行调整,实用中并不局限于此。 These include neighboring macroblocks contains the macro block (the determination result of the first step) with a motion flag, the macro block with motion vectors and macro-block of the same column if there are neighboring macroblocks in the horizontal direction with a certain range a macro block motion vectors and the like, this rule determined by experiment, can be adjusted according to the effect, practically is not limited thereto.

例如,可以采用如下失见则: For example, see the loss as follows:

1) 其左边宏块、左上边、正上、右上边宏块均为由第一步标记为可能静止宏块的宏块; 1) the macroblock to its left, upper left, is on the right upper macroblocks are marked by the first step may be stationary macroblock macroblock;

2) 其左边宏块、左上边、正上、右上边宏块均无运动矢量; 2) the macroblock to its left, upper left, is on the upper right side no macroblock motion vector;

3) 其前面三行相同列位置的宏块在其同一行上一定范围内左右两边不同时存在带运动矢量的非孤立宏块(这里指其周围至少存在一个带运动矢量的宏块)。 3) the front three rows of macro blocks in the same column position within a range of the same line on which the left and right sides of the absence of the non-isolated macroblock with a motion vector (here, around at least a macro block with motion vectors) at the same time. 具体实施时,可以将任一行上在不超过一定跨度的两个带运动矢量的非孤立宏块对之间的宏块加以特殊标记,如果前面三行相同列的宏块均未被打上此标记,则认为此宏块符合3)的规定。 In specific implementation, may be any special tag macroblocks between the non-isolated two macroblocks with motion vectors not exceed a certain line on the span, if the same three rows in front of the column of macroblocks are marked with this flag none it is considered to comply with the provisions of this macro block 3) of.

若以上三条规则同时满足则表示符合规定。 If the above three rules, it said while meeting compliance.

第三步,考虑到对于一些包含小面积运动对象的宏块,其SAD值可能小于第一步所设定的门限,并且其邻近宏块满足第二步设定的规则,故需要对可能静止宏块进行进一步的判断以排除其中的非静止宏块:根据其中各象素点与前一帧对应象素点亮度分量绝对差值的统计信息,判断其运动属性,其属性最终自定义为运动或静止两种。 The third step, taking into account some small area comprises a macro block of the moving object, which may be less than the first step SAD value set threshold, and which second step of neighboring macroblocks satisfy the rule set, you may still need to further determines macroblocks to eliminate non-stationary macroblocks therein: wherein each pixel dots in accordance with the corresponding pixel of the previous frame statistics point of the absolute difference of the luminance component of its motion is determined attribute, which is defined as the movement from the final or still two kinds. 具体包括以下步骤: It includes the following steps:

1)根据各对应象素点的亮度分量之间的绝对差值与门限的比较结果, 将象素点赋予运动或静止标记。 1) According to the absolute difference between the comparison result of the threshold luminance component of each corresponding pixel points, the pixel points given mark moving or static.

ii具体实施如下: ii particular embodiment as follows:

令£>砍=|"-y」,如果"砍>&,令M^,-1;否则,令M^,-0 Make £> cut = | "-y" if "cut> &, so that M ^, - 1; otherwise, let M ^, - 0

其中,"砍表示该宏块与前一帧对应宏块第z'个亮度分量之差的绝对 Wherein, "cut indicates the macroblock of the macroblock difference z 'of the luminance component of the previous frame corresponding to the absolute

值;y" K的定义与前面相同;M"^'为第^个象素点的标记;^为将象素点判断为运动的门限值,其取值由大量实验确定,本发明中取为8。 Value; y "K is the same as defined previously; M" ^ 'of marked pixel points ^; ^, whose value is determined by a large number of test points is determined by the pixel motion threshold, the present invention take 8.

2)在该宏块的亮度分量中以块为单位分别统计其运动象素点的数目并计算其总和; 2) In the luminance component of the macroblock in units of blocks in which the number of statistical points are motion pixels and calculating the sum thereof;

具体实施如下: Specific embodiments are as follows:

尸乂= J]M"A Sw/w = Dead qe = J] M "A Sw / w =

, y=】 , Y =]

其中^表示第7个块的运动象素点数;M表示一个块的亮度分量总数;S"w表示该宏块的总运动象素点数。 Wherein ^ represents the motion block of pixel dots 7; M represents the total number of the luminance component of a block; S "w represents the total movement of the macro block pixel points.

3 )由2)的结果,依照预定判决规则及门限,最终确定该宏块的属性。 3) a 2) the result of decision rules in accordance with a predetermined threshold and, ultimately determining an attribute of the macroblock. 具体实施如下: Specific embodiments are as follows:

如杲任一巧>7^或5^>?;,判断该宏块为运动宏块;否则,判断该宏 As any one of Gao Qiao> or 5 ^ 7 ^>;?, Determines the macroblock is a macroblock motion; otherwise, determines that the macro

块为静止宏块。 Block stationary macroblocks. 其中,门限值L、 ^分别表示块与宏块中运动象素点数门限,其取值由大量实验结果确定,本发明中分别取3、 6。 Wherein the threshold value L, ^ denote the block motion threshold macroblock pixel points, whose value is determined by a large number of experimental results, the present invention taken respectively 3, 6.

第四步,才艮据第三步判定的宏块属性,采取相应的滤波方法:对于运动宏块,不执行滤波操作,跳到第六步;对于静止宏块,继续执行第五步操作; A fourth step, only according to macroblock Gen third attribute determination step, take the appropriate filtering method: for the macroblock motion, does not perform filtering operation jumps to a sixth step; the stationary macroblock, proceed to Step fifth operation;

第五步,对于由第三步判定为静止宏块的宏块,4艮据其第一步计算的亮度分量的SAD值与设定门限的比较结果判定其属性(此处的属性自定义为绝对静止或一般静止),并根据其属性采糾目应的策略,具体实施如下: A fifth step, the third step for the macroblock by a determined stationary macroblocks, Gen 4 SAD value and the set according to the luminance component of the door which limit the first step in calculating the comparison result of the determination of its attribute (as defined herein from Usually absolute rest or stationary), and according to its properties correct entry should mining strategy, the following specific embodiments:

如果&4Z),T2,对宏块中各亮度分量:iX+^;';否则,对宏块中各亮度分量:K=y>'。 If & 4Z), T2, the respective macroblock luminance components: iX + ^; '; otherwise, the respective macroblock luminance components: K = y>'. 其中&为门限值,由实验确定,本发明中取400; y" K的定义与前面相同;"、 〃表示权值系数,由实验进行设置,也可以根据需要使其随宏块的SAD值或该象素与其对应象素亮度分量的绝对差值等信息自适应变化,本发明中各取0.5,相当于执行均值操作。 Wherein & is the threshold value, it is determined experimentally, the present invention taken 400; y "K the same meaning as defined above;" 〃 is the weight coefficient set by experiment, it may be a macroblock SAD with the necessary the pixel value or the like corresponding to its absolute difference of pixel luminance component adaptively changes, the present invention depicting 0.5, average equivalent operations performed.

第六步,结束本宏块亮度分量的时域滤波处理。 A sixth step, the end of the macro-block time-domain filtering of the luminance component.

(二)色度分量的具体实施方式 (Ii) the chrominance components of the specific embodiments

本发明的时域滤波方法中对色度分量的处理比较筒单,下面结合图2 进行说明。 Temporal filtering method of the present invention in a single process cartridge comparison of the chrominance component, will be described below in conjunction with FIG.

第一步,计算当前帧当前宏块与前一帧对应宏块色度分量的绝对差值总和(即SAD值)并判断其是否需要执行滤波操作。 The first step, the current frame is calculated and the sum of absolute difference between a current macroblock in a corresponding macro block of the chroma component (i.e., the SAD value) and determines whether it needs to perform a filtering operation. 具体实施如下: Specific embodiments are as follows:

丄,、l入雄-=1;—-一) 对任一宏块,令^ ; Into male Shang ,, l - = 1; - a) a block of Renyi Hong, so ^;

其中&4Z^为当前宏块与前一帧对应宏块色度分量的sAD值;w为宏 Wherein & 4Z ^ for the current chrominance component corresponding macro block of the previous frame macroblock sAD value; W macro

块的色度分量总数;[^为当前宏块第z'个色度分量的值;^^'为前一帧对应宏块第z'个色度分量的值。 The total number of chrominance component block; [^ macroblock of the current z 'value chrominance components; ^^' value of the previous frame corresponding macro block of z 'chrominance components.

第二步,根据该SAD值与门限的比较结果,分别釆拟目应的处理方法。 The second step, according to the comparison result of the SAD value and the threshold, respectively, preclude entry should the proposed processing method. 具体实施如下: Specific embodiments are as follows:

如果该宏块"化^713,则t/^"t/K,+"f7)/^否则,不执行滤波操作。 其中,^、 "^的定义与前同。 If the macroblock "of ^ 713, then t / ^" t / K, + "f7) / ^ otherwise, not perform the filtering operation wherein, ^" ^ the same as previously defined. ^为门限值,通过大量实验获得,本发明中取为200。 ^ Is a threshold, obtained by experiment, the present invention is taken as 200.

第三步,结束本宏块色度分量的时域滤波处理。 The third step, when the chrominance components of the present macroblock temporal filtering process ends.

时域滤波方法是视频压缩中视频序列预处理的一个重要技术,会直接影响图像的主观质量。 Temporal filtering method is an important technique in video compression video sequence pretreated, will directly affect the subjective quality of the image. 本发明的滤波方法可广泛应用于视频编码系统中, 方法计算复杂度较低,并且在一定的传输速率下可以获得实时高质的图像效果,具有较高的实用价值。 Filtering method according to the present invention can be widely used in video coding system, a low computational complexity method, and can obtain high quality images in real time effect under certain transmission rate, it has a high practical value.

13 13

Claims (11)

  1. 1、一种视频编码的时域滤波方法,其特征在于,当前帧当前宏块的滤波操作在其上一宏块的运动搜索之后、当前宏块的运动搜索之前进行,在同一宏块内部,对亮度分量和色度分量数据分别进行如下操作: (一)亮度分量的处理,包括以下步骤: 第一步:计算当前帧当前宏块与前一帧对应宏块中亮度分量的绝对差值总和(SAD)值,以初步将当前宏块赋予属性标记,所述属性标记是运动宏块或可能静止宏块; 第二步:根据第一步所判定的当前宏块的属性标记,及该当前宏块的邻近宏块的属性标记及运动矢量信息,分别进行下述处理:对于属性标记为运动宏块的当前宏块,或属性标记为可能静止宏块的当前宏块且该当前宏块的邻近宏块不满足预定规则,则不执行滤波操作,以当前宏块的亮度分量作为处理结果,跳到第六步;否则继续执行第三步操作; 第 1. A video encoding method for temporal filtering, characterized in that the current macroblock of the current frame after filtering operation on a macroblock motion search, prior to the current macro block motion search, inside the same macroblock, luminance and chrominance components of the data are as follows: (a) processing the luminance component, comprising: a first step: calculating the sum of absolute difference corresponding to the luminance component of the macroblock of the current frame and a previous frame of the current macroblock (SAD) value, the initial current macroblock to impart attribute tags, the tag attribute is a motion macroblock or macroblock may be stationary; Step: the first step in determining the properties of the current macro block flag, and the current attribute tags and the motion vectors of neighboring macroblocks of the macroblock, respectively, the following process: the current macro block or the tag for the attribute is marked as the current macroblock motion may still macroblock macroblock and the current macroblock neighboring macroblocks does not satisfy the predetermined rules, a filtering operation is not performed, the luminance components of the current macro block as a result of the processing, jumps to a sixth step; otherwise, proceed to step operation; first 步:根据当前宏块各象素点亮度分量与前一帧对应宏块的各对应象素点亮度分量的绝对差值的统计信息,判断当前宏块的运动属性,所述运动属性是运动或静止; 第四步:根据第三步判定的当前宏块的运动属性,分别采取相应的滤波方法:对于运动属性为运动的当前宏块,不执行滤波操作,跳到第六步;对于运动属性为静止的当前宏块,继续执行第五步操作; 第五步:将当前宏块的亮度分量的SAD值与设定门限相比较,判定当前宏块的属性,所述属性自定义为绝对静止或一般静止,如果被判定为绝对静止宏块,则采用前一帧对应宏块的亮度分量作为该当前宏块的亮度分量;反之,如果被判定为一般静止宏块,则将该当前宏块的各亮度分量与前一帧对应亮度分量采用预定权值逐一进行滤波,所产生的新亮度值为该当前宏块的处理结果; 第六步:结 Step: the current macroblock luminance component of each pixel point and statistics of the previous frame corresponding to the absolute difference of pixel dots corresponding to the respective macroblock luminance components of the motion is determined according to the properties of the current macroblock, the motion is a motion attribute or still; fourth step: the attribute of the current macroblock according to the motion of the third step of the determination, the corresponding filter were taken: for motion attribute current macroblock motion, does not perform filtering operation jumps to a sixth step; attribute for movement current macroblock is stationary, operations proceed to fifth step; step 5: the luminance component of the current macroblock SAD value and the set threshold comparison, determining an attribute of the current macro block, the custom properties as absolute rest or generally stationary, it is determined if the absolute rest macroblock, a luminance component corresponding to the previous frame is used as a macro block luminance component of the current macro block; the other hand, if it is determined that a general stationary macroblock, the current macroblock individually each of the luminance component with a predetermined weight value corresponding to the luminance component of a previous frame using the filter, the new value of the brightness of the generated current macroblock processing result; step Six: junction 本当前宏块亮度分量的时域滤波处理; (二)色度分量的处理,包括以下步骤: 第一步:计算当前帧当前宏块与前一帧对应宏块的色度分量的绝对差值总和; 第二步:根据所述色度分量的绝对差值总和,判断当前宏块是否需要执行滤波操作,若不需要执行滤波操作,则以当前宏块的色度分量作为处理结果,若需要执行滤波操作,则将该当前宏块的色度分量与前一帧对应色度分量采用预定权值逐一进行滤波,所产生的新象素值为色度分量的处理结果; 第三步:结束本当前宏块色度分量的时域滤波处理。 This time-domain filtering of the current macroblock luminance component; (ii) processing the chrominance components, comprising: a first step: calculating the absolute difference between the current macroblock of the current frame and the previous frame chrominance component corresponding macro block sum; the second step: the sum of the absolute difference chrominance components, it is determined whether the current macroblock filtering operation needs to be performed, performs a filtering operation if required, places the current macroblock as a chroma component of the processing result, if desired performs a filtering operation, then the processing result the chrominance component of the current macro block with a predetermined filter weights corresponding to the chrominance component using a previous frame one by one, the new pixel value is generated chrominance components; the third step: end temporal macro block filtering process according to the current chrominance component.
  2. 2、 根据权利要求1所述视频编码的时域滤波方法,其特征在于, 所述的当前帧指未经时域滤波的当前图像,所述的前一帧指经过时域滤波的前帧图像。 Temporal filtering method, according to claim 1 video encoding, wherein the current frame image refers to the current time domain without filtering, the previous frame image refers to a temporal filtering through .
  3. 3、 根据权利要求1所述视频编码的时域滤波方法,其特征在于, 所述亮度分量处理第一步中将当前宏块赋予属性标记具体包括以下步骤:将SAD值与门限进行比较,若SAD值大于所述门限,则将当前宏块的属性标记设为运动宏块;否则将当前宏块的属性标记设为可能静止宏块。 3 temporal filtering method, according to claim 1 video encoding, wherein the processing of the luminance component in the first step of the current macro block imparting property tag includes the following steps: SAD value is compared with a threshold, if SAD value is greater than the threshold, then the current attribute of the macroblock motion flag to macro blocks; otherwise the attribute flag to the current macro block may rest macroblocks.
  4. 4、 根据权利要求1所述视频编码的时域滤波方法,其特征在于, 所述亮度分量处理第三步具体包括以下步骤:1) 根据当前宏块与前一帧对应宏块各对应象素点的亮度分量之间的绝对差值与设定门限的比较结果,将象素点赋予运动或静止标记;2) 在该当前宏块的亮度分量中以块为单位分别统计块内运动象素点的数目并计算当前宏块运动象素的数目的总和;3) 由2)的结杲,依照预定判决规则及门限,最终确定该当前宏块的运动属性。 4 the temporal filtering method, according to claim 1 video encoding, wherein the luminance component processing the third step includes the following steps: 1) before the current macro block with a macro block corresponding to each of the corresponding pixel comparison with the absolute difference between the set threshold luminance component points, the pixel points given mark moving or still; 2) in the luminance component of the current macro block in units of blocks in the block moving pixel count respectively and calculating the sum of the number of points of the current macro-block moving pixel number; and 3) a 2) of junction Gao, in accordance with predetermined rules and decision thresholds, which ultimately determines the properties of the current macroblock motion.
  5. 5、 根据权利要求4所述视频编码的时域滤波方法,其特征在于,在步骤l)中,该当前宏块与前一帧对应宏块第/个亮度分量之差的绝^H直Z)砍—K一a , ^口果"砍>7;,令M"A,. =1,否贝寸令MaA,.-0;其中,y;、 y/分别是当前宏块和前一帧对应宏块第/个亮度分量; m"w^为第z个象素点的标记;rrf为将象素点判断为运动的门限值。 5 temporal filtering method, according to claim 4 video encoding, characterized in that in step l), the current macroblock of the macroblock absolute / differential luminance component of a previous frame corresponding to a linear Z ^ H ) a cut -K a, ^ mouth fruit "cut> 7 order ;, M" a ,. = 1, so that no shellfish inch MaA, .- 0; wherein, y ;, y / are the current and previous macroblock corresponding macro block of the frame / luminance component; m "w ^ z of the marked pixel points; RRF point is determined by the pixel value motion of the door.
  6. 6、 根据权利要求4所述4见频编码的时域滤波方法,其特征在于,在步骤2 )中,尸;=f]Ma^ , S柳=;其中^表示第/个块的运动象素点数;M表示一个块的亮度分量象素总数;S,表示该当前宏块的总运动象素点数。 6, according to claim see frequency coding temporal filtering method according to claim 4, wherein, in step 2), P; = f] Ma ^, S = Liu; where ^ represents the motion of / th block like Su points; M represents a total number of the luminance component of the pixel block; S, represents the total movement of the current macro block pixel points.
  7. 7、 根据权利要求4所述视频编码的时域滤波方法,其特征在于, 在步骤3)中,如果任一^>7;或^附>7;,则判断该当前宏块为运动宏块,否则判断该当前宏块为静止宏块;其中,门限值r。 7 temporal filtering method, according to claim 4 video coding, wherein, in step 3), if either ^> 7; ^ is attached or> 7 ;, it is judged that the current macroblock is a macroblock motion otherwise, the current macroblock is determined as the still macroblock; wherein the threshold value r. 、 7;分别表示块与宏块中运动象素点数门限。 , 7; represent the motion block and macro block pixel count threshold.
  8. 8、 根据权利要求1所述^L频编码的时域滤波方法,其特征在于, 所述亮度分量处理第五步中,判定当前宏块的属性的具体方式和相应采用的滤波器为:如果&4£>>72,则当前宏块的属性为一般静止,且采用的滤波器为:gc^+;^';否则当前宏块的属性为绝对静止,且采用的滤>发器为:《=17其中72为门限值;K、"'分别是当前宏块和前一帧对应宏块第f个亮度分量,g是当前宏块第/个亮度分量的滤波结果;《、"表示权值系数。 8. ^ L time-domain coding the frequency filtering method as claimed in claim 1, wherein said luminance component processing the fifth step determines the specific embodiment of the current macroblock attributes and the corresponding filter is used: if & 4 £ >> 72, the current attribute of the macroblock is generally stationary, and the filter is used: gc ^ +; ^ '; otherwise the attribute is the current macroblock absolute rest, and the use of the filter> hair is: " = 72 to 17, wherein the threshold value; K, " 'are the current macroblock and the previous frame corresponding macro block of the luminance component of f, g is the result of the filtering of the current macro block / luminance component;" "represents a weight value coefficient.
  9. 9、 根据权利要求8所述视频编码的时域滤波方法,其特征在于, 72取400; "、 /?各耳又0.5。 9 temporal filtering method, according to claim 8 video encoding, characterized in that, taking 72 400; "/ each ear and 0.5?.
  10. 10、 根据权利要求1所述视频编码的时域滤波方法,其特征在于, 所述色度分量处理第一步中,具体实施如下:对任一当前宏块,令<formula>formula see original document page 5</formula>其中&4Z^为当前宏块与前一帧对应宏块色度分量的SAD值;iV为宏块的色度分量象素总数;t/p;为当前宏块色度分量第/个象素值;"K为前一帧对应宏块色度分量第/个象素值。 10 temporal filtering method, according to claim 1 video encoding, wherein said first step of processing the chrominance component, the specific embodiments as follows: for any one of the current macroblock, so <formula> formula see original document page 5 </ formula> where ^ & macroblock corresponding to the current chrominance component of the previous frame macroblock SAD value 4Z; the iV chrominance component of the pixel is the total number of macroblocks; t / p; chroma component of the current macroblock second / pixel values; "K component of the first / pixel values ​​of the corresponding macroblock in the previous frame chrominance.
  11. 11、根据权利要求io所述视频编码的时域滤波方法,其特征在于, 所述色度分量处理第二步中,具体实施如下:如果该当前宏块<formula>formula see original document page 5</formula>否则不执4亍滤波操作;其中K为门限值。 11, the temporal filtering method according to claim io the video encoding claims, characterized in that, the chrominance component processing the second step, the specific embodiments as follows: If the current macroblock <formula> formula see original document page 5 < / formula> otherwise does not perform a filtering operation right foot 4; wherein K is a threshold value.
CN 200410088837 2004-11-05 2004-11-05 Time-domain filtering method for video code CN100562129C (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 200410088837 CN100562129C (en) 2004-11-05 2004-11-05 Time-domain filtering method for video code

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 200410088837 CN100562129C (en) 2004-11-05 2004-11-05 Time-domain filtering method for video code

Publications (2)

Publication Number Publication Date
CN1770880A true CN1770880A (en) 2006-05-10
CN100562129C true CN100562129C (en) 2009-11-18

Family

ID=36751826

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 200410088837 CN100562129C (en) 2004-11-05 2004-11-05 Time-domain filtering method for video code

Country Status (1)

Country Link
CN (1) CN100562129C (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101257630B (en) 2008-03-25 2012-06-06 浙江大学 Video frequency coding method and device combining with three-dimensional filtering
WO2012068085A3 (en) * 2010-11-19 2012-08-09 Analog Devices, Inc. Component filtering for low-light noise reduction
JP5907367B2 (en) * 2011-06-28 2016-04-26 ソニー株式会社 An image processing apparatus and method, program, and recording medium
CN102970541B (en) * 2012-11-22 2015-05-27 深圳市海思半导体有限公司 Image filtering method and device

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2000023190A (en) 1998-06-26 2000-01-21 Sony Corp Motion vector generation method, picture encoder, motion compensation method, motion compensation device and provision medium
US6175593B1 (en) 1997-07-30 2001-01-16 Lg Electronics Inc. Method for estimating motion vector in moving picture
JP2002369206A (en) 2001-06-07 2002-12-20 Matsushita Electric Ind Co Ltd Device and method for selective encoding of dynamic region and static region
CN1514649A (en) 2003-08-11 2004-07-21 深圳市中太多媒体通信有限公司 Image pretreatment method used for video frequency signal compression

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6175593B1 (en) 1997-07-30 2001-01-16 Lg Electronics Inc. Method for estimating motion vector in moving picture
JP2000023190A (en) 1998-06-26 2000-01-21 Sony Corp Motion vector generation method, picture encoder, motion compensation method, motion compensation device and provision medium
JP2002369206A (en) 2001-06-07 2002-12-20 Matsushita Electric Ind Co Ltd Device and method for selective encoding of dynamic region and static region
CN1514649A (en) 2003-08-11 2004-07-21 深圳市中太多媒体通信有限公司 Image pretreatment method used for video frequency signal compression

Also Published As

Publication number Publication date Type
CN1770880A (en) 2006-05-10 application

Similar Documents

Publication Publication Date Title
Liu et al. Efficient DCT-domain blind measurement and reduction of blocking artifacts
US5757969A (en) Method for removing a blocking effect for use in a video signal decoding apparatus
US6473532B1 (en) Method and apparatus for visual lossless image syntactic encoding
US6281942B1 (en) Spatial and temporal filtering mechanism for digital motion video signals
US20010036320A1 (en) Method and apparatus for dynamic loop and post filtering
US6807317B2 (en) Method and decoder system for reducing quantization effects of a decoded image
US5926572A (en) Method of determining coding type and coding mode for object shape coding
US7068722B2 (en) Content adaptive video processor using motion compensation
US20060245499A1 (en) Detection of artifacts resulting from image signal decompression
US20100080472A1 (en) Image processing apparatus, moving image decoding apparatus, moving image encoding apparatus and method
JPH09247681A (en) Dynamic image decoding method and dynamic image decoder
US6862372B2 (en) System for and method of sharpness enhancement using coding information and local spatial features
CN101448077A (en) Self-adapting video image 3D denoise method
CN1925614A (en) Rapid movement estimating method
US20100027905A1 (en) System and method for image and video encoding artifacts reduction and quality improvement
Töreyin et al. Moving object detection in wavelet compressed video
CN1444406A (en) Vidio motion estimation method
CN1465193A (en) Image encoder, image decoder, image encoding method, and image decoding method
JP2002051341A (en) Moving picture coding apparatus and apparatus of detecting special effect scene in the moving picture
CN1960491A (en) Real time method for segmenting motion object based on H.264 compression domain
CN1761286A (en) Method for detecting movement detection by using edge detection, and for removing ripple noise through medium filtering
CN101321287A (en) Video encoding method based on movement object detection
JPH08149479A (en) Extremely low bit rate video encoding method
CN102025997A (en) Method and device for concealing information as well as method and device for extracting concealed information
WO2002102086A2 (en) Apparatus and method for adaptive spatial segmentation-based noise reducing for encoded image signal

Legal Events

Date Code Title Description
C06 Publication
C10 Request of examination as to substance
C14 Granted
COR Bibliographic change or correction in the description

Free format text: CORRECT: ADDRESS; FROM: 518057 SHENZHEN, GUANGDONG PROVINCE TO: 210000 NANJING, JIANGSU PROVINCE

ASS Succession or assignment of patent right

Owner name: NANJING YUHUATAI INTELLECTUAL PROPERTY CENTER

Free format text: FORMER OWNER: ZTE CORPORATION

Effective date: 20140924

C41 Transfer of the right of patent application or the patent right
C41 Transfer of the right of patent application or the patent right
COR Bibliographic change or correction in the description

Free format text: CORRECT: ADDRESS; FROM: NANJING, JIANGSU PROVINCE TO: 210012 NANJING, JIANGSU PROVINCE

ASS Succession or assignment of patent right

Owner name: NANJING KUNQIAN COMPUTER TECHNOLOGY CO., LTD.

Free format text: FORMER OWNER: NANJING YUHUATAI INTELLECTUAL PROPERTY CENTER

Effective date: 20141209