CN101742086B - Method for eliminating image noise - Google Patents

Method for eliminating image noise Download PDF

Info

Publication number
CN101742086B
CN101742086B CN 200810174849 CN200810174849A CN101742086B CN 101742086 B CN101742086 B CN 101742086B CN 200810174849 CN200810174849 CN 200810174849 CN 200810174849 A CN200810174849 A CN 200810174849A CN 101742086 B CN101742086 B CN 101742086B
Authority
CN
Grant status
Grant
Patent type
Prior art keywords
image
data
portion
low
filter
Prior art date
Application number
CN 200810174849
Other languages
Chinese (zh)
Other versions
CN101742086A (en )
Inventor
江铭峰
Original Assignee
联咏科技股份有限公司
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

一种影像噪声消除方法。 An image noise cancellation method. 该影像噪声消除方法,包括接收一影像且对该影像进行一第一阶段处理,以得到在YCbCr座标空间(domain)下分离出对应于一像素阵列的一亮度信息Y以及一色彩信息Cb与/或Cr。 The image noise removing method, comprising receiving an image and the image for a first stage of treatment, to obtain a separated space in the YCbCr coordinates (Domain) corresponding to a pixel array of a luminance information Y and color information Cb and a / or Cr. 针对该亮度信息Y,进行一第二阶段处理,以至少减少一亮度噪声。 For the luminance information Y, performing a second process stage, at least a luminance noise reduction. 针对该色彩信息Cb与/或Cr,进行一第三阶段处理,以至少减少一色彩噪声。 For the color information Cb and / or Cr, performing a third stage of treatment, at least a color noise reduction. 将该亮度信息Y与该色彩信息Cb与/或Cr组合。 The luminance information Y and the color information Cb / Cr or the combination.

Description

影像噪声消除方法 Image noise elimination method

技术领域 FIELD

[0001] 本发明涉及一种影像处理技术,特别是涉及一种对亮度信息Y以及一色彩信息Cb与/或Cr分别滤波(filtering)的影像压缩技术。 [0001] The present invention relates to an image processing technique, particularly to a video of a luminance information Y and color information Cb and / or Cr each filter (Filtering) compression technique.

背景技术 Background technique

[0002] 影像处理技术例如随着近年数字相机的发展而要求更加有效的处理机制。 [0002] With the development of image processing technology in recent years, such as digital cameras and more effective claims handling mechanism. 一般消费者在数字相机的彩色照片,为了有防手振的要求,需要以低亮度拍摄,其也使得具有高感度(High ISO)消除噪声功效的要求越来越高。 The average consumer in a digital color photo camera, there are requirements for anti-shake, it is necessary to lower the brightness of the shooting, which also makes the claim has high sensitivity (High ISO) noise-canceling effect of higher and higher. 一般而言,越低的感度,其噪声较小。 In general, the lower the sensitivity, the smaller the noise. 越高的感度,其噪声也就越大。 The higher the sensitivity, the greater the noise. 如何抑制由于高感度所产生的影像噪声为一项课题。 How to suppress image noise generated by the high sensitivity to a topic. 其中噪声部分,尤其以彩色噪声令人更难以接受。 Wherein the noise component, particularly in the color noise What is even more difficult to accept.

[0003] 一般而言,要处理成一张数字相机的照片,数字影像处理器,一般而言需要有几个大的重要单元:感测接口模块(Sensor Interface Module),影像路径模块(ImagePipeline Module ),缩放模块(Scalar Module),以及压缩模块(Jpeg Module)。 [0003] In general, a digital camera to be processed as a photograph, a digital image processor, generally requires a large number of significant elements: a sensing interface module (Sensor Interface Module), the image path module (ImagePipeline Module) scaling module (Scalar module), and compression module (Jpeg module). 影像经过一级一级的处理,就可以得到一般的影像压缩文件。 After processing an image level, you can get a general image compression file.

[0004] 一般而言,由感测器输出的影像是像素阵列的影像,可能是RGGB的Bayer图案(pattern)或者是补色CMYG之类的图案。 [0004] In general, an image is output from the image sensor pixel array, a Bayer pattern may be RGGB (pattern) or CMYG complementary color pattern or the like. 后端影像处理的影像路径模块会转为RGB再转为YCbCr经过缩放模块的缩小或放大后给压缩模块去做压缩成一般所知的jpg文件。 Back-end image processing module will be converted into RGB image path and then converted jpg file YCbCr scaled to reduce or enlarge the module after compression module to compress into do generally known.

[0005] 然而传统影像处理中,没有对亮度噪声与色彩噪声做有效的抑制。 [0005] However, the conventional image processing, do not effectively suppress the luminance noise and color noise.

发明内容 SUMMARY

[0006] 本发明提供一种影像处理机制,在YCbCr的座标空间(domain)下,分别针对亮度信息Y以及一色彩信息CbCr做滤波处理。 [0006] The present invention provides an image processing mechanism, in the coordinate space of YCbCr (Domain), were done for filtering a luminance information Y and color information CbCr.

[0007] 本发明提出一种影像噪声消除方法,包括接收一影像且对该影像进行一第一阶段处理,以得到在YCbCr座标空间(domain)下分离出对应于一像素阵列的一亮度信息Y以及一色彩信息Cb与/或Cr。 [0007] The present invention provides an image noise removing method, comprising receiving an image and the image for a first stage of treatment, to obtain a luminance information corresponding to a pixel array is separated in the YCbCr space coordinates (Domain) a Y and color information Cb and / or Cr. 针对该亮度信息Y,进行一第二阶段处理,以至少减少一亮度噪声。 For the luminance information Y, performing a second process stage, at least a luminance noise reduction. 针对该色彩信息Cb与/或Cr,进行一第三阶段处理,以至少减少一色彩噪声。 For the color information Cb and / or Cr, performing a third stage of treatment, at least a color noise reduction. 将该亮度信息Y与该色彩信息Cb与/或Cr组合,其中该第二阶段处理与该第三阶段处理都是以藉由一低通滤波方式进行,该低通滤波方式包括: The luminance information / or Cr in combination with Y and the color information Cb, and wherein the second stage of the process at the third stage of the process are filtered by a low-pass manner, the low-pass filtering mode comprising:

[0008] 将在该像素阵列的行与列的一第一方向上的每一条数据串分成一第一部分数据以及一第二部分数据,该第一部分数据是该数据串,该第二部分数据相对该第一部分数据在该第一方向上相对移位η个像素,n ^ 1,其中该第二部分数据超出该第一部分数据的一边界区域的η像素被设定为一预定值; [0008] Each of the data string in a first direction of the row and column of the pixel array is divided into a first portion and a second data part of the data, the first portion of the data string data, data corresponding to the second portion the first portion of the data in the first direction relative displacement η pixels, n ^ 1, wherein the second portion η pixel data exceeds a boundary area of ​​the first portion of the data is set to a predetermined value;

[0009] 将该第一部分数据与该第二部分数据平均,得到一低通滤波影像;以及 [0009] The first partial data and the second partial data averaged to obtain a low pass filtered image; and

[0010] 根据该低通滤波影像以及该低通滤波方式的前述的步骤,进行一递回(recurrence)处理,以达到一所要的滤波阶数。 [0010] According to this low-pass filtering of the image and the low-pass filtering step mode, performing a recursive (Recurrence) process to achieve a number of the order to the filter.

[0011] 依据一实施例,在所述的影像噪声消除方法中,还包括将在该像素阵列的行与列的一第二方向上的每一条数据串分成一第一部分数据以及一第二部分数据,该第一部分数据是该数据串,该第二部分数据相对该第一部分数据在该第二方向上相对移位m个像素,m3 1,其中该第二部分数据超出该第一部分数据的一边界区域的m像素被设定为一预定值; [0011] According to one embodiment, the image noise eliminating process further comprising each of the data strings in a second direction of the row and column of the pixel array is divided into a first portion and a second portion of the data data, the first portion of the data string is a data, data corresponding to the second portion of the first portion of data in the second direction relative displacement m pixels, m3 1, wherein the second portion of the data exceeds a first portion of the data m pixel boundary region is set to a predetermined value;

[0012] 将该第一部分数据与该第二部分数据平均,得到一低通滤波影像;以及 [0012] The first partial data and the second partial data averaged to obtain a low pass filtered image; and

[0013] 根据该低通滤波影像以及该低通滤波方式的前述的步骤,进行一递回处理,以达到该所要的滤波阶数,且实现二维的一影像滤波处理。 [0013] According to this low-pass filtered image and the low-pass filtering of the embodiment step, a recursive process, the filter order to achieve the desired, and to realize the two-dimensional image of a filtering process.

[0014] 依据一实施例,在所述的影像噪声消除方法中,该所要的滤波阶数可由外部选择。 [0014] According to an embodiment, the image noise removing method, the order number of the filter may be external to be selected.

[0015] 依据一实施例,在所述的影像噪声消除方法中,还包括一边缘判断机制决定一边界区域,并且针对该边界区域调整该所要的滤波阶数。 [0015] According to an embodiment, the image noise removing method further includes a determining mechanism for determining a boundary of an edge region, and adjusting the filtering order of the desired number for the boundary region.

[0016] 依据一实施例,在所述的影像噪声消除方法中,在边界区域的该所要的滤波阶数可由外部调整。 [0016] one embodiment, the image noise elimination method, can be adjusted according to the external boundary of the filter order in the desired region.

[0017] 依据一实施例,在所述的影像噪声消除方法中,经过该低通滤波方式处理后的一低通滤波影像,还包括将该低通滤波影像与该影像的一高通滤波影像再做一权重处理。 [0017] According to one embodiment, the image noise eliminating method, after a low-pass filtering the image after the low pass filtering process embodiment, further comprising low pass filtering the image with a further image of the high pass filtered image do a weight processing.

[0018] 为使本发明的上述和其他目的、特征和优点能更明显易懂,下文特举较佳实施例,并结合附图详细说明如下。 [0018] The above and other objects of the present invention, features and advantages can be more fully understood by reading the following preferred embodiments, in conjunction with the accompanying drawings and described in detail below.

附图说明 BRIEF DESCRIPTION

[0019] 图1示出了依据本发明实施例,影像处理的方法示意图。 [0019] FIG. 1 shows a schematic diagram of a method embodiment, the image processing of the embodiment of the present invention.

[0020] 图2示出了Y、Cb、Cr对应像素阵列的分布。 [0020] FIG. 2 shows a Y, Cb, Cr corresponding to the distribution of the pixel array.

[0021] 图3示出了巴斯卡三角形的系数关系。 [0021] FIG. 3 shows a relationship between the coefficient Pascal triangle.

[0022] 图4示出了依据本发明实施例,巴斯卡三角形低通滤波器在X方向的滤波机制示意图。 [0022] FIG 4 illustrates an embodiment according to the present invention, a low pass filter schematic Pascal triangle filter mechanism in the X direction.

[0023] 图5示出了依据本发明实施例,巴斯卡三角形低通滤波器在y方向的滤波机制示意图。 [0023] FIG. 5 shows an embodiment according to the present invention, a low pass filter schematic Pascal triangle filter mechanism in the y-direction.

[0024] 附图符号说明 [0024] BRIEF DESCRIPTION OF REFERENCE NUMERALS

[0025] 100 〜116:步骤 [0025] 100 ~116 steps of:

[0026] 120:Y 阵列 [0026] 120: Y array

[0027] 122:Cb 阵列 [0027] 122: Cb array

[0028] 124:Cr 阵列 [0028] 124: Cr Array

[0029] 140,240:原始像素阵列 [0029] 140, 240: an original pixel array

[0030] 142,242:第一部份数据 [0030] 142, 242: first partial data

[0031] 144、242:第二部份数据 [0031] 144,242: The second part of the data

[0032] 146,246:第一阶滤波影像 [0032] 146, 246: a first order filter image

[0033] 148,248:第一部份数据 [0033] 148, 248: first partial data

[0034] 150,250:第二部份数据 [0034] 150, 250: The second part of the data

[0035] 152,252:第二阶滤波影像 [0035] 152, 252: second-order filter image

具体实施方式[0036] 本发明提供一种影像处理机制,在YCbCr的座标空间(domain)下,利用算数处理加速器实现一种阶数可调整的滤波器,并利用不同阶数对应到不同的频率来实现YCbCr空间上是可调阶数的且是有边缘判断的低通率波器,这样可有效针对不同的频率降低影像的亮度彩色噪声,并同时通过变形运算来增强影像边缘。 DETAILED DESCRIPTION [0036] The present invention provides an image processing mechanism, in the coordinate space of YCbCr (Domain), using an arithmetic processing accelerator realize an adjustable filter order and use different orders corresponding to the number of different frequencies to achieve the YCbCr space are adjustable and are of the order of determination of a low-pass edge of the wave, so that can be effective for different frequency color noise reduced luminance image, and simultaneously to enhance the image edges by deforming operation.

[0037] 基本上,本发明将影像分为Y亮度及C彩度。 [0037] Basically, the present invention is the image into luminance Y and chroma C. 一般而言,彩度的变化并不会像亮度的变化那样的强烈,也就是说彩度的空间频率相较于亮度的空间频率低很多。 Changes in general, saturation and not as strong as changes in brightness, saturation spatial frequency that is much lower compared to the spatial frequency brightness. 本发明利用如此的影像特性,针对亮度与彩度分别应用不同阶数,提出更具有物件边缘判断的低通滤波器。 The present invention utilizes such characteristics of the image, the luminance and saturation are different application order, a low pass filter made more object edge determination. 彩度部份可为阶数较高的低通滤波器,亮度可为阶数较低的滤波器,来改善影像的亮度及彩度噪声,以提升影像视觉效果,更例如较能维持物件边缘的显示,不会被过度模糊化。 Saturation portion may be higher-order low-pass filter, the luminance can be lower order filter, to improve the image brightness and chroma noise to improve image visual effects, to maintain more than the article edge e.g. the show will not be over-blurred.

[0038] 图1示出了依据本发明实施例,影像处理的方法示意图。 [0038] FIG. 1 shows a schematic diagram of a method embodiment, the image processing of the embodiment of the present invention. 参阅图1,本发明的影像处理的方法,可以设置在一影像处理装置上。 Referring to Figure 1, the image processing method of the present invention may be provided in an image processing apparatus. 影像处理装置例如是数字照相机,影像装置或是电脑的处理系统等,藉由硬件或是软件的处理,以达到较佳品质的影像。 Image processing means such as a digital camera, an image processing apparatus or a computer system, processing by hardware or software, to achieve a better quality image. 于步骤100,例如藉由照相机的影像感应模块取得影像的原始文件数据。 In step 100, for example, acquired by the image sensor of the camera module of the original image data file. 步骤102进行对取得的影像做第一阶段的处理,例如包括白平衡、像素内插、噪声消除、伽玛(Gamma)色彩校正,尺寸缩放等。 Step 102 performs the first stage of the process to make the image acquired, for example including white balance, pixel interpolation, noise reduction, gamma (the Gamma) color correction, scaling size. 步骤104进行在YCbCr空间的Y亮度及C彩度的分离,在步骤106得到对应像素阵列的Y亮度信息,在步骤108得到C彩度信息。 In step 104 the separation space YCbCr luminance Y and chroma C, Y obtained in step 106 corresponding to the brightness information of the pixel array, at step 108 obtained chroma information C. 图2示出了Y、Cb、Cr对应像素阵列的分布。 FIG 2 shows a Y, Cb, Cr corresponding to the distribution of the pixel array. Y阵列120是売度像素值。 Y 120 array of pixel values ​​is bai. Cb阵列122是Cb像素值。 Pixel array 122 is Cb Cb values. Cr阵列124是Cr像素值。 Cr is the Cr 124 array of pixel values. 又或是Cb、Cr也可合成CbCr。 And or Cb, Cr can be synthesized CbCr. 每一种像素阵列,需要对应的滤波处理。 Each of the pixel array, corresponding to the required filtering.

[0039] 在步骤110,针对Y亮度信息至少做噪声的抑制。 [0039] In step 110, the luminance information Y do for at least suppressing noise. 又例如可以配合考虑物件边缘的影像显示,同时进行边缘的增强滤波(Edge Enhancement Filter)。 Another example can be considered with an edge of the object image display, at the same time the edge enhance filter (Edge Enhancement Filter). 于步骤112,针对CbCr色彩信息至少做噪声的抑制。 In step 112, the color information for at least do CbCr noise suppression. 还例如可以配合考虑物件边缘的影像显示,同时进行边缘的增强滤波。 Also for example, consider an object with the edge of the video display simultaneously the edge enhancement filtering. 步骤110与步骤112的较详细机制会描述于后。 Step 110 and step 112 will be more detailed mechanism is described in the following.

[0040] 当分别处理完成的Y亮度信息与CbCr色彩信息一般会例如再组合后进行压缩步骤114,压缩成一般的Jpeg/Jpg的影像文件。 [0040] Step 114 is compressed when the processing is complete, respectively, the luminance information Y and color information generally CbCr recombined e.g., compressed into a general Jpeg / Jpg image file. 压缩完成后就可以输出116。 116 may be output after the completion of the compression. 压缩步骤114仅是一般影像处理步骤。 Compression step 114 merely general image processing steps. 然而,本发明将影像已将噪声做有效滤除,以提升影像品质。 However, the present invention is effective to make the image noise has been filtered to enhance the image quality.

[0041] Y亮度信息与CbCr色彩信息的影像效果不同。 [0041] Y luminance information and color information of the image effects CbCr different. 例如在传统方法中,物件边缘的色彩信息容易被模糊掉。 For example, in the conventional method, the color information of the object is easily obscured edges. 本发明于步骤110与步骤112,提出可以简单实现,并且滤波阶数是可调的滤波器,可称为巴斯卡三角形低通滤波器。 The present invention is in the step 110 and step 112, may be made simple implementation, and the tunable filter is a filter order, Pascal triangle may be referred to as a low-pass filter.

[0042] 所谓的巴斯卡三角形,其关系如图3所示。 [0042] The so-called Pascal triangle, the relationship as shown in FIG. 巴斯卡三角形的关系是一般所知的数学常识,可以规则推算出每一阶的系数。 Pascal's triangle relationship is a mathematical knowledge generally known, can rule calculate the coefficient for each order. 本发明的一实施例所提出的低通滤波器,其滤波形式的结构,随着阶数的增加也会有如巴斯卡三角形的规则变化,因此称为巴斯卡三角形低通滤波器。 An embodiment of the present invention proposed a low-pass filter, its filter structure of the form, as the number of order of the rules will change like Pascal triangle, so called low-pass filter Pascal triangle. 也因此允许滤波阶数是可调的,其可以有初始设定值(default value),但是也可以由使用者从外部调整,供使用者的选择,以符合所要的影像效果。 Thus allowing filter order is adjustable, which can be initial set value (default value), but may be adjusted by the user from the outside, for the user's selection, the image to meet the desired effect.

[0043] 以下举一实施例描述阶数可调的巴斯卡三角形低通滤波器,但是其不是本发明的唯一选择。 [0043] For the following description of embodiments of tunable order low-pass filter a Pascal triangle embodiment, but it is not the only option for the present invention. 图4示出了依据本发明实施例,巴斯卡三角形低通滤波器在χ方向的滤波机制示意图。 FIG 4 illustrates an embodiment of the present invention, a low pass filter schematic Pascal triangle in the filtering mechanism based χ direction. 参阅图4,对于一张影像的像素阵列140以A(x,y)来表示,参数χ与y对应像素的位置。 Referring to Figure 4, an image for the pixel array 140 A (x, y) is represented, the parameter χ and y positions corresponding to the pixel. A(x,y)代表例如PxQ解析度的一像素阵列的像素值。 A (x, y) represents, for example the pixel values ​​of a pixel array of PxQ resolution.

[0044] A(x,y)的影像需要消除噪声。 [0044] A (x, y) of images necessary to eliminate noise. 因此,针对次一阶的滤波即是{11}的阶数,可以将原有的A(x, y)数据140当作第一部份数据A(x, y) 142。 Thus, for a filtering time order, i.e., {11} is the order, the original may be A (x, y) data 140 as the first partial data A (x, y) 142. 先就对χ的正方向进行噪声消除的状况来描述,然而如果对χ的负方向进行噪声消除,其机制仍相似。 First on the positive direction χ of noise cancellation to describe the situation, but if the negative direction χ noise cancellation, the mechanism is still similar. 另外,χ方向的解析度是以O到ρ-1的P个像素为例。 Further, the direction of resolution is χ ρ-1 P O to the pixels as an example. 一第二部分影像B(x+n,y) 144是取自原有的A(x,y)数据,但是属于移位η个像素后的数据,即是第η个像素到第ρ-1个像素的数据,做为第二部分影像B的第O个像素到第p-1-n个像素。 A second partial image B (x + n, y) 144 is taken from the original A (x, y) data, but the data pixel belonging shifted η, i.e., the first pixel to a first η ρ-1 pixel data, pixel O as the first B image of the second portion into the first p-1-n pixels. 而由于χ方向的解析度有P个像素,因此第二部分影像B的第ρ-η个像素到第ρ-1个像素,是一边界区域(boundary region),可以填入一设定值。 Since the resolution χ direction with P pixels, pixels of ρ-η B of the second portion to the first image ρ-1 th pixel, is a boundary region (boundary region), a predetermined value can be filled. 边界区域例如可以填入都相同大小的设定值,其更例如取A(x,y)数据的最后一个像素的像素值。 For example, the boundary region can be filled are the same size setting value, which is more a final example, take the pixel value A (x, y) data.

[0045] 接着、对于{11}阶的滤波而言,将第一部份数据A与第二部份数据B做平均得到在{11}阶的第一阶滤波影像C 146,也就是C= (A+B)/2。 [0045] Next, the order of the filter {11}, the first portion of the data A and the data B averaging the second portion obtained in step {11} to a first order filtering image C 146, i.e. C = (A + B) / 2. 接着,递回(recur)前述的方式可以得到次一阶{121}的滤波。 Next, the recursive (Recur) the manner previously described {121} can be obtained a secondary filtering step. 以第一阶滤波产生的影像C 146开始,依照相同原则再度分出第一部份数据C(x,y) 148与第二部份数据C(x+n,y) 150。 Image C 146 starts a first-order filter produced in accordance with the same principles as the first portion was separated again data C (x, y) 148 and the second portion of data C (x + n, y) 150. 接着平均后可以得到第{121}阶的影像,即是第二阶滤波影像152。 Then the average can be obtained after the first order image {121}, i.e. the second order filter 152 image. 其他阶数的滤波器可由相同方式进行递回所要的次数而得到。 Other times of the order recursive filter to be obtained may be in the same manner.

[0046] 上述是再χ方向的一维的滤波方式,相同的方式可以在y方向做滤波。 One-dimensional filtering methods [0046] The χ direction, then, in the same manner in the y-direction filtering can be done. 图5示出了依据本发明实施例,巴斯卡三角形低通滤波器在I方向的滤波机制示意图。 FIG 5 illustrates an embodiment of the present invention, a low pass filter schematic Pascal triangle in the filtering mechanism based on the I direction. 将原有的A(x,Y)数据240当作第一部份数据A(x,y)242。 The original A (x, Y) data 240 as the first partial data A (x, y) 242. 对于y方向的正方向进行噪声消除的状况来描述,然而如果对y的负方向进行噪声消除,其机制仍相似。 Carried out for the positive y-direction noise cancellation to describe the situation, but if the negative direction of the y noise cancellation, the mechanism is still similar. 另外,y方向的解析度是以O到q-Ι的q个像素为例。 Further, the resolution in the y direction is O q pixels to q-Ι example. 一第二部分影像B(x,y+m) 244是取自原有的A(x,y)数据,但是属于移位m个像素后的数据,即是第m个像素到第q-Ι个像素的数据,做为第二部分影像B的第O个像素到第q-1-m个像素。 A second partial image B (x, y + m) 244 is taken from the original A (x, y) data, but the data belonging to the shifted m pixels, that is, the m-th pixel to the first q-Ι pixel data, pixel O as the first image the second portion B of the first to q-1-m pixels. 而由于y方向的解析度有q个像素,因此第二部分影像B的第qm个像素到第q-Ι个像素,是y方向上的边界区域,如x方向的方式可以填入一设定值。 Since the resolution in the y direction are of q pixels, the first pixel qm to the second portion of the video B iota q-th pixel, is a boundary region in the y-direction, the x-direction such a manner can be filled to a set value.

[0047] 接着、对于{11}阶的滤波而言,将第一部份数据A 242与第二部份数据B 244做平均得到在{11}阶的第一阶滤波影像C 246,也就是C= (A+B)/2。 [0047] Next, the order of the filter {11}, the first part of the second partial data A 242 and data B 244 do {11} obtained average first order filter stage image C 246, i.e. C = (A + B) / 2. 类似地,藉由递回前述的方式可以得到影像(:(1,7)248、(:(1,7+111)250,进而得到次一阶{121}的滤波影像,是第二阶滤波影像252。其他阶数的滤波器可由相同方式进行递回所要的次数而得到。 Similarly, by the above-described embodiment can be handed back to the image obtained (: (1,7) 248 (1,7 :( 111 +) 250, 121 {} to give further filtering the image once the first order, second order filter 252. other filters of the order of the image by the same number of recursive manner to obtain desired.

[0048] 上述是一维的滤波方式。 [0048] The filter is a one-dimensional manner. 就滤波器而言,其例如也可简化成处理一串像素数据,例如是一像素行(pixel column)或是一像素列(pixel row)的数据。 On the filter, its simplified example, to process the series of pixel data, for example, a pixel row (pixel column) or a pixel row (pixel row) of data.

[0049] 然而由于影像是二维的像素阵列,影像一般会需要二维的滤波效应。 [0049] However, since the images are two-dimensional array of pixels, two-dimensional images typically require filtering effect. 二维的滤波效应可以根据上述的方式,例如完成χ方向或是y方向的一方向上的滤波后,再对另一方向做滤波。 After the two-dimensional filtering effect according to the above-described manner, χ direction one direction or the y-direction filtering such completion, and then do other filtering direction. 又例如,二维滤波的方法可以先在一方向完成所要阶数的滤波后,才在另一方向再进行第二方向的滤波。 As another example, two-dimensional filtering process can be completed in a desired direction, the order number of the filter, filtering it and then in the other direction is a second direction. 然而这方法仅是多种选择方式之其一。 However, this method is only one of several ways to select it. 另外例如,二维滤波的方式也可以在一个方向于每一次或是几次的递回后,就换另一个方向做滤波动作。 Further, for example, two-dimensional filtering may be in the way or several times after each recursion, do they try another direction filtering operation in one direction.

[0050] 二维滤波的是系数,以{121}阶为例,其二维系数的分布如下: [0050] The two-dimensional filter coefficients is to step {121}, for example, the distribution coefficients of two-dimensionally as follows:

[0051] 121 [0051] 121

[0052] 242 [0052] 242

[0053] 121。 [0053] 121.

[0054] 又以{1331}阶为例,二维系数的分布如下: [0054] {1331} order again as an example, the two-dimensional distribution coefficients as follows:

[0055] 1331[0056] 3993 [0055] 1331 [0056] 3993

[0057] 3993 [0057] 3993

[0058] 1331。 [0058] 1331.

[0059] 其他阶数也可以依相同方式得到。 [0059] other order can be obtained according to the same manner. 另外,X方向与y方向的滤波阶数可相同或不同。 Further, the filter order in the X direction and the y-direction may be the same or different.

[0060] 另外,藉由物件边缘(object edge)的判断机制,例如可以将此区块的影像做其他阶数的滤波。 [0060] Further, by the edge of the object (object edge) of the mechanism is determined, for example, this image block do other filtering degree. 换句话说,例如整个一张影像的不同区块,可以分别单独做适当阶数的滤波处理,无须全部都是相同阶数的滤波。 In other words, for example, different blocks of a whole image, the filtering process alone may each be appropriate degree without filter are all the same degree. 还例如,整个一张影像先做相同阶数的滤波后,再针对须要考虑的物件边缘区块,再做进一步滤波处理。 After further example, an entire image of the order of the same filter to do first, and then an edge object block for the need to consider However, further filtering process. 换句话说,依照相同的巴斯卡三角形滤波机制,其应用于各种方式的滤波安排。 In other words, following the same mechanism Pascal triangle filter, the filter arrangement applied in various ways.

[0061] 在硬件实现上,因硬件体架构及成本考虑,对于有限长度频率响应(FiniteImpulse Response,FIR)的阶数通常是受限的。 [0061] In the hardware implementation, because hardware architectures and cost considerations, the frequency response for a finite length (FiniteImpulse Response, FIR) generally of the order is limited. 本发明提出一个可无限扩增阶数的实现方法。 The present invention provides a method to achieve unlimited amplification degree. 此方法例如搭配一影像运算加速器。 This method, for example, with an image arithmetic accelerator. 此影像运算加速器可对两个影像来源A,B做算数运算,结果再存至C。 This video image arithmetic accelerator may be two sources A, B do arithmetic, and then save the results to C. 例如C = A/2+B/2。 For example, C = A / 2 + B / 2. 利用B(x,y) = A(x+n,y)或B(x,y) = A(x,y+m)来跟A(x,y)做运算,来达到可以可调整阶数的水平及垂直方向的低通率波器。 Using B (x, y) = A (x + n, y), or B (x, y) = A (x, y + m) to do with the operation A (x, y), can be adjusted to achieve the order a low-pass wave filter of the horizontal and vertical directions. 参数η与m可以相等或不相等。 Parameters η and m may be equal or unequal. 例如η = 1,2,...;m = I, 2,...ο E.g. η = 1,2, ...; m = I, 2, ... ο

[0062] 以下举一运算实例来描述如何藉有平均与递回方式得到所要的滤波阶数。 [0062] For a calculation of the following examples to describe how the filter order, by any recursive manner to obtain the average desired. A(x,y)是指在χ,y座标开始的一个影像区块,以PxP的区块为例,且取图4所述中,移位一个像素,即是η = I的情形。 A (x, y) refers to the [chi], y coordinates of a start image block to block PxP example, and taken to FIG. 4, the shift a pixel, that is, η = I situation. χ = O〜Ρ-1, y = O〜Ρ-1。 χ = O~Ρ-1, y = O~Ρ-1.

[0063] 选择Al(x,y) = A(x,y)做为第一部分数据,另外选择BI (x,y) =A(x+l,y)。 [0063] Select Al (x, y) = A (x, y) as the first partial data, choose another BI (x, y) = A (x + l, y). 接着做C= (Α1+Β1)/2 的运算。 C is then made calculation = (Α1 + Β1) / 2 in. 如此,Cl(x,y) = (A(x,y)+A(x+1,y))/2,如此可以得到{11}阶的高斯模糊滤波器。 Thus, Cl (x, y) = (A (x, y) + A (x + 1, y)) / 2, thus obtained {11} order Gaussian blur filter.

[0064]接着以 Cl (χ, y)开始,重复上述的运算,取B2(x, y) = Cl (x+1, y), A2(x, y) =CI (x,y) ο 将B2 (x, y)与A2 (x, y)作平均, [0064] Next to Cl (χ, y) starts repeating the above operations, taking B2 (x, y) = Cl (x + 1, y), A2 (x, y) = CI (x, y) ο The B2 (x, y) and A2 (x, y) are averaged,

[0065] C2 (x, y) = (A2+B2) /2 = (CI (x, y) +CI (x+1) )/2 = (A (x, y)+2*A(x+l,y)+A(x+2,y))/4,其系数是对应{121}阶的滤波器。 [0065] C2 (x, y) = (A2 + B2) / 2 = (CI (x, y) + CI (x + 1)) / 2 = (A (x, y) + 2 * A (x + l, y) + a (x + 2, y)) / 4, which is a coefficient of {121} correspond order filter.

[0066] 如此再以C2(x,y)开始,重复上述的运算可以得到{1331}阶的滤波器。 [0066] In such a re-C2 (x, y) starts to repeat the above operation can be obtained {1331} order filter. 若是再递回一次就可以得到{14641}阶的滤波器。 If once again be obtained recursively {14641} order filter. 换句话说,硬体的滤波器不必实质增加,就可以达到任一阶数的滤波器。 In other words, the filter need not be a substantial increase in hardware, you can achieve any order of a filter. 递回的次数因此可以随时做调整。 The number of recursive therefore can be adjusted at any time.

[0067] 例如,就以整体演算法流程,可以定义数学式子如下: [0067] For example, with regard to the overall process algorithms, mathematical equations may be defined as follows:

[0068] I [η]是输入影像,可以是Y亮度或彩度Cb或彩度Cr。 [0068] I [η] is an input image, or may be a Y luminance or chroma chroma Cb Cr.

[0069] I_Modl[n]是修正的输入影像。 [0069] I_Modl [n] is modified input image.

[0070] I_Mod2[η]是修正的输入影像。 [0070] I_Mod2 [η] of the input image is corrected.

[0071] I_Mod3 [η]是修正的输入影像。 [0071] I_Mod3 [η] of the input image is corrected.

[0072] LPF[I [η]]是高斯模糊低通滤波器Α,可利用阶数来调整选择不同频率。 [0072] LPF [I [η]] low-pass filter is a Gaussian blur [alpha], may be utilized to adjust the order of selecting a different frequency.

[0073] G_ForEdge [η]是高斯模糊低通滤波器B,可利用阶数来调整选择不同频率。 [0073] G_ForEdge [η] is a low-pass Gaussian blur filter B, may be utilized to adjust the order of selecting a different frequency.

[0074] EdgeMap {.}是输入的外部选择参数,以去除不必要的噪声干扰。 [0074] EdgeMap {.} Is an external input selection parameter, to remove unwanted noise.

[0075] EhnEdgeMap {.}是输入的另一外部选择参数,以去除不必要的噪声干扰。 [0075] EhnEdgeMap {.} Is another external input selection parameter, to remove unwanted noise.

[0076] G_ForEdgeEhn[η]是高斯模糊低通滤波器C,可利用阶数来调整选择不同频率。 [0076] G_ForEdgeEhn [η] low-pass filter is a Gaussian blur C, can be adjusted using the order of selecting a different frequency. [0077] De_Edge [n]是输入影像边缘检测输出,并标准化在O〜I之间。 [0077] De_Edge [n] is the input image edge detection output and normalized between O~I. 此高频滤波器可由调整选择不同频率的低通滤波器,来达到检测不同频率的边缘。 This high-frequency filter may be adjusted to select a different frequency of the low pass filter, to achieve a different frequency edge detection.

[0078] EE_Edge[n]是边缘增强的部份。 [0078] EE_Edge [n] is the edge enhancement part. 此高频滤波器可由调整不同频率的低通滤波器,来达到检测不同频率的边缘。 This high-frequency filter may be to adjust the low-pass filter with different frequencies, different frequencies to achieve the edge detection.

[0079] O [η]是输出影像,可以是Y亮度或彩度Cb或彩度Cr的分别输出。 [0079] O [η] is the output image, each output may be Y or luminance and chroma chroma Cb and Cr.

[0080] Det_Edge [η] [0080] Det_Edge [η]

[0081 ] = EdgeMap{abs(I_Modl[n]-G_ForEdge [η])}。 [0081] = EdgeMap {abs (I_Modl [n] -G_ForEdge [η])}.

[0082] EE_Edge [η] [0082] EE_Edge [η]

[0083] = EhnEdgeMap(I_Mod2[η]-G_ForEdgeEhn[η])。 [0083] = EhnEdgeMap (I_Mod2 [η] -G_ForEdgeEhn [η]).

[0084] O[η] = (Det_EdgeO_2[η]*(I_Mod3[η]+EE_Edge[η])) [0084] O [η] = (Det_EdgeO_2 [η] * (I_Mod3 [η] + EE_Edge [η]))

[0085] +((l_Det_Edge[η])*LPF[I[η])。 [0085] + ((l_Det_Edge [η]) * LPF [I [η]).

[0086] 此运算是藉由权重(weighting)的方式来调整输出。 [0086] This operation is accomplished by the weight (weighting) manner to adjust the output. Det_EdgeO_2 [η] * (1_Mod3[n] +EE_Edge[η]))是原影像加上增强的部份。 Det_EdgeO_2 [η] * (1_Mod3 [n] + EE_Edge [η])) is part of the original video plus enhanced. ((l_Det_Edge[n])*LPF[I [η])是滤掉噪声的部份。 ((L_Det_Edge [n]) * LPF [I [η]) is to filter the noise part. 利用这样的数学式子,例如可以同时达到,消除亮度及彩度不同频率的噪声,以及增强影像不同频率边缘强度。 With such a mathematical expression, for example, can be achieved at the same time, eliminate noise and luminance saturation at different frequencies, different frequencies and enhanced edge strength image. 然而如此的运算仅是本发明应用的其一。 However, such operation is only one application of the present invention.

[0087] 本发明较佳地利用不同阶数的高斯模糊滤波器,来对应到不同的频率,也就可以产生对应不同频率的低通滤波器,再由不同频率的低通滤波器,利用一些简单的转换,例如HPF{X[n]} =X[n]-LPFl {X[η]},便可以得到对应不同频率的高频滤波器。 [0087] The present invention preferably utilize different order Gaussian blur filter to correspond to a different frequency, it may be generated corresponding to different frequencies of the low-pass filter, then from the low pass filter with different frequencies, using some of the simple conversion, e.g. HPF {X [n]} = X [n] -LPFl {X [η]}, we can obtain high-frequency filters corresponding to different frequencies. 通过不同频率的调整,可以使我们噪声消除(Noise Reduction)控制上更有弹性。 By adjusting the different frequencies, so that we can eliminate noise more flexibility on (Noise Reduction) control.

[0088] 本发明提出的方法,例如可以实现(implement)在有影像处理的装置中,例如实现于数字摄像或摄影装置中,可以有效消除影像上不同频率亮度跟彩度的噪声,尤其在色彩噪声部份,除了消除噪声,也能同时也能加强边缘强度的演算法。 [0088] The method proposed by the present invention, for example, may be implemented (Implement) in the image processing apparatus has, for example, implemented in a digital camera or a photographic apparatus, can effectively eliminate the luminance image with a different frequency chroma noise, especially in color noise part, in addition to the elimination of noise, can also can enhance the strength of the edge algorithms. 另外影像处理的装置也可以例如是需要做影像处理的电脑系统。 Further the image processing apparatus may be, for example, a computer system need to do image processing.

[0089] 虽然本发明已以较佳实施例揭示如上,然其并非用以限定本发明,本领域的技术人员在不脱离本发明的精神和范围的前提下可作若干的更动与润饰,因此本发明的保护范围以本发明的权利要求为准。 [0089] While the present invention has been disclosed in the preferred embodiment described above, they are not intended to limit the invention, those skilled in the art without departing from the spirit and scope of the present invention may be made several modifications and variations, Therefore, the scope of the invention as claimed in the invention claims and their equivalents.

Claims (9)

  1. 1.一种影像噪声消除方法,包括: 接收一影像; 对该影像进行一第一阶段处理,以得到在YCbCr座标空间下分离出对应于一像素阵列的一亮度信息Y以及一色彩信息Cb与/或Cr ; 针对该亮度信息Y,进行一第二阶段处理,以至少减少一亮度噪声; 针对该色彩信息Cb与/或Cr,进行一第三阶段处理,以至少减少一色彩噪声;以及将该亮度信息Y与该色彩信息Cb与/或Cr组合, 其中该第二阶段处理与该第三阶段处理都是以藉由一低通滤波方式进行,其中该低通滤波方式包括: 将在该像素阵列的行与列的一第一方向上的每一条数据串分成一第一部分数据以及一第二部分数据,该第一部分数据是该数据串,该第二部分数据相对该第一部分数据在该第一方向上相对移位η个像素,n ≥1,其中该第二部分数据超出该第一部分数据的一边界区域的η像素被设定为一预定值; 将该第 An image noise removing method, comprising: receiving an image; an image for a first stage of the process, to give the separated at a space of YCbCr luminance information Y coordinates corresponding to a pixel array, and a color information Cb and / or of Cr; for the luminance information Y, performing a second process stage, at least a luminance noise reduction; for the color information Cb and / or Cr, performing a third stage of treatment to reduce at least a color noise; and the luminance information Y and the information Cb / Cr or the combination of colors, wherein the second process stage and the third stage of the process is to perform a low-pass filtering by mode, wherein the low-pass filtering mode comprises: in the each of the data string in the direction of a first row and column of the pixel array is divided into a first portion and a second data part of the data, the data portion of the first data string, the second portion of the data corresponding to the data in the first portion the first direction relative displacement η pixels, n ≥1, wherein the second portion η pixel data exceeds a boundary area of ​​the first portion of the data is set to a predetermined value; the first 一部分数据与该第二部分数据平均,得到一低通滤波影像;以及根据该低通滤波影像以及该低通滤波方式的前述的步骤,进行一递回处理,以达到一所要的滤波阶数。 A portion of data associated with the second portion of data is averaged to obtain a low pass filtered image; and based on the low-pass filtered image and the the low pass filtering mode step, a recursive process to achieve a number of filtering order to the.
  2. 2.如权利要求1所述的影像噪声消除方法,还包括将在该像素阵列的行与列的一第二方向上的每一条数据串分成一第一部分数据以及一第二部分数据,该第一部分数据是该数据串,该第二部分数据相对该第一部分数据在该第二方向上相对移位m个像素,m >I,其中该第二部分数据超出该第一部分数据的一边界区域的m像素被设定为一预定值; 将该第一部分数据与该第二部分数据平均,得到一低通滤波影像;以及根据该低通滤波影像以及该低通滤波方式的前述的步骤,进行一递回处理,以达到该所要的滤波阶数,且实现二维的一影像滤波处理。 2. The image noise removing method according to claim 1, further comprising each of the data strings in a second direction of the row and column of the pixel array is divided into a first portion and a second data part of the data, the second data portion of the data string, the second portion of the data corresponding to the first data portion in the second direction relative displacement m pixels, m> I, wherein the second portion of the data exceeds a boundary region of the first portion of the data m pixel is set to a predetermined value; the first partial data and the second partial data averaged to obtain a low pass filtered image; and an image based on the low-pass filtering and low pass filtering the aforementioned embodiment step, a recursive processing, the filter order to achieve the desired, and achieve a two-dimensional image filtering.
  3. 3.如权利要求2所述的影像噪声消除方法,其中η = 1,m = I。 3. The image noise elimination according to claim 2, wherein η = 1, m = I.
  4. 4.如权利要求2所述的影像噪声消除方法,其中n = m。 4. The image noise elimination according to claim 2, wherein n = m.
  5. 5.如权利要求1所述的影像噪声消除方法,其中η = I。 5. The image noise cancellation method of claim 1, wherein η = I.
  6. 6.如权利要求1所述的影像噪声消除方法,其中该所要的滤波阶数可由外部选择。 Image noise as claimed in claim 1 wherein the filter order may be outside the desired selection request cancellation method.
  7. 7.如权利要求1所述的影像噪声消除方法,还包括一边缘判断机制决定一边界区域,并且针对该边界区域调整该所要的滤波阶数。 7. The image noise removing method according to claim 1, further comprising a determination mechanism for determining a boundary of an edge region, and adjusting the filtering order of the desired number for the boundary region.
  8. 8.如权利要求7所述的影像噪声消除方法,其中在边界区域的该所要的滤波阶数可由外部调整。 8. The image noise elimination according to claim 7, wherein the external adjustment by the filter order in the border area desired.
  9. 9.如权利要求1所述的影像噪声消除方法,其中经过该低通滤波方式处理后的一低通滤波影像,还包括将该低通滤波影像与该影像的一高通滤波影像再做一权重处理。 9. The image noise cancellation method of claim 1, wherein after a low-pass filtering the image after low-pass filtering process embodiment, further comprising low pass filtering the image with a high pass filter do images of the image a weight deal with.
CN 200810174849 2008-11-07 2008-11-07 Method for eliminating image noise CN101742086B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 200810174849 CN101742086B (en) 2008-11-07 2008-11-07 Method for eliminating image noise

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 200810174849 CN101742086B (en) 2008-11-07 2008-11-07 Method for eliminating image noise

Publications (2)

Publication Number Publication Date
CN101742086A true CN101742086A (en) 2010-06-16
CN101742086B true CN101742086B (en) 2013-05-15

Family

ID=42464919

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 200810174849 CN101742086B (en) 2008-11-07 2008-11-07 Method for eliminating image noise

Country Status (1)

Country Link
CN (1) CN101742086B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1183696A (en) 1996-09-30 1998-06-03 三星电子株式会社 Image quality enhancement circuit and method therefor
CN1335579A (en) 2000-07-27 2002-02-13 诺日士钢机株式会社 Image processing method, apparatus and recording medium for recording & executing the same method program
US20070097391A1 (en) 2005-11-01 2007-05-03 Hwang Sung-Hyun Apparatus and method for improving image quality of image sensor
CN101009771A (en) 2006-01-23 2007-08-01 株式会社理光 Image processing apparatus, imaging apparatus, image processing method, and computer program product
CN101119448A (en) 2006-07-31 2008-02-06 株式会社理光 Image processing apparatus, imaging apparatus, image processing method, and computer program product

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1183696A (en) 1996-09-30 1998-06-03 三星电子株式会社 Image quality enhancement circuit and method therefor
CN1335579A (en) 2000-07-27 2002-02-13 诺日士钢机株式会社 Image processing method, apparatus and recording medium for recording & executing the same method program
US20070097391A1 (en) 2005-11-01 2007-05-03 Hwang Sung-Hyun Apparatus and method for improving image quality of image sensor
CN101009771A (en) 2006-01-23 2007-08-01 株式会社理光 Image processing apparatus, imaging apparatus, image processing method, and computer program product
CN101119448A (en) 2006-07-31 2008-02-06 株式会社理光 Image processing apparatus, imaging apparatus, image processing method, and computer program product

Also Published As

Publication number Publication date Type
CN101742086A (en) 2010-06-16 application

Similar Documents

Publication Publication Date Title
US20080175510A1 (en) Imaging apparatus, noise removing device, noise removing method, program for noise removing method, and recording medium for recording the same
US20120020556A1 (en) Image processor, image processing method, and computer-readable medium
US7065246B2 (en) Image processing apparatus
US20070229863A1 (en) Tone Correction Apparatus, Mobile Terminal, Image Capturing Apparatus, Mobile Phone, Tone Correction Method and Program
US20090160992A1 (en) Image pickup apparatus, color noise reduction method, and color noise reduction program
WO1994018801A1 (en) Color wide dynamic range camera using a charge coupled device with mosaic filter
JP2005520442A (en) Method and apparatus for processing a sensor image
US7417670B1 (en) Digital video camera with binning or skipping correction
US20050025379A1 (en) Digital imaging systems, articles of manufacture, and digital image processing methods
US20080240602A1 (en) Edge mapping incorporating panchromatic pixels
US20090219417A1 (en) Image capturing system and computer readable recording medium for recording image processing program
JPH08163408A (en) Noise detection circuit, noise elimination circuit and control emphasis circuit
US7609300B2 (en) Method and system of eliminating color noises caused by an interpolation
JP2005269645A (en) High-quality gradient correction linear interpolation for demosaic processing of color image
US20080298722A1 (en) Apparatus for Color Interpolation Using Adjustable Threshold
US20050025380A1 (en) Digital image processing methods, digital image devices, and articles of manufacture
JP2007208399A (en) Image processing unit, image processing method, program thereof, and recording medium for recording the same
US20050244052A1 (en) Edge-sensitive denoising and color interpolation of digital images
US20110050934A1 (en) Image processing apparatus, image processing method, and program
JP2011003048A (en) Image processing apparatus and image processing program
US20080240601A1 (en) Edge mapping using panchromatic pixels
US7751642B1 (en) Methods and devices for image processing, image capturing and image downscaling
JP2008147980A (en) Image processor, image sensing device, method for processing image, program and recording medium
JP2008147981A (en) Image processing apparatus, imaging apparatus, image processing method, program, and storage medium
JP2010178302A (en) Image processing apparatus and image processing method

Legal Events

Date Code Title Description
C06 Publication
C10 Entry into substantive examination
C14 Grant of patent or utility model
EXPY Termination of patent right or utility model