CN105976331A - Method for adjusting white balance of image and system - Google Patents

Abstract

The invention discloses a method for adjusting the white balance of an image and a system for adjusting the white balance of the image. The method includes the following steps that: white balance characteristic data of a preview image are extracted from the preview image and are denoted as F(x); the Euclidean distances from each sample to the F(x) are calculated based on the F(x) and a preset sample set A, and a minimum Euclidean distance dmi is found out; the minimum Euclidean distance dmi is compared with a similarity threshold value T; when the dmi is smaller than T, all the samples in the sample set are sequenced according to a small-to-large sequence of the Euclidean distances; the number of the N-th samples in each sample group is calculated based on the N-th samples which are selected by a user and are located ahead after the samples are sequenced, wherein N is an odd number; the image corresponding to the F(x) is distributed to a sample group including the most samples; and interpolation is carried out on the sample group including the most samples based on the F(x), so that a white balance adjustment parameter can be obtained. With the method and system of the invention adopted, supplementation can be carried out on a scene with which the user is unsatisfied, and the white balance parameter of the image can be adjusted.

Description

A kind of method and system regulating image white balance

Technical field

The invention belongs to technical field of image processing, particularly to a kind of method regulating image white balance and System.

Background technology

White refers to that the light reflexing in human eye is owing to three kinds of coloured light ratios blue, green, red are identical and have The vision response that certain brightness is formed.White light is by red, orange, yellow, green, cyan, blue, purple seven Plant coloured light composition, and these seven kinds of coloured light be to be mixed to form by different proportion by Red Green Blue, Popular understanding, white is free from the brightness of colored composition.

White balance is a very abstract concept, the most popular understanding allow exactly white imaging be still White, if Bai Shibai, the image of other scenery can be closer to the color vision custom of human eye.Bai Ping Weighing apparatus adjusts typically three kinds of modes on early stage equipment: preset white balance, manual white balance adjust and automatic Follow the tracks of blank level adjustment.

Under the light source of different-colour, human eye is different to the understanding of white, and usual colour temperature is the highest, light Color is the most blue, and colour temperature is the lowest the reddest.Therefore, under the light source of different-colour, need to pass through white balance Regulation algorithm, uses different red green, blue green ratio also colour prime white, to meet people as far as possible to image Visual custom.In theory, white balance algorithm perfect can not process the light source scene of 100%.

Colour temperature, it is simply that represent color with kelvin degree (K) quantitatively.The light that tengsten lamp is sent is Yellow hue, reason is that colour temperature is relatively low；The flame of natural gas is blue, and reason is that colour temperature is higher.A certain When kind coloured light is higher than the colour temperature of other coloured light, illustrate that this coloured light is more blue than other coloured light, on the contrary the reddest； Equally, illustrate that the colour temperature of this coloured light is higher when a kind of coloured light is partially more blue than other coloured light, on the contrary on the low side.

As a example by the mobile phone camera of high-pass platform, platform provides can be to different light sources (8 kinds), difference Brightness (3 kinds) uses different red green, blue green ratio to be adjusted respectively, enters mixing light source The method regulated again after row interpolation.Due to this control method scene segmentation deficiency to mixing light source, combination Limited amount, occurs that same group of parameter cannot meet two or more similar scene simultaneously, causes same The problem occurring under mixing light source, the similar lightness environment of sample can not distinguish.

The present invention uses the classificating thought of KNN, unsatisfactory after regulating for current white balance system Scene is supplemented.

Summary of the invention

It is an object of the invention to provide a kind of method and system regulating image white balance, it is possible to for user Unsatisfied scene is supplemented, the white balance parameter of regulation image.

For achieving the above object, one aspect of the present invention provides a kind of method regulating image white balance, Described method includes: extracts the white balance characteristic of this two field picture from a two field picture of preview, is designated as F(x)；Based on white balance characteristic F (x) and default sample set A, calculate each sample A₁₁,…,A_miEuclidean distance d to white balance characteristic F (x)₁₁,…,d_mi, find minimum euclidean distance d_mi；Wherein, sample set A includes multiple sample group A₁,…,A_n, each sample group A_nIncluding multiple samples This A₁₁,…,A_ni, i >=1, m >=n >=1；；Judge minimum range d_miBig with similarity threshold T Little；Work as d_miDuring ＜ T, according to Euclidean distance d₁₁,…,d_miOrder from small to large, to described sample set Close all sample A in A₁₁,…,A_miIt is ranked up；Based on odd number forward after the sequence that user chooses Sample, adds up the sample size comprising described odd number sample in each sample group；Described white balance is special Levy this image division corresponding to data F (x) to comprising in the sample group that described sample size is most；To put down in vain Weighing apparatus characteristic F (x) makees interpolation with the described sample group most containing sample size, obtains white balance adjusting Parameter.

Wherein, d is worked as_miDuring >=T, using described white balance characteristic as white balance adjusting parameter.

Wherein, described method also includes: previously generate the step of described sample set, comprising: from many Two field picture extracts the white balance characteristic of every two field picture；Record the plurality of white balance characteristic； The plurality of white balance characteristic is normalized, generates described sample set A.

Wherein, under described white balance characteristic includes each shelves warm colour temperature ash point quantity number, intensity of illumination, R/G ratio and B/G ratio.

Wherein, described based on the forward odd number sample after the sequence chosen, add up in each sample group Comprise the step of the sample size of sample in described odd number sample to include: from all samples after sequence, Forward odd number sample is selected to obtain sample set C；Based on the sample belonging to sample each in sample set C This group, adds up and comprises the sample size of sample in described sample set C in each sample group.

Wherein, white balance adjusting parameter R/G and B/G are obtained based on following formula as interpolation；

R/G=Crg* (1-Abs (Frg-Crg)/Max (Frg, Crg))；

B/G=Cbg* (1-Abs (Fbg-Cbg)/Max (Fbg, Cbg))；

Wherein, Frg represents that the B/G that the R/G of F (x), Fbg represent F (x), Crg represent and comprises sample number The R/G, Cbg that measure most sample groups represent that the B/G, Abs that comprise the most sample group of sample size are Taking absolute value, Max is for taking maximum.

Wherein, described distance calculation module calculates each sample A based on following formula₁₁,…,A_miArrive The Euclidean distance d of F (x)₁₁,…,d_mi:

$d_{mi}=\sqrt{\underset{k=1}{\overset{N}{\Σ}}{(F{\left(x\right)}_k-A_{{\mathrm{mi}}_k})}^2}$

Wherein, F (x)_kRepresent the kth dimensional feature numerical value of F (x),Represent A_miKth dimensional feature numerical value, N represents dimension.

Another aspect of the present invention provides a kind of system regulating image white balance, and described system includes: Data extraction module, distance calculation module, judge module, order module, statistical module, sort module, Interpolating module and parameter adjustment module；Data extraction module, should for extracting from a two field picture of preview The white balance characteristic of two field picture, is designated as F (x)；Distance calculation module, based on white balance characteristic F (x) and the sample set A preset, calculates each sample A₁₁,…,A_miTo white balance characteristic F (x) Euclidean distance d₁₁,…,d_mi, find minimum euclidean distance d_mi；Wherein, sample set A includes multiple sample This group A₁,…,A_n, each sample group A_nIncluding multiple sample A₁₁,…,A_ni, i >=1, m >=n >=1；； Judge module, is used for judging minimum range d_miSize with similarity threshold T；Order module, is used for Work as d_miDuring ＜ T, according to Euclidean distance d₁₁,…,d_miOrder from small to large, in described sample set A All sample A₁₁,…,A_miIt is ranked up；Statistical module, leans on after the sequence chosen based on user Front odd number sample, adds up the sample size comprising described odd number sample in each sample group；Classification Module, by this corresponding for described white balance characteristic F (x) image division to comprising described sample size In many sample groups；Interpolating module, white balance characteristic F (x) is most containing sample size with described Sample group make interpolation, obtain white balance adjusting parameter；Parameter adjustment module, for based on described white flat Image is adjusted by weighing apparatus regulation parameter.

Wherein, described parameter adjustment module performs following operation: work as d_miDuring >=T, described white balance is special Levy data F (x) as white balance adjusting parameter.

Wherein, described system also includes: sample set generation module, is used for previously generating described sample set Closing, it performs following operation: extract white balance characteristic F (x) of every two field picture from multiple image； Record the plurality of white balance characteristic F (x)；The plurality of white balance characteristic F (x) is returned One change processes, and generates described sample set A.

Wherein, ash point quantity number, the illumination under described white balance characteristic F (x) includes each shelves warm colour temperature Intensity, R/G ratio and B/G ratio.

Wherein, described statistical module performs following operation: from all samples after sequence, select forward Odd number sample obtain sample set C；Based on the sample group belonging to sample each in sample set C, system Count and each sample group comprises the sample size of sample in described sample set C.

Wherein, described interpolating module (70) based on following formula as interpolation obtain white balance adjusting parameter R/G and B/G；

R/G=Crg* (1-Abs (Frg-Crg)/Max (Frg, Crg))；

B/G=Cbg* (1-Abs (Fbg-Cbg)/Max (Fbg, Cbg))；

Wherein, Frg represents that the B/G that the R/G of F (x), Fbg represent F (x), Crg represent and comprises sample number The R/G, Cbg that measure most sample groups represent that the B/G, Abs that comprise the most sample group of sample size are Taking absolute value, Max is for taking maximum.

Wherein, described distance calculation module (20) calculates each sample based on following formula A₁₁,…,A_miEuclidean distance d to white balance characteristic F (x)₁₁,…,d_mi:

$d_{mi}=\sqrt{\underset{k=1}{\overset{N}{\Σ}}{(F{\left(x\right)}_k-A_{{\mathrm{mi}}_k})}^2}$

Wherein, F (x)_kRepresent the kth dimensional feature numerical value of white balance characteristic F (x),Represent A_mi's Kth dimensional feature numerical value, N represents dimension.

As it has been described above, by the method and system of the regulation image white balance of the present invention, it is possible to for user Unsatisfied scene is supplemented, the white balance parameter of regulation image.

Accompanying drawing explanation

Fig. 1 is the schematic flow sheet of the method for the regulation image white balance of the present invention；

Fig. 2 is the schematic flow sheet of step S20 of the present invention；

Fig. 3 is the schematic flow sheet of step S5 of the present invention；

Fig. 4 is the structural representation of the system of the regulation image white balance of the present invention.

Detailed description of the invention

For making the object, technical solutions and advantages of the present invention of greater clarity, below in conjunction with being embodied as Mode referring to the drawings, the present invention is described in more detail.It should be understood that these describe simply example Property, and it is not intended to limit the scope of the present invention.Additionally, in the following description, eliminate known knot Structure and the description of technology, to avoid unnecessarily obscuring idea of the invention.

It should be understood that to obtain white balance the most accurately, can choose a digital camera on manually white Balance mode, allows camera according to object of reference colour temperature under light source, determines how accurate reproduction.Manually White balance function, is substantially all the object using a color neutral and makees benchmark, such as standard gray, standard white The object of reference of color.As long as it is, in principle, that object of reference is without any colour cast, it is possible to allow camera obtain Good benchmark data, it is thus achieved that optimal white balance reduction.

Fig. 1 is the schematic flow sheet of the method for the regulation image white balance of the present invention.

As it is shown in figure 1, the schematic flow sheet of method of the regulation image white balance of the present invention, it include as Lower step:

Step S1, extracts the white balance characteristic of this two field picture from a two field picture of preview, is designated as F(x)。

In this step, camera from preview to multiple image random extract a two field picture, from this two field picture Middle extraction white balance characteristic, is designated as F (x), and described white balance characteristic F (x) includes each shelves warm colour Ash point quantity number, intensity of illumination, R/G ratio and B/G ratio under Wen.Concrete, described white balance feature Data include: the ash point quantity under each shelves colour temperature (is such as, one grade, 1000-2000 with 0-1000K Being one grade, 2000-3000 is one grade etc.), intensity of illumination, R/G and B/G value.

Step S2, based on white balance characteristic F (x) and default sample set A, calculates each sample A₁₁,…,A_miEuclidean distance d to white balance characteristic F (x)₁₁,…,d_mi, find minimum euclidean distance d_mi, wherein, sample set A includes multiple sample group A₁,…,A_n, each sample group A_nIncluding multiple samples This A₁₁,…,A_ni, i >=1, m >=n >=1；.

In this step, based on white balance characteristic F (x) extracted in step S1 and the sample preset Set A, calculates each sample A₁₁,…,A_miEuclidean distance to white balance characteristic F (x) d₁₁,…,d_mi, from the plurality of Euclidean distance d₁₁,…,d_miIn find minimum euclidean distance d_mi。

Concrete, calculate each sample A based on following formula₁₁,…,A_miEurope to white balance characteristic F (x) Formula distance d₁₁,…,d_mi:

$d_{mi}=\sqrt{\underset{k=1}{\overset{N}{\Σ}}{(F{\left(x\right)}_k-A_{{\mathrm{mi}}_k})}^2}$

Wherein, F (x)_kRepresent the kth dimensional feature numerical value of F (x),Represent A_miKth dimensional feature numerical value, N represents dimension.

Step S3, it is judged that minimum range d_miSize with similarity threshold T.

In this step, relatively described minimum range d_miWith the size of similarity threshold T, so that it is determined that be No needs carries out classified statistics to sample set A.

Step S4, works as d_miDuring ＜ T, according to Euclidean distance d₁₁,…,d_niOrder from small to large, to institute State all sample A in sample set A₁₁,…,A_miIt is ranked up.

In this step, work as d_miDuring ＜ T, according to Euclidean distance d₁₁,…,d_miOrder from small to large is right All sample A in described sample set A₁₁,…,A_miSort successively.

Step S5, based on odd number sample forward after the sequence that user chooses, adds up in each sample group Comprise the sample size of described odd number sample.

In this step, from all samples after sequence, select forward odd number sample (this odd number The value of the Euclidean distance of sample is less than normal), comprise institute based in this odd number sample statistics each sample group State the sample size of sample in odd number sample.

Step S6, by this corresponding for described F (x) image division to comprising the sample that described sample size is most In group.

In this step, by samples most for this image division to the model quantity comprising sample in odd number sample In this group.

Step S7, makees interpolation by F (x) with the described sample group most containing sample size, obtains white balance Regulation parameter.

In this step, sample groups most with the model quantity comprising sample in odd number sample for F (x) is entered Row makees interpolation, obtains final white balance adjusting parameter.

Further, described method also includes: step S8, works as d_miDuring >=T, by described white balance feature Data are as white balance adjusting parameter.Concrete, work as d_miDuring >=T, directly F (x) is adjusted as white balance JIESHEN number.

Wherein, white balance adjusting parameter R/G and B/G are obtained based on following formula as interpolation；

R/G=Crg* (1-Abs (Frg-Crg)/Max (Frg, Crg))；

B/G=Cbg* (1-Abs (Fbg-Cbg)/Max (Fbg, Cbg))；

Wherein, Frg represents that the B/G that the R/G of F (x), Fbg represent F (x), Crg represent and comprises sample number The R/G, Cbg that measure most sample groups represent that the B/G, Abs that comprise the most sample group of sample size are Taking absolute value, Max is for taking maximum.

Fig. 2 is the schematic flow sheet of step S20 of the present invention.

As in figure 2 it is shown, described method also includes: step S20, previously generate the step of described sample set Suddenly, step S20 farther includes:

Step S21, extracts the white balance characteristic of every two field picture from multiple image.

In this step, from multiple image, extract the white balance characteristic of every two field picture, every two field picture pair Answer a white balance characteristic, as it was previously stated, described white balance characteristic includes: under each shelves colour temperature Ash point quantity (such as, with 0-1000K be one grade, 1000-2000 be one grade, 2000-3000 is one Shelves etc.), intensity of illumination, R/G and B/G value.

Step S22, records the plurality of white balance characteristic.

In this step, record the plurality of white balance characteristic.

Step S23, is normalized the plurality of white balance characteristic, generates described sample Set A.

In this step, the plurality of white balance characteristic is normalized, generates described sample Set A.

In the present invention, the purpose of normalized is to described ash point quantity, intensity of illumination, R/G and B/G Value converts so that ash point quantity, intensity of illumination, R/G and B/G value after conversion are between 0-1.

Fig. 3 is the schematic flow sheet of step S5 of the present invention.

As it is shown on figure 3, described step S5 farther includes:

Step S51, from all samples after sequence, selects forward odd number sample to obtain sample set Close C.

In this step, further from all samples after sequence, forward odd number sample is selected to form Sample set C.

Step S52, based on the sample group belonging to sample each in sample set C, in the different sample group of statistics Comprise the sample size of sample in described sample set C.

In this step, according to the sample group belonging to sample each in sample set C, the different sample group of statistics Comprise the sample size of sample in described sample set C.

For example, sample set A includes 3 sample groups A₁,A₂, each sample group A₁Including multiple samples This A₁₁,A₁₂,A₁₃,A₂Including multiple sample A₂₁,A₂₂,A₂₃；A₁₁,A₁₂,A₁₃,A₂₁,A₂₂,A₂₃Respectively Corresponding Euclidean distance is 1,2,4,3,7,6.According to Euclidean distance order from small to large, to sample This A₁₁,A₁₂,A₁₃,A₂₁,A₂₂,A₂₃Being ranked up, ranking results is as follows: A₁₁,A₁₂,A₂₁,A₁₃,A₂₃,A₂₂, it is assumed that user chooses 3 sample A that sequence is forward₁₁,A₁₂,A₂₁, Add up this 3 sample A₁₁,A₁₂,A₂₁Affiliated sample group, sample A₁₁,A₁₂Belong to sample group A₁, sample This A₂₁Belong to sample group A₂, it can be seen that in 3 samples that sequence is forward, sample group A₁Middle sample This quantity is 2, sample group A₂The quantity of middle sample is 1, therefore F (x) is divided into sample group A₁In.

Fig. 4 is the structural representation of the system of the regulation image white balance of the present invention.

As shown in Figure 4, the system of the regulation image white balance of the present invention, described system includes: data carry Delivery block 10, distance calculation module 20, judge module 30, order module 40, statistical module 50, point Generic module 60, interpolating module 70 and parameter adjustment module 80.

Data extraction module 10, for extracting the white balance feature of this two field picture from a two field picture of preview Data, are designated as F (x), and wherein, described white balance characteristic includes the ash point quantity under each shelves warm colour temperature Number, intensity of illumination, R/G ratio and B/G ratio.

Distance calculation module 20 is connected with described data extraction module 10, based on F (x) and default sample Set A, calculates each sample A₁₁,…,A_miEuclidean distance d to F (x)₁₁,…,d_mi, find minimum Europe Formula distance d_mi；

Wherein, sample set A includes multiple sample group A₁,…,A_n, each sample group A_nIncluding multiple samples This A₁₁,…,A_ni, i >=1, m >=n >=1；；

Wherein, described distance calculation module 20 calculates each sample A based on following formula₁₁,…,A_miTo F's (x) Euclidean distance d₁₁,…,d_mi:

$d_{mi}=\sqrt{\underset{k=1}{\overset{N}{\Σ}}{(F{\left(x\right)}_k-A_{{\mathrm{mi}}_k})}^2}$

Wherein, F (x)_kRepresent the kth dimensional feature numerical value of F (x),Represent A_miKth dimensional feature numerical value, N represents dimension.

Judge module 30 and described distance calculation module 20, be used for judging minimum range d_miWith similarity threshold The size of value T；

Order module 40 is connected with described judge module 30, for working as d_miDuring ＜ T, according to Euclidean distance d₁₁,…,d_miOrder from small to large, to all sample A in described sample set A₁₁,…,A_miCarry out Sequence；

Statistical module 50 and described order module 40, forward strange after the sequence chosen based on user Several samples, add up the sample size comprising described odd number sample in each sample group；

Sort module 60 is connected with described data extraction module 10 and described statistical module 50 respectively, by institute State this corresponding for F (x) image division to comprising in the sample group that described sample size is most；

Interpolating module 70 is connected with described sort module 60, and F (x) is most containing sample size with described Sample group make interpolation, obtain white balance adjusting parameter；

Wherein, described interpolating module 70 obtains white balance adjusting parameter R/G and B/G based on following formula as interpolation；

R/G=Crg* (1-Abs (Frg-Crg)/Max (Frg, Crg))；

B/G=Cbg* (1-Abs (Fbg-Cbg)/Max (Fbg, Cbg))；

Wherein, Frg represents that the B/G that the R/G of F (x), Fbg represent F (x), Crg represent and comprises sample number The R/G, Cbg that measure most sample groups represent that the B/G, Abs that comprise the most sample group of sample size are Taking absolute value, Max is for taking maximum.

Parameter adjustment module 80 is connected with described interpolating module 70, for joining based on described white balance adjusting Several image is adjusted.

Wherein, described parameter adjustment module 80 performs following operation: work as d_miDuring >=T, by described white flat Weighing apparatus characteristic is as white balance adjusting parameter.

In one embodiment, described system also includes: sample set generation module 90, for pre-Mr. Becoming described sample set, it performs following operation: the white balance extracting every two field picture from multiple image is special Levy data；Record the plurality of white balance characteristic；The plurality of white balance characteristic is returned One change processes, and generates described sample set A.

In one embodiment, described statistical module 50 performs following operation: all samples after sequence In, select forward odd number sample to obtain sample set C；Based on belonging to sample each in sample set C Sample group, add up and each sample group comprise the sample size of sample in described sample set C.

As it has been described above, describe the method and system of the regulation image white balance of the present invention, the present invention in detail Supplement for the unsatisfied scene of user, the white balance parameter of regulation image.

It should be appreciated that the above-mentioned detailed description of the invention of the present invention is used only for exemplary illustration or explanation The principle of the present invention, and be not construed as limiting the invention.Therefore, without departing from the present invention spirit and Any modification, equivalent substitution and improvement etc. done in the case of scope, should be included in the guarantor of the present invention Within the scope of protecting.Additionally, claims of the present invention be intended to fall into scope and Whole in the equivalents on border or this scope and border change and modifications example.

Claims (14)

1. the method regulating image white balance, it is characterised in that described method includes:

The white balance characteristic of this two field picture is extracted from a two field picture of preview；

With default sample set, calculate each sample in the sample set preset special to described white balance Levy data Euclidean distance, wherein, sample set includes that multiple sample group, each sample group include Multiple samples；

Judge whether described minimum euclidean distance is less than and similarity threshold T；

If so, according to the size order from small to large of Euclidean distance, to the institute in described sample set Sample is had to be ranked up；

The multiple samples chosen based on user, add up the plurality of sample sample number in different sample groups Amount；

By described image division corresponding for described white balance characteristic to comprising described sample size In many sample groups；

White balance characteristic is made interpolation with the described sample group comprising sample size most, obtains white Balance adjustment parameter.

Method the most according to claim 1, wherein, when described minimum euclidean distance is more than or equal to Described similarity threshold T >=time, using described white balance characteristic as white balance adjusting parameter.

Method the most according to claim 1, wherein, described method also includes: previously generate institute State the step of sample set, comprising:

The white balance characteristic of every two field picture is extracted from multiple image；

Record the plurality of white balance characteristic；

The plurality of white balance characteristic is normalized, generates described sample set.

4. according to the method described in any one of claim 1-3, wherein, described white balance characteristic Including ash point quantity number, intensity of illumination, R/G ratio and B/G ratio under each shelves colour temperature.

5. according to the method described in any one of claim 1-3, wherein, described according to Euclidean distance Size, the step being ranked up all samples in described sample set includes:

According to Euclidean distance order from small to large, all samples in sample set are ranked up.

6. according to the method described in any one of claim 1-3, wherein, described choose based on user Multiple samples, add up the plurality of sample and include in the step of the sample size of different sample groups:

From all samples after sequence, choose forward odd number sample；

Based on the sample group belonging to each sample in described odd number sample, add up described odd number sample Sample size in different sample groups.

7., according to the method described in any one of claim 1-3, wherein, obtain as interpolation based on following formula White balance adjusting parameter R/G and B/G；

R/G=Crg* (1-Abs (Frg-Crg)/Max (Frg, Crg))；

B/G=Cbg* (1-Abs (Fbg-Cbg)/Max (Fbg, Cbg))；

Wherein, Frg represents that the R/G of white balance characteristic, Fbg represent white balance characteristic B/G, Crg represent that the R/G, Cbg that comprise the most sample group of sample size represent and comprise sample size The B/G, Abs of most sample groups are for taking absolute value, and Max is for taking maximum.

8. the regulation system regulating image white balance, it is characterised in that described system includes:

Data extraction module (10), for extracting the white flat of this two field picture from a two field picture of preview Weighing apparatus characteristic, is designated as F (x)；

Distance calculation module (20), based on white balance characteristic F (x) and default sample set A, Calculate each sample A₁₁,…,A_miEuclidean distance d to white balance characteristic F (x)₁₁,…,d_mi, Find minimum euclidean distance d_mi；

Wherein, sample set A includes multiple sample group A₁,…,A_n, each sample group A_nIncluding multiple Sample A₁₁,…,A_ni, i >=1, m >=n >=1；

Judge module (30), is used for judging minimum range d_miSize with similarity threshold T；

Order module (40), for working as d_miDuring ＜ T, according to Euclidean distance d₁₁,…,d_miFrom little to Big order, to all sample A in described sample set A₁₁,…,A_miIt is ranked up；

Statistical module (50), odd number sample forward after the sequence chosen based on user, system Count the sample size comprising described odd number sample in each sample group；

Sort module (60), by this corresponding for described white balance characteristic F (x) image division to bag Containing in the sample group that described sample size is most；

Interpolating module (70), white balance characteristic F (x) is most containing sample size with described Sample group makees interpolation, obtains white balance adjusting parameter；

Parameter adjustment module (80), for being adjusted image based on described white balance adjusting parameter.

System the most according to claim 8, wherein, described parameter adjustment module (80) performs Below operation:

Work as d_miDuring >=T, described white balance characteristic F (x) is as white balance adjusting parameter.

System the most according to claim 8, wherein, described system also includes:

Sample set generation module (90), is used for previously generating described sample set, below its execution Operation:

White balance characteristic F (x) of every two field picture is extracted from multiple image；

Record the plurality of white balance characteristic F (x)；

The plurality of white balance characteristic F (x) is normalized, generates described sample set A。

11. systems described in-10 any one according to Claim 8, wherein, described white balance feature Data F (x) include ash point quantity number, intensity of illumination, R/G ratio and B/G ratio under each shelves warm colour temperature.

12. systems described in-10 any one according to Claim 8, wherein, described statistical module (50) Operation below performing:

From all samples after sequence, forward odd number sample is selected to obtain sample set C；

Based on the sample group belonging to sample each in sample set C, add up in each sample group and comprise institute State the sample size of sample in sample set C.

13. systems described in-10 any one according to Claim 8, wherein, described interpolating module (70) White balance adjusting parameter R/G and B/G is obtained as interpolation based on following formula；

R/G=Crg* (1-Abs (Frg-Crg)/Max (Frg, Crg))；

B/G=Cbg* (1-Abs (Fbg-Cbg)/Max (Fbg, Cbg))；

Wherein, Frg represents that the B/G that the R/G of F (x), Fbg represent F (x), Crg represent and comprises sample The R/G, Cbg of the sample group that quantity is most represents the B/G, Abs comprising the most sample group of sample size For taking absolute value, Max is for taking maximum.

14. systems described in-10 any one according to Claim 8, wherein, described distance calculates mould Block (20) calculates each sample A based on following formula₁₁,…,A_miEuropean to white balance characteristic F (x) Distance d₁₁,…,d_mi:

$d_{mi}=\sqrt{\underset{k=1}{\overset{N}{\Σ}}{(F{\left(x\right)}_k-A_{{\mathrm{mi}}_k})}^2}$

Wherein, F (x)_kRepresent the kth dimensional feature numerical value of white balance characteristic F (x),Represent A_mi Kth dimensional feature numerical value, N represents dimension.