CN102496144A - NSCT (nonsubsampled contourlet transform) sonar image enhancement method based on HSV (hue, saturation and value) color space - Google Patents

Abstract

The invention provides a NSCT (nonsubsampled contourlet transform) sonar image enhancement method based on HSV (hue, saturation and value) color space, which includes steps of converting sonar images in a RGB (red, green and blue) space to the HSV space, splitting HSV images into the hue H, saturation S and value V, performing median filtering on the hue H and the saturation S and performing NSCT enhancing on the value V to obtain enhanced H', S' and V', compositing the H', S' and V' into a new HSV' image, performing median filtering on the new HSV' image, and inversely converting the HSV' into the RGB space to obtain the enhanced image. Using the NSCT sonar image enhancement method based on the HSV color space can evidently improve amount of information and definition of original sonar images, border and detail information can be obtained clearly after enhancement, and the method is greatly helpful to further analysis and the like.

Description

NSCT underwater sound image enchancing method based on the HSV color space

Technical field

What the present invention relates to is the image enchancing method in a kind of digital image processing techniques field.

Background technology

Along with the development of sonograms technology, sonar image is increasingly extensive in the application in ocean development field, and the Target Recognition of sonar image becomes an important topic in digital image processing field.Though sonar image is the same with optical imagery; All be the plane or the spatial distribution map of energy in itself; But because the seawater medium of acoustic intelligence transmission channel (underwater acoustic channel) and border (seabed, sea) thereof have complicated and changeable characteristic; And the transmissison characteristic of sound wave itself, in the process of obtaining and transmitting, all receive noise pollution in various degree.The noise of image is very big for influences such as further analysis, compressions, and therefore being necessary to carry out enhancement process obtains high s/n ratio and details distinct image.The characteristic of sonar image is different with optical imagery aspect a lot; Yet; Compare with the maturation that becomes of optical image treatment technology, at present people are to the understanding of sonar image characteristic and also have very big development space to the field of sonar image processing specially.

Non-down sampling contourlet transform (NSCT) is a kind of directivity multi-scale image analytical approach effectively, i.e. Contourlet conversion.The Contourlet conversion uses laplacian pyramid to realize multiple dimensioned decomposition, utilizes the directivity bank of filters to realize the directivity decomposition.Owing to utilized geometry flow information, Contourlet conversion can realize, like the application of denoising and texture recovery aspect than the analysis of wavelet transform better pictures.Owing to the up-sampling and the down-sampling that wherein exist, the Contourlet conversion does not possess translation invariant shape again.For in Image Edge-Detection, profile is described and the figure image intensifying has translation invariance, has made up NSCT, and the purpose that this method is based upon on the non-desampling fir filter group basis of iteration is to describe in order to obtain a kind of directivity multi-scale image.

The NSCT conversion has combined non-sampling pyramid transform (NSP) and non-to fall sampling DFB (NSDFB), after the conversion on each yardstick the size of all directions subband identical with original image.NSCT has overcome the defective that the Contourlet conversion does not have the translation invariant characteristic effectively; And NSCT has redundancy than Contourlet conversion and human-eye visual characteristic (Human Visual System; HSV) more approaching; Therefore in image processing field, NSCT is more superior than Contourlet conversion.

Summary of the invention

The object of the present invention is to provide the NSCT underwater sound image enchancing method based on the HSV color space that can effectively keep edge of image and target property that accurate target characteristic and edge conservation degree also are provided for subsequent treatment and analysis.

The objective of the invention is to realize like this:

The present invention is based on the NSCT underwater sound image enchancing method of HSV color space, it is characterized in that:

(1) underwater sound image of rgb space is carried out the conversion of color space, be converted to the HSV space by traditional rgb space;

(2) the HSV spatial image is isolated colourity H, saturation degree S, three components of numerical value V;

The HSV color system is based on cylindrical coordinate, and the colourity H that separates, saturation degree S, numerical value V are:

$H=\left\{\begin{array}{cc}\mathrm{\θ}& G\≥B\\ 2\mathrm{\π}-\mathrm{\θ}& G\≤B\end{array}\right.,$ $S=1-\frac{\sqrt{3}}{V}\min (R,G,B),$ $V=\frac{1}{\sqrt{3}}(R+G+B),$

Wherein $\mathrm{\θ}={\mathrm{Cos}}^{-1}\left[\frac{\frac{1}{2}[(R-G)+(R-B)]}{\sqrt{{(R-G)}^2+(R-B)(G-B)}}\right];$

(3) carry out medium filtering for colourity H and two components of saturation degree S, implement NSCT for numerical value V and strengthen, obtain three new component H ', S ', V ';

(4) with H ', S ', the synthetic new HSV ' image of V ';

(5) carry out medium filtering for the HSV ' image that obtains;

(6) with the inverse transformation of filtered HSV ' image to rgb space, the image after being enhanced.

The present invention can also comprise:

1, the described method that numerical value V enforcement NSCT is strengthened is:

Logarithm value V carries out the NSCT conversion, and carry out hard-threshold for the coefficient that produces and handle, the component image after being enhanced through the NSCT inverse transformation at last, wherein threshold value is confirmed specifically to carry out as follows:

(1) logarithm value V component carries out the 2-d wavelet decomposition, obtains branch solution vector C and matrix S;

(2) 2-d wavelet is decomposed extraction all directions detail coefficients, obtain horizontal direction X respectively, three groups of coefficients of vertical direction Y and diagonal D

(3) with the noise variance σ of each layer coefficients of prior estimate form estimation, wherein intermediate value is got in the median representative:

${\mathrm{\σ}}_1=\frac{\mathrm{median}\left(\right|X\left|\right)}{0.6745},$

${\mathrm{\σ}}_2=\frac{\mathrm{median}\left(\right|Y\left|\right)}{0.6745},$

${\mathrm{\σ}}_3=\frac{\mathrm{median}\left(\right|D\left|\right)}{0.6745},$

(4) size of image is m * n, according to noise variance each layer is calculated corresponding Donoho threshold value th:

${\mathrm{th}}_1={\mathrm{\σ}}_1\sqrt{2{\log }_{10}(m\×n)},$ ${\mathrm{th}}_2={\mathrm{\σ}}_2\sqrt{2{\log }_{10}(m\×n)},$ ${\mathrm{th}}_3={\mathrm{\σ}}_3\sqrt{2{\log }_{10}(m\×n)};$

(5) final threshold value is the average of three: $T=\frac{{\mathrm{Th}}_1+{\mathrm{Th}}_2+{\mathrm{Th}}_3}{3}.$

Advantage of the present invention is: the present invention is significantly improved for the quantity of information and the sharpness of initial condition acoustic image, after enhancing, can more clearly obtain edge and detailed information, is very helpful for later further analyzing and processing etc.

Description of drawings

Fig. 1 is a process flow diagram of the present invention;

Fig. 2 is the demonstrative structure figure according to the underwater sound figure image intensifying process of the embodiment of the invention;

Fig. 3 is the demonstrative structure figure according to the underwater sound Image Intensified System of the embodiment of the invention;

Fig. 4 is according to the underwater sound image color space conversion of the embodiment of the invention and each component reinforced effects synoptic diagram;

Fig. 5 is the initial condition acoustic image and enhancing back underwater sound image comparison effect synoptic diagram according to the embodiment of the invention.

Embodiment

For example the present invention is done description in more detail below in conjunction with accompanying drawing:

In conjunction with Fig. 1～5, the objective of the invention is to utilize the combination of color space transformation and different Enhancement Method, can effectively keep edge of image and target property, for subsequent treatment and analysis provide accurate target characteristic and edge conservation degree.

Realize that technical scheme of the present invention is is instrument with medium filtering and NSCT; Adopt HSV color space model that the initial condition acoustic image is carried out resolution process; Difference figure layer component taked different Enhancement Method; Each component after will strengthening then is fused into and strengthens the back image, and final inversion gains rgb color space, the underwater sound image after being enhanced.

The objective of the invention is to realize like this:

(1.1) accomplish by the conversion of rgb space to the HSV space

(1.2) the HSV picture breakdown is become colourity H, saturation degree S, three components of numerical value V

(1.3) colourity H component and saturation degree S component are carried out medium filtering, logarithm value V component carries out NSCT and strengthens three new component H ' after being enhanced respectively, S ', V '

(1.4) with chromatic component H ', saturation degree component S ', and the synthetic new HSV ' image of new magnitude component V '

(1.5) for the new HSV ' image that obtains and carry out medium filtering

(1.6) with new HSV ' image inverse transformation to rgb space, the image after being enhanced

The embodiment of the invention has also proposed a kind of underwater sound Image Intensified System, uses the underwater sound image energy that the present invention strengthened and brings the better visual experience of function for the user and be beneficial to further processing operation.

Core technology content of the present invention is the conversion in image color space and the integrated application of medium filtering and NSCT.

The technical scheme of the embodiment of the invention is achieved in that the conversion for the color space of the laggard circumstances in which people get things ready for a trip of the rgb space underwater sound image that reads; After being converted to the HSV space; Handle respectively and isolate colourity H, saturation degree S, three components of numerical value V; Use easier medium filtering to strengthen for H and two components of S, then separately it is implemented NSCT for the numerical value V component that extracts and strengthen.New V ' component and filtered H ', S ' component are synthesized new HSV image, the new HSV ' image that obtains is carried out going back to rgb space through inverse transformation behind the medium filtering, accomplish overall enhanced operation underwater sound image.

A kind of underwater sound image enchancing method that the present invention includes, its main contents are:

The conversion in image color space, and decompose to extract and to obtain the pictures different component is selected different Enhancement Method based on the different images component, and each component after will strengthening again synthesizes image after the new enhancing, accomplishes the inverse transformation of color space at last.

A kind of underwater sound Image Intensified System that the present invention includes, its main contents are:

Image color space conversion unit is used for conversion and the inverse transformation thereof of image in the different color space; The color component enhancement unit is used for taking different Enhancement Method to strengthen different components, specifically is divided into medium filtering enhancement unit and NSCT enhancement unit;

In embodiments of the present invention, read the underwater sound image of a width of cloth rgb space after, it is carried out the conversion of color space, be converted to the HSV space by traditional rgb space after, isolate colourity H, saturation degree S, three components of numerical value V.

The HSV color system is based on cylindrical coordinate, and the conversion formula that RGB is converted to HSV is:

$H=\left\{\begin{array}{cc}\mathrm{\θ}& G\≥B\\ 2\mathrm{\π}-\mathrm{\θ}& G\≤B\end{array}\right.---\left(1\right)$

$S=1-\frac{\sqrt{3}}{V}\min (R,G,B)---\left(2\right)$

$V=\frac{1}{\sqrt{3}}(R+G+B)---\left(3\right)$

$\mathrm{\θ}={\cos }^{-1}\left[\frac{\frac{1}{2}[(R-G)+(R-B)]}{\sqrt{{(R-G)}^2+(R-B)(G-B)}}\right]---\left(4\right)$

Be converted to the inverse transformation that rgb space is above-mentioned formula from HSV.

Use easier medium filtering to strengthen for colourity H and two components of saturation degree S, medium filtering specifically carries out as follows:

1. construct a two dimension pattern plate---3 * 3 zone,

2. the plate interior pixel according to pixels is worth size and sorts, generate the dull 2-D data sequence that rises.

3. be output as g (x, y)=med{f (x-k, y-l), (k, l ∈ W) }, i.e. the intermediate value of all pixel gray-scale values in the designated field, wherein (x, y), (x y) is respectively original image and handles the back image g f, and W is 3 * 3 template zone.

Then separately it being implemented NSCT for the numerical value V component that extracts strengthens.At first it is carried out the NSCT conversion, carry out hard-threshold for the coefficient that produces and handle the component image after being enhanced through the NSCT inverse transformation at last.

Wherein threshold value is confirmed specifically to carry out as follows:

1. at first logarithm value V component carries out three layers of 2-d wavelets decomposition, obtains branch solution vector C and matrix S

2. 2-d wavelet is decomposed and extract all directions detail coefficients, obtain horizontal direction H respectively, three groups of coefficients of vertical direction V and diagonal D

3. estimate the noise variance σ of each layer coefficients with the prior estimate form, wherein intermediate value is got in the median representative

${\mathrm{\σ}}_1=\frac{\mathrm{median}\left(\right|H\left|\right)}{0.6745}---\left(5\right)$

${\mathrm{\σ}}_2=\frac{\mathrm{median}\left(\right|V\left|\right)}{0.6745}---\left(6\right)$

${\mathrm{\σ}}_3=\frac{\mathrm{median}\left(\right|D\left|\right)}{0.6745}---\left(7\right)$

4. the size of image is m * n, according to noise variance each layer is calculated corresponding Donoho threshold value

$\mathrm{th}{\mathrm{th}}_1={\mathrm{\σ}}_1\sqrt{2{\log }_{10}(m\×n)}---\left(8\right)$

${\mathrm{th}}_2={\mathrm{\σ}}_2\sqrt{2{\log }_{10}(m\×n)}---\left(9\right)$

${\mathrm{th}}_3={\mathrm{\σ}}_3\sqrt{2{\log }_{10}(m\×n)}---\left(10\right)$

5. final threshold value is the average of three

$T=\frac{{\mathrm{th}}_1+{\mathrm{th}}_2+{\mathrm{th}}_3}{3}---\left(11\right)$

New V component and filtered H, S component are synthesized new HSV image, the new HSV image that obtains is carried out medium filtering, go back to rgb space through inverse transformation at last, accomplish overall enhanced operation, promote the visual effect of image underwater sound image.

Fig. 1 is the exemplary flow chart according to the underwater sound image enchancing method of the embodiment of the invention.

As shown in Figure 1, its key step is:

1. accomplish by the conversion of rgb space to the HSV space

2. colourity H component and saturation degree S component are carried out medium filtering, logarithm value V component carries out the NSCT adaptive threshold and strengthens three new component H ' after being enhanced, S ', V '

3. with filtered chromatic component H ', saturation degree component S ', and the synthetic new HSV ' image of the magnitude component V ' after strengthening carries out medium filtering another mistake and is converted into rgb space, the image after being enhanced.

Fig. 2 is that concrete steps are illustrated in figure 2 as according to the demonstrative structure figure of the underwater sound figure image intensifying process of the embodiment of the invention:

1. accomplish by the conversion of rgb space to the HSV space

2. the HSV picture breakdown is become colourity H, saturation degree S, three components of numerical value V

3. colourity H component and saturation degree S component are carried out medium filtering, logarithm value V component carries out NSCT and decomposes, and adopts the mode of adaptive threshold to strengthen three new component H ' after being enhanced respectively, S ', V '

4. with chromatic component H ', saturation degree component S ', and the synthetic new HSV ' image of new magnitude component V '

5. new HSV ' image is carried out medium filtering;

With new HSV ' image inverse transformation to rgb space, the image after being enhanced.

Based on above-mentioned image enhancement processing method, the present invention also provides a kind of Image Intensified System.

Fig. 3 is the demonstrative structure figure according to the underwater sound Image Intensified System of the embodiment of the invention.As shown in Figure 3, this system comprises the colour space transformation unit, enhancement unit and color space inverse transformation block

The colour space transformation unit will be accomplished from the conversion of rgb space to the HSV space.

Enhancement unit will utilize medium filtering to chromatic component H, and saturation degree component S carries out denoising and strengthens, and logarithm value V component carries out NSCT and strengthens.

The color space inverse transformation block will be accomplished the spatial alternation of above-mentioned enhancing underwater sound image.

Fig. 4 is according to the underwater sound image color space conversion of the embodiment of the invention and each component reinforced effects synoptic diagram.First row is respectively the initial condition acoustic image of rgb space from left to right, is converted to the initial condition acoustic image in HSV space and the HSV space underwater sound image after the enhancing; Second row is respectively the tone H component that the decomposition of initial condition acoustic image obtains under the HSV space, saturation degree S component, numerical value V component from left to right; The 3rd row is respectively the tone H ' component that adopts after medium filtering strengthens from left to right, the numerical value V ' component after saturation degree S ' component and NSCT strengthen.

Fig. 5 is the initial condition acoustic image and enhancing back underwater sound image comparison effect synoptic diagram according to the embodiment of the invention.Noise circumstance shown in the figure is the zero-mean Gaussian noise of variance 10, can find out that through the contrast of two figure the underwater sound image after the enhancing has been removed the noise spot in the original image, has promoted contrast, and the edge is more clear.

In sum, the underwater sound image enchancing method that the present invention combines based on NSCT conversion and medium filtering under the HSV color space can know that through experiment this method is significantly improved for the quantity of information and the sharpness of initial condition acoustic image.After enhancing, can more clearly obtain edge and detailed information, this advantage is very helpful for later further analyzing and processing etc.Design though The present invention be directed to the characteristics of underwater sound image, based on different applications, the present invention goes for other relevant image processing field equally through suitable modification.

Claims (2)

1. based on the NSCT underwater sound image enchancing method of HSV color space, it is characterized in that:

(1) underwater sound image of rgb space is carried out the conversion of color space, be converted to the HSV space by traditional rgb space;

(2) the HSV spatial image is isolated colourity H, saturation degree S, three components of numerical value V;

The HSV color system is based on cylindrical coordinate, and the colourity H that separates, saturation degree S, numerical value V are:

$H=\left\{\begin{array}{cc}\mathrm{\θ}& G\≥B\\ 2\mathrm{\π}-\mathrm{\θ}& G\≤B\end{array}\right.,$ $S=1-\frac{\sqrt{3}}{V}\min (R,G,B),$ $V=\frac{1}{\sqrt{3}}(R+G+B),$

Wherein $\mathrm{\θ}={\mathrm{Cos}}^{-1}\left[\frac{\frac{1}{2}[(R-G)+(R-B)]}{\sqrt{{(R-G)}^2+(R-B)(G-B)}}\right];$

(3) carry out medium filtering for colourity H and two components of saturation degree S, implement NSCT for numerical value V and strengthen, obtain three new component H ', S ', V ';

(4) with H ', S ', the synthetic new HSV ' image of V ';

(5) carry out medium filtering for the HSV ' image that obtains;

(6) with the inverse transformation of filtered HSV ' image to rgb space, the image after being enhanced.

2. the NSCT underwater sound image enchancing method based on the HSV color space according to claim 1 is characterized in that: the described method that numerical value V enforcement NSCT is strengthened is:

Logarithm value V carries out the NSCT conversion, and carry out hard-threshold for the coefficient that produces and handle, the component image after being enhanced through the NSCT inverse transformation at last, wherein threshold value is confirmed specifically to carry out as follows:

(1) logarithm value V component carries out the 2-d wavelet decomposition, obtains branch solution vector C and matrix S;

(2) 2-d wavelet is decomposed extraction all directions detail coefficients, obtain horizontal direction X respectively, three groups of coefficients of vertical direction Y and diagonal D

(3) with the noise variance σ of each layer coefficients of prior estimate form estimation, wherein intermediate value is got in the median representative:

${\mathrm{\σ}}_1=\frac{\mathrm{median}\left(\right|X\left|\right)}{0.6745},$

${\mathrm{\σ}}_2=\frac{\mathrm{median}\left(\right|Y\left|\right)}{0.6745},$

${\mathrm{\σ}}_3=\frac{\mathrm{median}\left(\right|D\left|\right)}{0.6745},$

(4) size of image is m * n, according to noise variance each layer is calculated corresponding Donoho threshold value th:

${\mathrm{th}}_1={\mathrm{\σ}}_1\sqrt{2{\log }_{10}(m\×n)},$ ${\mathrm{th}}_2={\mathrm{\σ}}_2\sqrt{2{\log }_{10}(m\×n)},$ ${\mathrm{th}}_3={\mathrm{\σ}}_3\sqrt{2{\log }_{10}(m\×n)};$

(5) final threshold value is the average of three: $T=\frac{{\mathrm{Th}}_1+{\mathrm{Th}}_2+{\mathrm{Th}}_3}{3}.$

Images