DE102005032287B4 - Processing method for a two-dimensional initial image and objects corresponding thereto - Google Patents

Abstract

Processing method for a two-dimensional initial image (B) by a computer,
- wherein the computer decomposes the initial image (B) into partial images (T _i ) and a residual image (R),
- Each sub-image (T _i ) contains those portions of the two-dimensional initial image (B), which vary locally with frequencies that are in a predetermined, for the respective sub-image (T _i ) characteristic, the frequency zero non-containing field frequency range, and the residual image (R) contains at least a DC component of the initial two-dimensional image (B) that is locally invariable,
- Wherein the computer for each pixel of each sub-image (T _i ) determines a specific for the respective pixel of the respective sub-image (T _i ) factor (g), the respective pixel of the respective sub-image (T _i ) with that for the respective pixel of the respective Sub-image (T _i ) multiplied by the specific factor (g) and thus processed partial images (T _i '),
Wherein the computer determines a processed residual image (R ') on the basis of the residual image (R),
- where the calculator the edited fields ...

Description

The The present invention relates to a processing method for a two-dimensional one Initial image. The present invention further relates to a disk with a stored on the disk Computer program for a computer to carry out such a processing method. Finally, the present concerns Invention still a computer with such a disk.

initial images in the context of the present invention know z. B. two-dimensional Image signals are from X-ray transmission systems to be delivered. These are usually discrete Image signals. The image signals can Be raw signals or already partially processed signals. They often contain medical meaning content and become in particular for diagnosis and partly also in the context of therapy or intervention used. They usually have high dynamics on. Furthermore, the images often contain fine details, contain the relevant image information, z. B. small vessels that filled with contrast agent are, or guidewires. In many cases However, the attending physician may not or only poorly mapping parameters and adjust rendering parameters such that the image is treated well. Nevertheless, especially in scenic imaging, so if there is a sequence of initial pictures, the one shown Picture also immediately show all relevant details.

Around too big To avoid dynamics in the images shown, are in the state The technique often harmonizes with a given fixed Filter core made essentially a high-pass filter equivalent. A major disadvantage of such high-pass filters is however the education of overshooters On edges that cause the doctor in the extreme case even to misdiagnosis can, but usually cause at least an unfamiliar image impression.

Farther the noise impression of the picture is of importance to the viewer. This is especially true in the X-ray based Imaging. Because if the useful signal and the noise are not from each other can be separated is the choice for the right contrast very difficult. If the contrast is too big, it will Eye of the viewer too much irritated by the noise. is the contrast is too low, the signal is barely or even even no longer recognizable.

• – Der Rechner zerlegt das Anfangsbild in Teilbilder und ein Restbild. Jedes Teilbild enthält dabei diejenigen Anteile des zweidimensionalen Anfangsbildes, die mit Frequenzen örtlich variieren, die in einem vorbestimmten, für das jeweilige Teilbild charakteristischen, die Frequenz Null nicht enthaltenden Teilbildfrequenzbereich liegen. Das Restbild enthält zumindest einen Gleichanteil des zweidimensionalen Anfangsbildes, der örtlich invariabel ist.
• – Für jedes Pixel jedes Teilbildes ermittelt der Rechner einen für das jeweilige Pixel des jeweiligen Teilbildes spezifischen Faktor, mit dem er dieses Pixel multipliziert. Auf diese Weise ermittelt der Rechner bearbeitete Teilbilder.
• – Weiterhin ermittelt der Rechner anhand des Restbildes ein bearbeitetes Restbild.
• – Die bearbeiteten Teilbilder und das bearbeitete Restbild kombiniert der Rechner zu einem Endbild.
• – Wenn für das Pixel eine vorbestimmte Bedingung erfüllt ist, setzt der Rechner den Faktor auf einen Wert oberhalb eines Minimalwertes. Anderenfalls setzt er den Faktor auf den Minimalwert. Der Minimalwert ist dabei für alle Pixel des jeweiligen Teilbildes der gleiche.

For image processing is - for example from the US 6,252,931 B1 - the following procedure is known:

• - The computer decomposes the initial image into partial images and a residual image. In this case, each partial image contains those portions of the two-dimensional initial image that vary locally with frequencies which are in a predetermined partial image frequency range that is characteristic of the respective partial image and does not contain the frequency zero. The residual image contains at least a DC component of the two-dimensional initial image, which is locally invariable.
• For each pixel of each partial image, the computer determines a specific factor for the respective pixel of the respective partial image with which it multiplies this pixel. In this way, the computer determines processed partial images.
• - Furthermore, the computer determines a processed residual image based on the residual image.
• - The edited partial images and the processed residual image are combined by the calculator into a final image.
• If the pixel meets a predetermined condition, the calculator sets the factor to a value above a minimum value. Otherwise it sets the factor to the minimum value. The minimum value is the same for all pixels of the respective sub-image.

In the US 6,252,931 B1 enter into the determination of the factor sizes that are not part of the respective sub-image, for example, pixels of the sub-image with the next lower field frequency range or the initial image. The factor can be a minimum of 1, but it can also be greater than.

The procedure of US 6,252,931 B1 often leads to the fact that - according to objective, signal-technically verifiable criteria - the image quality is improved and the recognizability of relevant image information is improved. The procedure of US 6,252,931 B1 However, it is also associated with a change in the subjective noise impression of the viewer. This change in the impression of noise is often perceived as irritating and disturbing, sometimes even unacceptable.

from Inventors of the present invention recognized that the change the noise impression in the process of the prior art their The reason for this is that when determining the edited drawing files each attempt is made to leave the useful signal unaffected, the To dampen noise, on the other hand.

From the DE 101 46 582 A1 For example, a processing method for a two-dimensional initial image by a computer is known. In this processing method, the computer decomposes the initial image into partial images and a residual image. Each field contains those parts of the two-dimensional An Initial image, which vary locally with frequencies that are in a predetermined, characteristic of the respective sub-image, the frequency zero not containing field frequency range. The residual image contains at least a DC component of the two-dimensional initial image, which is locally invariable. The computer determines a modified pixel value for each pixel of each partial image and thus determines processed partial images. It also determines a processed residual image based on the remaining image. The processed partial images and the edited residual image are combined by the calculator into a final image. The computer modifies the respective pixel value of a pixel by setting the pixel value to zero when the magnitude of the pixel value is below a threshold. Otherwise, the calculator reduces the amount of the pixel value by the threshold.

Further processing methods for two-dimensional initial images are known from the US 5,526,446 A , of the US 5,561,724 A , of the US 5,602,934 A and the reference book "Digital Image Processing" by RC Gonzalez and RE Woods, 2nd edition, Addison-Wesley, 2002, pages 239-241.

The The object of the present invention is a machining method of the type mentioned above and the corresponding objects such to design that improving the recognizability of the relevant Image information is retained, the noise impression but as subjective is felt much more pleasant.

• – dass der Rechner zum Ermitteln des für das jeweilige Pixel des jeweiligen Teilbildes spezifischen Faktors ermittelt, ob er in dem jeweiligen Teilbild innerhalb eines zweidimensionalen Filterkerns um de jeweilige Pixel herum eine Struktur erkennt,
• – dass der Rechner den Faktor bejahendenfalls auf einen wert oberhalb des Minimalwertes und verneinendenfalls auf den Minimalwert setzt,
• – dass der Rechner zum Erkennen der Struktur innerhalb des Filterkerns um das jeweilige Pixel des jeweiligen Teilbildes herum die statistischen Werte von mehreren Pixelgruppen bildet, wobei die Pixel jeder Pixelgruppe zumindest im Wesentlichen auf einer Geraden liegen und als Zentralpixel das jeweilige Pixel enthalten, und
• – dass der Rechner den Faktor für das jeweilige Pixel des jeweiligen Teilbildes anhand eines für die Schwankung der statistischen Werte charakteristischen Wertes ermittelt.

The task is solved by

• In that the computer determines, for determining the factor specific to the respective pixel of the respective partial image, whether it recognizes a structure in the respective partial image within a two-dimensional filter core around the respective pixels,
• - that the computer sets the factor to a value above the minimum value and, if so, to the minimum value,
• - That the calculator for recognizing the structure within the filter core around the respective pixel of the respective field around the statistical values of several pixel groups, the pixels of each pixel group are at least substantially on a straight line and as the central pixel containing the respective pixel, and
• - That the calculator determines the factor for each pixel of each sub-image using a value characteristic of the fluctuation of the statistical values.

Herewith Correspondingly, the task is handled by a data medium, on the a computer program for a computer to carry out of such a processing method is stored. Also will solved the problem by a computer with such a disk.

The Determine the for the fluctuation of statistical values of characteristic value For example, the calculator can use at least two of the values Make the maximum, minimum and average of the statistical values.

When suitable statistical values of the pixel groups come in particular their mean values and their variances are considered.

Preferably then the calculator sets the factor to the minimum value if the fix is characteristic of the fluctuation of statistical values Value is below a threshold. Furthermore, the calculator lifts the factor is preferably linear with that for the fluctuation of the statistical values characteristic value, if necessary for the pivoting of the statistical Values characteristic value is above the threshold.

Of the Threshold is preferably for all partial images the same. He can from the computer based on the initial image and / or determined by a user input. For example can be a percentage of the average brightness value of the Initial image used to calculate the threshold, the Percentage can be specified by the user.

Analogous the threshold is preferably also the minimum value for all fields the same. Alternatively, it is also possible that the minimum value individually for the respective partial image is. Also the minimum value or the minimum values can (preferably) The computer can be specified by a user.

It is possible, that the residual image is exclusive contains the DC component. But it is also possible that the residual image in addition to the DC component those parts of the two-dimensional initial image contains those with frequencies locally which are smaller than the frequencies of the lowest field frequency range are.

Of the Computer determines the processed residual image preferably by that he treats the rest of the picture like a partial picture. Alternatively it is but it is also possible that the computer for determining the edited residual image, the pixels of the residual image multiplied by a location-independent residual image factor.

Of the Filter core has a filter core size and at least one other Parameters. At least the other parameter should preferably for all Partial images will be the same.

Further advantages and details result from the following description of an embodiment in conjunction with the drawings. Here are a schematic diagram:

1 a basic structure of a computer,

2 a flow chart,

3 a block diagram of an image processor,

4 another flowchart,

5 a first two-dimensional filter core,

6 a second two-dimensional filter core and

7 a third flowchart.

According to 1 a computer has among other things a central unit 1 and a mass storage 2 on. The mass storage 2 can be configured for example as a hard disk. In the mass storage 2 is a computer program 3 deposited, at the call of the computer performs a processing method for a two-dimensional initial image B, the computer via a suitable interface 4 is supplied.

The computer program 3 can the computer, for example, via an interface 5 to a computer network 6 have been fed. The computer network 6 In this case, in particular, the Internet or World Wide Web can be. But it is also possible that the computer program 3 the computer via a data carrier 7 is supplied, which is designed as a removable medium. On such a disk 7 is for example the computer program 3 stored in machine-readable form only. In this case, the disk becomes 7 in a suitable reading device 8th of the computer. The computer then reads the computer program 3 by means of the reading device 8th from the disk 7 and stores it in the mass storage 2 from. In embodiment of the data carrier 7 as a CD-ROM, the reading device 8th For example, be designed as a CD-ROM drive or as a DVD drive.

When calling the computer program 3 the computer reads the computer program 3 from the mass storage 2 off and start it. Under execution of the computer program 3 then the computer executes the processing method according to the invention for the beginning picture B. This processing method will be described below in connection with 2 explained in more detail.

According to 2 the computer takes in a step S1 from a user 9 via an input device 10 initially opposed by a basic factor F1. The basic factor F1 is greater than 0 and a maximum of 1. Then takes the computer from the user 9 in a step S2, a minimum value M or a plurality of minimum values M _i . Furthermore, the calculator takes from the user 9 in a step S3 against a brightness value H.

The steps S1 to S3 are - alone or together - only optional. So they are not mandatory. That's why they are in 2 only dashed lines. As far as the steps S1 to S3 omitted, the corresponding input variables F1, M or M _i and H are set by the computer to predefined values.

In a step S4, the computer takes over the interface 4 the initial image B opposite. The initial image B is supplied to the computer, for example, directly and online from an X-ray device. The X-ray device is in 1 however, not shown for the sake of clarity.

In a step S5, the calculator then determines based on the basic factor F1, an additional factor F2 and the initial image B a threshold value t. He ascertains the average brightness value of all pixels of the initial image B and multiply this average value by the basic factor F1 and the additional factor F2.

In a step S6, the computer decomposes the initial image B into partial images T _i (i = 0... N) and a residual image R. The decomposition can take place in any desired manner. For example, a decomposition can take place by means of a Gaussian pyramid. According to embodiment - see 3 - The initial image B is block in a Laplace pyramid by means of a number of reduction blocks 11 and expansion blocks 12 as well as summation points 13 into the partial images T _i and the residual image R decomposed.

The structure and operation of the Laplace pyramid is well known to those skilled in the art: the reduction blocks 11 Perform a low-pass filtering and a sub-sampling by a factor of 2 in both dimensions of each image supplied to them. The following expansion blocks 12 perform a corresponding expansion and interpolation with the factor 2, preferably also in both dimensions.

As a result, the partial image T ₀ thus contains those portions of the initial image B which vary locally with frequencies which lie in a highest partial image frequency range that is characteristic of the partial image T ₀ . The sub-picture T ₁ contains those portions of the initial image B, which varies locally with frequencies which lie in a corresponding second highest field frequency range. Analogous explanations apply to the other partial images T _i . For the corresponding field frequency ranges, the respective frequencies are becoming smaller and smaller.

The Residual image R contains at least a DC component of the two-dimensional initial image B. It contains So the proportion of the two-dimensional initial image B, which is not locally varied. If the Laplace pyramid is high enough, contains the residual image R even exclusively the DC component. If the Laplace pyramid, however, not high is enough contains the residual image R in addition to the DC component those parts of the two-dimensional initial image B, which with frequencies locally which are smaller than the frequencies of the lowest field frequency range are.

In a step S7 checks the calculator next, whether he should treat the residual image R as a partial image. If this the case is, the calculator in a step S8, the residual image R as (additional) partial image one. Otherwise, the computer multiplies in a step S9 the pixels of the residual image R with a location-independent residual image factor F3 and determines a processed residual image R '. The residual image factor F3 is calculated by the computer determined so that the pixels of the processed residual image R 'is the brightness value H have.

In a step S10, the computer selects one of the partial images T _i and in a step S11 a pixel of the selected partial image T _i .

In a step S12, the computer then checks whether it can recognize a structure within a two-dimensional filter kernel around the pixel selected in step S11. Depending on this, the computer then calculates a factor g for this pixel of this partial image T _i and stores this factor g. Step S12 will be described later in connection with 4 will be explained in more detail.

In a step S13, the computer checks whether it has already performed step S12 for all the pixels of the selected subpicture T _i . If this is not the case, the computer goes to a step S14 in which he selects another, previously untreated pixel of the selected field T _i . Then it returns to step S12.

By contrast, if the computer has already determined the corresponding pixel-specific factor g for all pixels of the selected partial image T _i , the computer transfers to a step S 15. In step S15, the computer multiplies each pixel of the selected partial image T _i by the corresponding pixel-specific factor g and thus determines a processed partial image T _i '.

According to 3 this multiplication takes place in multiplier blocks 14 , The reference symbols G _i (i = 0..., N or n + 1) each designate a matrix with the pixel-specific factors g of the corresponding partial image T _i , possibly also of the residual image R.

In a step S16, the computer checks whether it has already processed all partial images T _i (possibly including the residual image R, if this was classified as partial image in step S8). If this is not the case, the computer selects in a step S17 another, previously unprocessed partial image T _i (possibly also the residual image R) and returns to step S11. Otherwise, in a step S18, the computer combines the processed partial images T _i 'and the processed residual image R' into an end image B '. The combining takes place - see 3 - in summation points 15 , if necessary, the processed partial images T _i 'and the processed residual image R' are in expansion blocks 16 expands and interpolates.

If the residual image R has been classified as a partial image in step S8, a step S19 is further executed. In step S19, the computer determines the average brightness value of the pixels of the final image B '. He then determines a location-independent final image factor F4 and multiplies the final image B 'with this final image factor F4. The final image factor F4 is determined such that the average brightness value of the pixels of the final image B 'is scaled to the brightness value H. This step S19 is according to 3 for example, in a multiplier block 17 executed. On the other hand, if the residual image R has been treated differently than the partial images T _i , that is, if step S 9 has been passed through, step S 19 may be omitted. The step S19 is therefore in 2 shown only by dashed lines.

Finally, in step S20, the final image B 'is finally sent by the computer via a suitable output device 18 , z. B. a viewing device, to the user 9 output.

When last checks the computer in a step S21, whether another initial image B to be edited. If this is the case, it becomes the step S4 declined. Otherwise, the execution of the Processing procedure ended.

For determining whether a structure is within the two-dimensional filter kernel around a pixel in one of the partial images T _i detected, so the implementation of step S12, according to 4 preferably proceed as follows:
First, the computer forms average values μj of several pixel groups within the filter kernel around the respective pixel of the respective partial image T _i . The pixels of each pixel group are at least essentially on a straight line and contain the respective selected pixel as the central pixel. For example, if the two-dimensional filter kernel - see 5 - Has a filter core size of 3 × 3 pixels, thus averages μ1 to μ4 can be determined. If the filter kernel according to 6 For example, having a filter kernel size of 5x5 pixels may, in addition to the mean values, be μ1 through μ4 (these are by design 5 visible and in 6 for reasons of clarity not shown) also mean values μ5 to μ8 are determined.

In In a step S32, the computer then determines an average value μ of the mean values μj and Minimum value μ 'of the mean values μj. Further determined he in step S32, the difference δ of Average μ and Minimum value μ 'of the mean values μj. These Difference is characteristic for the fluctuation of the mean values μj.

In a step S33, the computer checks whether the difference δ is greater than the threshold value t determined in step S5. If this is the case, in a step S34 the computer determines the pixel-specific factor g for (1 + δ -t) M or for (1 + δ -t) M _i , respectively, depending on whether the minimum value M applies to all partial images T _{i is} the same or whether the minimum value M _i for the respective field T _{i is} individual. Otherwise, ie if the difference δ is not greater than the threshold value t, the computer executes a step S35 by setting the pixel-specific factor g to the minimum value M or M _i .

The calculator then sets the factor to the minimum value M or M _i only when the value δ characteristic of the fluctuation of the average values μj is below the threshold value t. On the other hand, if the characteristic value δ is above the threshold value t, the calculator increases the factor linearly with this value δ. In any case, however, the factor for the respective pixel of the respective partial image Ti is determined on the basis of the value δ.

The mean values μj are statistical values of the pixel groups. However, other statistical values can also be used, in particular the variances σ of the pixel groups. In this case - as an alternative to the procedure of 4 - according to 7 proceed as follows:
First, the computer determines in a step S41 within the filter kernel around the selected pixel of the respective subpicture T _i variances σj of several pixel groups. The pixel groups are the same pixel groups already in connection with the 4 to 6 were explained.

In In a step S42, the computer then determines the maximum σ and the Minimum σ 'of the variances σj and the Difference δ this both values σ, σ '. This difference δ is again characteristic of the Variation of variances σj. In particular, it indicates whether the considered pixel is part of it an image edge is.

The steps S43 to S45 correspond to the steps S33 to S35 of FIG 4 and will therefore not be discussed again in detail below.

From the above remarks to the 4 and 7 it also follows that the filter kernel in addition to its filter core size (see 5 and 6 ) has a further parameter t, namely the threshold value t. This parameter t is the same for all fields T _i . But he can, depending on whether the procedure according to 4 or the procedure according to 7 take a specific value for the respective procedure.

As already mentioned, the minimum value M or M _{i can} alternatively be the same for all partial images T _i or else individually for the respective partial image T _i . If the minimum value M _{i is} individual to the respective subpicture T _i , it may in particular be the weighting factor of the patent application filed shortly before the present invention titled "Processing method for a two-dimensional initial image and objects corresponding thereto", inventors Philipp Bernhardt and Markus Lendl, filing date at the German Patent Office 22.06.2005, official file number 10 2005 028 892.8 , correspond. Regardless of which procedure is taken, the minimum value M or M _i, however, is the same for all pixels of the respective subpicture T _i .

By means of the procedure according to the invention it is possible in particular to obtain the natural noise impression and nevertheless to apply a good noise reduction method. Ultimately, this is the intuitive recognition of useful signals by the user 9 improved and thus causes a significant patient and user benefits.

Claims (19)

Processing method for a two-dimensional initial image (B) by a computer, - wherein the computer decomposes the initial image (B) into partial images (T _i ) and a residual image (R), - wherein each partial image (T _i ) those portions of the two-dimensional initial image (B ), which vary locally with frequencies which are in a predetermined, for the respective field (T _i ) characteristic, the frequency zero not containing field frequency range, and the residual image (R) at least a DC component of the two-dimensional initial image (B), the is locally invariable, - wherein the computer for each pixel of each field (T _i ) one for the respective pixel of the respective Teilbil of the (T _i ) specific factor (g), the respective pixel of the respective partial image (T _i ) is multiplied by the factor (g) specific to the respective pixel of the respective partial image (T _i ) and partial images (T _i ') thus processed the computer determines a processed residual image (R ') on the basis of the residual image (R), the computer combining the processed partial images (T _i ') and the processed residual image (R ') into an end image (B'), Wherein the computer determines, for determining the factor (g) specific to the respective pixel of the respective subpicture (T _i ), whether it recognizes a structure in the respective subpicture (T _i ) within a two - dimensional filter kernel around the respective pixel, the computer sets the factor (g) to a value above a minimum value (M or M _i ) and, if so, to the minimum value (M or M _i ), the minimum value (M or M _i ) for all pixels of the respective subpicture (T _i ) is the same, wherein the computer for detecting the structure within the filter core around the respective pixel of the respective field (T _i ) around statistical values (μ _j , σ _j ) of several Pi xelgruppen forms, the pixels of each pixel group at least in lie substantially on a straight line and contain the respective pixel as the central pixel, and - wherein the computer calculates the factor (g) for the respective pixel of the respective partial image (T _i ) on the basis of a value characteristic of the fluctuation of the statistical values (μj, σj) ( δ). Processing method according to claim 1, characterized that the calculator based on at least two of the values maximum (σ), minimum (μ ', σ') and average value (μ) the statistical values (μj, σj) for the fluctuation of statistical values (μj, σj) characteristic Value (δ) determined. Processing method according to claim 1 or 2, characterized in that the statistical values (μj) of the pixel groups are their mean values (μj) are. Processing method according to claim 1 or 2, characterized characterized in that the statistical values (σj) of the pixel groups their variances (σj) are. Processing method according to one of the preceding claims, characterized in that the computer sets the factor (g) to the minimum value (M or M _i ) if and only if the value characteristic of the fluctuation of the statistical values (μj, σj) ( δ) is below a threshold value (t). Processing method according to claim 5, characterized in that that the calculator the factor (g) linear with that for the fluctuation of the statistical Values (μj, σj) characteristic Value (δ) raises when the for the fluctuation of the statistical values (μj, σj) characteristic value (δ) above of the threshold (t). Processing method according to claim 5 or 6, characterized in that the threshold value (t) for all sub-images (T _i) is the same. Processing method according to claim 5, 6 or 7, characterized in that the computer uses the threshold value (t) of the initial image (B). Processing method according to one of claims 5 to 8, characterized in that the computer the threshold value (t) determined by a user input. Processing method according to one of claims 1 to 9, characterized in that the minimum value (M) for all sub-images (T _i ) is the same. Processing method according to one of claims 1 to 9, characterized in that the minimum value (M _i ) is individual for the respective subpicture (T _i ). Processing method according to any of the above claims, characterized in that the minimum value (M or M _i) the computer (from a user 9 ) is given. Processing method according to one of claims 1 to 12, characterized in that the residual image (R) exclusively the DC component contains. Processing method according to one of claims 1 to 12, characterized in that the residual image (R) in addition to DC component those parts of the two-dimensional initial image Contains (B) those with frequencies locally which are smaller than the frequencies of the lowest field frequency range are. Processing method according to one of claims 1 to 14, characterized in that the computer, the processed residual image (R ') thereby determined that he treats the residual image (R) like a partial image. Processing method according to one of claims 1 to 14, characterized in that the computer for determining the processed Residual image (R ') the pixels of the residual image (R) with a location-independent residual image factor (F3) multiplied. Processing method according to one of the above claims, characterized in that the Filter core has a filter core size (eg 3 × 3 or 5 × 5) and at least one other parameter (t) and that the other parameter (t) for all fields (T _i ) is the same. Data carrier with a computer program stored on the data medium ( 3 ) for a computer for performing a processing method according to any one of the above claims. Computer with a data carrier ( 2 ) according to claim 18.