BE1022166B1 - IMAGE COMPRESSION METHOD - Google Patents

Abstract

A process for creating, from a digital image, a foreground layer for use in image compression such as MRC (Mixed Raster Content). A mask is first determined. Each pixel in the mask corresponds to a scaling window in the digital image. Black pixels in the mask correspond to foreground elements, and white pixels in the mask correspond to background elements. Then, a transition image is determined in which white pixels in the mask get an empty value and black pixels in the mask get a value determined from the colors of the pixels in an averaging window, said averaging window of one. pixel comprising more pixels than the scaling window of said pixel.

Description

"Image compression process"

Field of the invention The invention relates to a method of compressing images. Specifically, the invention relates to a method for creating from a digital image a foreground layer for use in image compression.

Prior art of the invention

Image compression (printing, scanning, binarization) changes the colors of an image, especially the edge of elements as characters where the contrast between adjacent pixels is high. This modification introduces a certain inhomogeneity in the colored elements of the image, which results in a certain inhomogeneity of the foreground layer used in the MRC compression. The fact that MRC compression, as described for example in US2012 / 0093412A1, can not provide such a high compression ratio for an inhomogeneous foreground layer as for a homogeneous foreground layer is a problem. '

Another problem is that a deterioration in the homogeneity of the foreground layer causes a deterioration in the readability of the image after compression. Summary of the invention

Therefore, an object of the present invention is to provide a method for creating a foreground image of good homogeneity for use in image compression.

Another object of the present invention is a compression method using a foreground image of good homogeneity.

Another object of the present invention is a non-temporary computer readable medium storing a program causing a computer to execute a method for creating a foreground layer of good homogeneity for use in image compression.

Another object of the present invention is a system comprising a processor and a non-temporary computer readable medium storing a program causing a computer to execute a method for creating a foreground layer of good homogeneity to be used in the image compression.

This object is achieved according to embodiments of the invention.

According to one aspect of the present invention, there is provided a method for creating, from an initial digital image, a foreground image for use in image compression, including the steps of creating the image. a binary mask image from the initial digital image, each pixel of the binary mask image being created from a corresponding first pixel window in the initial digital image, and said binary mask image including a first set of pixels which are of a first value and a second set of pixels which are of a second value; b / generating a transition image that has an amount of pixels equal to the pixel amount of the bit mask image by determining for each pixel of the transition image corresponding to a pixel of the first set of pixels in the image. binary mask image a corresponding second pixel window in the initial digital image, the second pixel window comprising more pixels than the first pixel window; determining a color for each pixel of the transition image corresponding to a pixel of the first set of pixels in the bitmask image by a calculation based on the colors of the pixels in the corresponding second window; and establishing each pixel of the transition image corresponding to a pixel of the second set of pixels in the binary mask image as empty; and cl creating the foreground image from the transition image by keeping the color of the non-empty pixels and determining a color for the empty pixels using the neighboring non-empty pixels.

Since the second window has more pixels than the first window, the colors determined for nearby pixels are closer than if the colors were determined on the second window. This induces more similar colors on the pixels of the transition image and therefore more similar colors on the pixels of the foreground layer. This good homogeneity allows good compression and readability of text characters compressed in this way.

According to one aspect of the invention, the first pixel window comprises a pixel. This is the case when the binary mask image is the same size as the initial digital image.

According to another aspect of the invention, the first pixel window comprises a window η xn of pixels, n being at least 2. This is the case when the bit mask image has a lower resolution than the digital image initial.

According to one aspect of the invention, the calculation based on the colors of the pixels in the corresponding second window is an averaging.

According to one aspect of the invention, the location of the center of a pixel in the transition image corresponds to the location of the center of the corresponding second pixel window in the initial digital image. The second window indicates the pixels of the initial digital image used to determine the color of the pixels of the transition image. It is therefore advantageous that the location of the center of the second window corresponds to the center of the corresponding pixel of the transition image.

According to one aspect of the invention, the method comprises a step of reducing the resolution of the transition image between its generation and the creation of the foreground image. Such compression may be advantageous to achieve higher compression and lower resolution for the foreground image.

According to one aspect of the invention, the color of the empty pixels is determined from non-empty pixels using a two-step sequential algorithm.

According to one aspect of the invention, the method comprises the step of splitting the initial digital image into groups of pixels, the color of each pixel in the transition image being determined independently for each group of pixels. According to one aspect of the invention, the method comprises the step of splitting the bit mask image into groups of pixels corresponding to the groups of pixels in the initial digital image.

According to one aspect of the invention, each group of pixels in the initial image contains at least one of: a single connected component, a single textual element, a single lineart element, or a single character. This aspect of the invention is particularly advantageous when elements (connected component, letters, symbols, solid lines, ...) of different colors are close to each other in the initial digital image. The elements are distributed in groups of pixels. This aspect of the invention causes these colors of different groups of pixels to not mix in the foreground image.

According to one aspect of the invention, the method comprises the steps of: detecting whether, in a group of pixels in the initial digital image, the pixels corresponding to the pixels of the first set of pixels of the bit mask image have colors that are significantly different, and to determine, if the pixels in the initial digital image do not have significantly different colors, the average pixel color in the initial digital image and to apply the average color on each pixel in the transition image corresponding to the group of pixels in the initial digital image and corresponding to the pixels, in the first set of pixels of the bit mask image.

This aspect of the invention is particularly advantageous if the pixels in an element corresponding to a group of pixels are of similar colors. In such a case, it is sufficient to calculate once the color for the group and to give this color to all the pixels of the transition image corresponding to this group of pixels and which are of first value, for example black, in the binary mask. This additional process reduces the number of calculation steps in the process, since averaging is performed only once per group and provides higher homogeneity, since all pixels of the element in the image of transition have the same color. This averaging over the entire group is however not used if the pixels of the group have too different colors.

According to one aspect of the invention, there is provided a compression method for compressing an initial digital image of a scanned document, said compression method comprises the step of segmenting the initial digital image into multiple layers of an image comprising a foreground image containing color information for the foreground elements of said document, a background image containing color information for the background elements of said document and an image binary for selecting between pixels in said foreground image and said background image, the foreground image being created according to any of the methods for creating a foreground image according to the invention.

According to one aspect of the invention, there is provided a non-temporary computer readable medium storing a program causing a computer to execute a method according to any of the methods for creating a foreground image according to the invention. According to one aspect of the invention there is provided a system comprising a processor and a non-temporary computer readable medium storing a program causing the processor to execute any of the methods for creating a foreground image according to the present invention. 'invention. The method of the invention is advantageously implemented by a computer program.

According to one aspect of the invention, the system includes an imaging device for capturing the initial digital image. An imaging device may introduce inhomogeneity into the initial digital image, which is at least partially compensated in the image resulting from the compression by the method according to the invention.

Brief description of the figures

For a better understanding of the present invention, reference will now be made, by way of example only, to the accompanying drawings in which:

Figure 1 shows a flowchart of an image compression method.

Figure 2 shows a flow diagram of an image segmentation method.

Figure 3 shows a flowchart of a method of constructing a foreground layer.

Figure 4 shows a flowchart of a method of constructing a foreground layer according to an embodiment of the invention.

Fig. 5 shows a flowchart of a method for determining a transition image according to an embodiment of the invention.

Figure 6 illustrates a method of determining a transition image according to an embodiment of the invention.

Fig. 7 illustrates a method of determining a transition image according to an embodiment of the invention.

Figure 8 shows a flowchart of a method for determining a transition image according to an embodiment of the invention.

Figure 9 illustrates a method of determining a transition image according to an embodiment of the invention.

Fig. 10 shows a flowchart of a method for determining a transition image according to an embodiment of the invention.

Description of the invention

The present invention will be described in connection with particular embodiments and with reference to certain drawings but the invention is however not limited thereto. The drawings described are only schematic and are non-limiting. In the drawings, the size of some of the elements may be exaggerated and not drawn to scale for illustrative purposes.

In addition, the terms first, second, third and similar in the description and in the claims are used to distinguish between similar elements and not necessarily to describe a sequential or chronological order. The terms are interchangeable under appropriate circumstances and the embodiments of the invention may operate in other sequences than those described or illustrated herein.

In addition, the various embodiments, although referred to as "preferred", should be interpreted as exemplary ways in which the invention can be implemented rather than as limiting the scope of the invention.

The term "comprising", as used in the claims, should not be construed as being limited to the means or steps listed below; it does not exclude other elements or steps. It must be interpreted as specifying the presence of the elements, integers, steps or components referred to but does not exclude the presence or addition of one or more other elements, integers, steps or components or groups of these. Therefore, the scope of the term "a device comprising A and B" should not be limited to devices comprising only components A and B, on the contrary, with respect to the present invention, the only listed components of the device are A and B, and the claim should also be interpreted to include equivalents of these components.

In the figures, identical or similar elements may be designated by the same number. .

In the following, aspects of the invention applied in an image compression method will be described. It should be noted that many of the steps or algorithms described can also be used for other applications, for example for text recognition or other. In addition, many modifications can be made to the steps and algorithms described without departing from the scope of the invention.

As used herein, the term "digital image" is meant to refer to a matrix of pixels where each pixel has a value that corresponds to its color in said image. A pixel may have an "empty" value which means that no color corresponds to said pixel in said image. In what follows, the words "value" and "color" are equivalent when they refer to a pixel. Digital images are stored in memory devices readable by electronic devices such as computers.

As used herein, the word "lineart" referring to an element is meant to refer to connected lines or curves placed against a background (usually plain), the lineart element having no gradation shade (blackness) or color tone (color). A black element on a white background is an example.

As used herein, the term "continuous tone" referring to an element is meant to refer to an element in which colors and grayscale melt smoothly in adjacent colors or greyscale levels. instead of producing distinct, clearly delimited areas of color or gray level.

As used herein, the term "connected component" in an image made of color pixels and empty pixels is meant to refer to a group of pixels connected by color pixels.

As used herein, the term "connected component" in a black and white image is meant to refer to a group of black pixels that are connected to each other by black pixels. A connected component can be called a "blobM.

As used herein, the term "image of a connected component" is meant to refer to a digital image that includes a single connected component.

The image compression method shown in Figure 1 is based on the MRC model. The MRC model is supposed to produce a good "mixed" image compression that contains your continuous (like a photo) and lineart (like solid lines, characters).

An original image 1 is taken as input by an initial image processing step 2, resulting in a full-size image 11 which is a digital image. The full-size image 11 is taken as input by a segmentation 100. The segmentation 100 results in three digital images: a foreground layer 12, a mask 13, and a background layer 14. The segmentation 100 uses as parameters a reduction factor of the foreground resolution, a reduction factor of the mask resolution and a reduction factor of the background resolution. The foreground layer 12 is compressed by a compression step 18 which uses as a parameter foreground quality 19 and generates a compressed foreground layer 28. The mask 13 is compressed by a compression step Which uses a mask quality 21 as a parameter and generates a compressed mask layer 30. The background layer 14 is compressed by a compression step 22 which uses a background quality 23 as a parameter and generates a Compressed background layer 32. After the compression of images 10, the three compressed images (foreground 28, mask 30 and background 32) are used in a subsequent image processing 24. The original image 1 may be a paper image or a digital image. It can come from an impression, a writing ..... it can include text, figures, photos, barcodes, ... The initial image processing 2 scans the original image 1 if it was not digital. It may include image processing. The full size image 11 resulting from the initial image processing 2 is stored in a memory device. The initial image processing 2 can use an imaging device such as a scanning digitizer, a camera, a mouse, a digital pen .....

The initial image processing 2 can use a computing device such as a computer, a smart phone (smartphone), a tablet, .... The full size image 11 is in color, grayscale ... .. The full size image 11 may contain photos and text with characters to be recognized at a later stage of the image processing. The segmentation step 100 is described in detail later in this document. The foreground layer 12 and the background layer 13 are in color, in grayscale, according to the full size image 11. The mask 13 is a binary image, which means that the pixels of the mask 13 take binary values (black or white, on or off, 1 or

The compression 18 of the foreground layer 12 can use a lossy continuous tone compression technique such as JPEG or JPEG-2000. The compression of the mask 13 may use a binary compression technique such as JBIG2 or TIFF G4. The compression 22 of the background layer 14 can use a lossy continuous tone compression technique such as JPEG or JPEG-2000. The use of a lossy continuous tone technique for the background and foreground layer provides high compression while maintaining the color gamut.

The subsequent image processing 24 typically comprises decompression of the three layers (foreground 28, mask 30 and background 32) and reconstruction of an image from the decompressed layers. The reconstruction selects the color of the pixels as follows. A black pixel, ie, lit or equal to 1, in the mask indicates that during the reconstruction, the value of said pixel is equal to the value of said pixel in the foreground layer 12. A white pixel in the mask 13 (ie off or having a value equal to 0) indicates that during the reconstruction, the value of said pixel is equal to the value of said pixel in the background layer 14.

The subsequent image processing 24 may include: segmentation, normalization, feature extraction, classification, character recognition, blob determination, image transfer, image storage, ...

Image compression 10 can be used to create a PDF document. The PDF document includes the three compressed images (foreground, mask, and background). When the PDF document is open, the reading software, Acrobat Reader for example, uncompresses the three compressed images and performs the reconstruction of images, during the opening of the document.

FIG. 2 illustrates the segmentation 100 of the full-size image 11 in three digital images: foreground layer 12, mask 13 and background layer 14. A binarization step 101 takes as input the full image size 11 and generates as output a black and white image 102 which is a digital image. A determination of the foreground elements 103 takes as input the black and white image 102 and possibly the full-size image 11 and generates as output an initial mask 104 which is a digital image. A step of reduction of the resolution (or reduction of scale) 105 takes as input the initial mask 104 and the reduction factor 16 of the mask resolution and generates the mask 13. A step of constructing the front layer plane 200 takes as inputs the full-size image 11 and the initial mask 104 and generates the foreground layer 12. A step of constructing the background layer 106 takes as inputs the full-size image 11 and the initial mask 104 and generates the background layer 14.

The binarization 101 may be performed by an adaptive local algorithm or other binarization algorithm to generate the black and white image 102 which is a binary version of the full-size image 11.

The foreground elements are then selected in the black and white image 102 in the determination step 103 to form the black pixels in the initial mask 104. Text elements such as characters and punctuation marks are typically foreground elements. In one embodiment of the invention, any non-continuous element is taken as a foreground element. As a result, text elements, solid lines, and table frames are foreground elements.

The non-foreground elements form the white pixels in the initial mask 14. The elements that are not foreground are background elements. The determination 103 may be performed by filtering that uses the minimum description length principle, which is known in the art. It is decided here for each set of black pixels and each set of white pixels if it is a foreground element or a background element. An example of this technique is described in US patent 8331706.

The mask 13 is generated from the reduction of the resolution 105 of the initial mask 104. The reduction step of the resolution 105 is optional. It improves the compression but deteriorates the resolution.

The construction of the foreground layer 200 is described in more detail later in this document. The construction of the background layer 106 may be performed by any technique known in the art. An example of this technique is described in US patent 8331706. ·

A construction of the foreground layer 200 is shown in Fig. 3. In order to distinguish between several methods of constructing the foreground layer, the construction of the foreground layer 200 illustrated in FIG. Fig. 3 is designated 200A. The full-size image 11 and the initial mask 104 are used as inputs in the selection step 201 of the pixels of the foreground elements in the full-size image 11, which generates a image of the foreground elements 202, which is a digital image, as an output. The initial mask 104 and a reduction factor 252 of the operating mask resolution are used as input to a resolution reduction step 251 which generates an operating mask 253, which is a binary digital image. The image of the foreground elements 202 and the operating mask 253 are then used as input to a determination step 300 of a transition image, which determines a transition image 203, which is a digital image. A fill step 204 is performed on the transition image 203 to generate the foreground layer 12. The step 201 of selecting pixels of the foreground elements in the full-size image 11 builds an image digital, the image of the foreground elements 202, where a pixel that is black in the initial mask 104 takes the same value as in the full-size image 11, and a pixel that is white in the initial mask 104 is established as "empty." The resolution reduction step 251 creates the operating mask 253 by reducing the resolution of the initial mask 104 by a factor n equal to the reduction factor 252 of the operating mask resolution. a pixel in the operating mask 253 corresponds to a first window of η xn pixels in the initial mask 104 and therefore in the full-size image 11, with n = 1, 2, ... In alternative embodiments, the first window used to generate the operating mask 253 from the initial mask 104 has a rectangular shape, a polygonal shape, a circular or elliptical shape, ... The determination step 300 of a transition image introduces a downscaling such that the image transition 203 resulting from step 300 has the same scale as the operating mask 253, which means that the transition image 203 is scaled down by a reduction factor of the resolution n with respect to the image of the elements foreground 202 and thus in the full-size image 11. The step of determining 300 of a transition image is described in more detail later in this document. The transition image 203 resulting from the step 300 has the same scale as the operating mask 253, which means that the transition image 203 is scaled down by a reduction factor of the resolution n relative to the image of the foreground elements 202. The reduction factor of the foreground resolution is therefore equal to the reduction factor 252 of the mask resolution operating in the foreground layer construction. 200A. Pixels that are white in the operating mask 253 are empty in the transition image 203.

In one embodiment of the invention, the selection step 201 of the foreground elements and the step of determining 300 of a transition image are combined in a single step. The transition image 203 is directly calculated from the full size image, the initial mask 104 and the operating mask 253 without calculating the image of the foreground elements 202 as such. Skipping the calculation and storage of this foreground image 202 reduces the memory required for the computer calculation. The filling step 204 fills the empty pixels using neighboring non-empty pixels, for example by a two-step sequential algorithm. Such an algorithm is described in US patent 8331706. The filling step 204 generates the foreground layer 12 as an output.

The construction of the foreground layer 200 according to an embodiment of the present invention is illustrated in FIG. 4. To differentiate between several methods of constructing the foreground layer, the construction of the layer foreground 200 shown in Figure 4 is designated 200B. The full-size image 11 and the initial mask 104 are used as inputs in the selection step 201 of the pixels of the foreground elements in the full-size image 11, which generates an image of before elements plan 202, which is a digital image, as an output. The initial mask 104 and a reduction factor 252 of the operating mask resolution are used as input to a resolution reduction step 251 which generates an operating mask 253, which is a binary digital image. The image of the foreground elements 202 and the operating mask 253 are then used as input to a determination step 300 of a transition image, which determines a transition image 203, which is a digital image. A fill step 204 is performed on the transition image 203 to generate the foreground layer 12. The step 201 of selecting pixels of the foreground elements in the full-size image 11 builds an image digital, the image of the foreground elements 202, where a pixel that is black in the initial mask 104 takes the same value as in the full-size image 11, and a pixel that is white in the initial mask 104 is established as "empty". The resolution reduction step 251 creates the operating mask 253 by reducing the resolution of the initial mask 104 by a factor n equal to the reduction factor 252 of the operating mask resolution. Consequently, a pixel in the operating mask 253 corresponds to a first window of η xn pixels in the initial mask 104, with n = 1, 2, ... In variant embodiments, the first window used to generate the operating mask 253 from the initial mask 104 has a rectangular shape, a polygonal shape, a circular or elliptical shape,. The step 300 of determining a transition image introduces a downscaling such that the transition image 203 resulting from step 300 has the same scale as the operating mask 253, which means that the transition image 203 is scaled down by a reduction factor of the resolution n with respect to the image of the foreground elements 202. The step of determining 300 of a transition image is described in more detail later in this document. The transition image 203 resulting from the step 300 has the same scale as the operating mask 253, which means that the transition image 203 is scaled down by a reduction factor of the resolution n relative to the image of the foreground elements 202. Pixels which are white in the operating mask 253 are empty in the transition image 203.

In one embodiment of the invention, the selection step 201 of the foreground elements and the step of determining 300 of a transition image are combined in a single step. The transition image 203 is directly calculated from the full-size image, the initial mask 104 and the operating mask 253 without calculating the image of the foreground elements 202 as such. Skipping the calculation and storage of this foreground image 202 reduces the memory required for the computer calculation.

The reduction of the resolution 205 reduces the resolution of the transition image 203. Since the foreground layer 12 is in total reduced scale compared to the full-size image 11 by the reduction factor of the resolution of the foreground 15 rfg, the reduction of the resolution 205 has a reduction factor 206 of the resolution equal to the reduction factor 15 of the resolution of the foreground divided by the reduction factor 252 of the resolution of the mask operating in the construction of the foreground layer 200B, In other words, the reduction of the resolution 205 reduces the scale of the transition image 203 by a factor 206 equal to rfg / n, rfg being greater to n. The reduction factor of the foreground resolution is therefore greater than the reduction factor 252 of the mask resolution operating in the construction of the foreground layer 200B. The filling step 204 fills the empty pixels using neighboring non-empty pixels, for example by a two-step sequential algorithm. The filling step 204 generates the foreground layer 12 as an output. The step 300 of determining a transition image according to an embodiment of the present invention is illustrated in FIG. 5. To differentiate between several embodiments of the present invention, the determination step 300 of a transition image of Figure 5 is designated 3Q0A. The step 3Q0A determination of a transition image begins with the creation 350 of an image 351 of empty pixels of the same size as the operating mask 253. A pixel in the image 351 of empty pixels corresponds to a single pixel in the operating mask 253. The determining step 300A of a transition image then comprises a loop 310 on the black pixels 301 of the operating mask 253. The loop 310 replaces, in the image 351 of empty pixels, each pixel that corresponds to a black pixel in the operating mask 253 by a color pixel 304.

Each black pixel 301 of the operating mask 253 is considered in turn. A color averaging step 302 uses as inputs a black pixel 301 of the operating mask 253, an averaging factor m 303 and the image of the foreground elements 202, and generates a color pixel 304 at the location of said pixel. The color pixel 304 replaces an empty pixel of the image 351 of empty pixels. Then, another black pixel 301 of the operating mask 253 is taken into consideration and so on for all the black pixels of the operating mask 253. Instead of a loop 310 on the black pixels 301 of the operating mask 253, said black pixels 301 could be considered in parallel. At the end of the loop, the transition image 203 is formed by the color pixels 304 which are located at the locations of the black pixels 301 of the operating mask 253 and by empty pixels which are located at the locations of the white pixels of the operating mask. 253 and which come from the 351 image of empty pixels. Therefore, the transition image 203 has the same scale as the operating mask 253. The color averaging 302 for a given black pixel 301 of the operating mask 253 is a color averaging of the colors of a pixel group in the image of the foreground elements 202, said pixels of said group have two characteristics. First, they do not have the value "empty" in the image of the foreground elements 202, which means that they correspond to the black pixels in the initial mask 104. Second, they are in a window of mxm pixels in the image of the foreground elements 202 which is centered on the given black pixel 301, where m, the color averaging factor 303, is higher than rfg, the reduction factor 15 of the resolution of the pixel. 'foreground.

Therefore, said black pixel 301 corresponds in terms of its location to a scaling window of rfg x rfg pixels in the full-size image 11 and is the color mean over an averaging window of mxm pixels in the full-size image 11 (where the pixels that are white in the initial mask 104 are not taken into account), with rfg <m. This method where the color averaging is performed on a window larger than the scaling window creates more homogeneous colors in the transition image 203, and thus in the foreground layer 12, a method where color averaging is performed on the scaling window.

The averaging window can be called "second window" as opposed to the "first window" which is the window of scaling.

If the centering of the averaging window on the location of said black pixel 301 is not possible, for example because rfg is equal to 2 and m is equal to 3, the center of the averaging window can be moved relative to in the center of the scaling window.

The averaging window can have any shape, any size, as long as it includes more pixels than the scaling window. If the operating mask 253 has the same scale as the color image 11 because no resolution reduction 251 has been performed, the color averaging 302 is performed on a window of at least two pixels of the image foreground elements 202. In one embodiment of the invention, the averaging window varies in size and / or shape according to a predetermined rule. For example, the averaging window may be larger in regions where pixel color variation is small and narrow in regions where the color varies greatly from pixel to pixel.

In one embodiment of the invention, the color averaging 302 is performed from the pixel values of an averaging window in the full-size image 11, taking into consideration only pixels that are black in the mask This provides the same transition image 203 as the method so far described which uses the foreground element image 202 without requiring this intermediate image of the foreground elements 202.

The color averaging 302 for the black pixel 301 provides the color 304 of said black pixel 301 in the transition image 203. Then, a next black pixel 301 is taken into consideration and its color 304 in the transition image 203 is provided. by the color averaging 302, until the color in the transition image of all the black pixels 301 of the operating mask 253 has been calculated.

In one embodiment of the invention, the averaging window does not have the same size for all the pixels of the transition image 203. In one embodiment of the invention, the averaging window is not no larger than the scaling window for some pixels of the transition image 203. In one embodiment of the invention, the averaging window is the same size as the scaling window. for some pixels of the transition image 203 and is larger than the scaling window for other pixels of the transition image 203,

In one embodiment of the invention, the color averaging 302 is replaced by another operation which determines a color from the colors of the pixels in the averaging window. This operation can be a determination of a median, a statistical mode, a geometric mean, a weighted average, ..

FIG. 6 provides an illustration of the determination of the transition image 300A described with reference to FIG. 5. FIG. 6 is one-dimensional, taking into account a pixel line, but the described method can be immediately generalized to FIG. case of a matrix of two-dimensional pixels. Line 601 provides the pixel index of a horizontal line. The index will be used later in this document. Line 602 provides the value of the pixels in an original grayscale image 1, the value of a pixel indicating the gray level of said pixel. Line 603 provides the pixel value determined by scanning the original greyscale image 1. Line 603 is the full-size image 11 referred to in the description of the present invention. Line 604 is the initial mask 104 determined from the full-size image 11 (line 603) by the binarization steps 101 and the determination of the foreground elements 103. The line 605 is the image of the d-elements. 206. The line 606 is the operating mask 253, determined from the initial mask 104 by the reduction of the resolution 251. The line 607 is the transition image 203 resulting from the determination of the transition image. 300A according to an embodiment of the present invention. Line 608 is the transition image 203 resulting from a known method of determining the transition image. Line 609 provides the index of pixels at a reduced scale on the horizontal line.

A higher pixel value in lines 602, 603, 605, 607, and 608 indicates a gray level close to white, and a lower value indicates a gray level close to black. The original image includes a first clear part (pixels a and b), a dark part (pixels c to g), a pixel between light and dark (pixel h) and a second clear part (pixels i and j). The full size image (line 603) is not exactly equal to the original image (line 602) because the color at the edge between the light and dark portions (pixels c and g) is a mixture between the parts dark and light, this may be due to printing, scanning, or other physical or digital image processing. The initial mask (line 604) is determined by considering that all the pixels with a value lower than 100 are black (value = 1) and the others are white (value = 0). The image of the foreground elements 202 (line 605) is determined from the full-size image 11 (line 603) and the initial mask (line 604) by setting the white pixels of the initial mask as empty. assigning the black pixels of the initial mask their value in the full size image (line 603). The operating mask (line 606) is determined from the initial mask (line 604) by scaling by a reduction factor of resolution n equal to 2. The transition image resulting from the image determination 300A transition device according to one embodiment of the present invention (line 607) has the same scale as the operating mask (line 606). The pixels ab and ij which are white (value = 0) in the mask (line 606) are set as empty in the transition image (line 607). The value of the pixels cd, ef and gh which are black (value = 1) in the operating mask (line 606) is determined by an average of the values of the non-empty pixels of the image of the foreground elements (line 605). ) that are not empty and are in a window of four (m = 4) pixels centered on the black pixels of the operating mask. More specifically, the value of pixel cd in line 607 is the average of c, d and e of line 605; the value of the pixel ef in the line 607 is the average of d, e, f and g of the line 605; and the value of the pixel gh in the line 607 is the average of f and g of the line 605. The transition image resulting from a known method of determining the transition image (line 608) has the same scale as the operating mask (line 606). The pixels ab and ij which are white (value = 0) in the mask (line 606) are set as empty in the transition image (line 608). The value of the pixels cd, ef and gh which are black (value = 1) in the operating mask (line 606) is determined by an average of the values of the non-empty pixels of the image of the foreground elements (line 605). ) using the same window for averaging as for scaling. More specifically, the value of pixels cd in line 608 is the average of c and d of line 605; the value of the pixel ef in line 608 is the average of e and f of line 605; and the value of the pixel gh in the line 608 is equal to the value of the pixel g of the line 605. It appears that the values in the line 607 are more homogeneous than the values in the line 608.

FIG. 7 provides a second illustration of the determination of the transition image 300A described with reference to FIG. 5. A matrix 701 provides the value of the pixels in an original grayscale image 1, the value of one pixel indicating the gray level of said pixel. The original image represents a first character 702.and a second character 703 which are darker (less number in the gray scale) than the remainder of said original image. Matrix 704 provides the pixel value determined by scanning the original greyscale image 1. Matrix 704 is the full-size image 11 to which reference is made in the description of the present invention. The two characters are called 705 and 706 in 704. A matrix 707 is the initial mask 104 determined from the full-size image 11 (matrix 704) by the binarization steps 101 and the determination of the foreground elements 103 The two characters are called 708 and 709 in 707. A matrix 710 is the image of the foreground elements 202. The two characters are called 711 and 712 in 710. A matrix 713 is the operating mask 253, determined at from the initial mask 104 by the reduction of the resolution 251. The two characters are called 714 and 715 in 713. A matrix 716 is the transition image 203 resulting from the determination of the transition image 300A according to one embodiment. of the present invention. The two characters 717 and 718 are present in the matrix 716. One of the pixels of the matrix 716 has received a reference number 719. A matrix 722 is the image of the foreground elements 202 where the setting window is indicated. at the 723 scale and the averaging window 724 corresponding to the pixel 719 in the transition image 203 (matrix 716). The full size image (matrix 704) is not exactly equal to the original image (matrix 701), mainly because the color at the edge between the light and dark parts is a mixture between the dark and light parts. The initial mask (matrix 707) is determined by considering that all the pixels with a value lower than 100 are black and the others are white. The image of the foreground elements 202 (matrix 710) is determined from the full size image (matrix 704) and the initial mask (matrix 707) by setting the white pixels of the initial mask as empty and assigning to the black pixels of the initial mask their value in the full-size image (array 704). The operating mask (matrix 713) is determined from the initial mask (matrix 707) by scaling by a reduction factor of resolution n equal to 2. Different scaling formulas are possible. The preferred scaling method is to consider a pixel resulting from scaling equal to 1 if at least one pixel in the scaling window equals 1. The transition image (Matrix 716) resulting from the determination of the transition image 300A according to one embodiment of the present invention on the same scale as the operating mask (array 713). The white pixels of the operating mask (array 713) are empty pixels in the transition image (array 716). The black pixels of the operating mask (matrix 713) are calculated from an average of a window of 16 (four x four) non-empty pixels (m = 4) of the image of the foreground elements centered on the black pixels of the mask, the position of the pixel 719 corresponds to the window 723 in the image of the foreground elements (matrix 722) and the value (or color) of the pixel 719 is calculated as the average relating to the values of the pixels. non-empty pixels of window 724. The step 300 of determining a transition image according to another embodiment of the present invention is illustrated in FIG. 8, to differentiate between several embodiments of the present invention. the step of determining 300 of a transition image of FIG. 8 is designated 3Q0B. The determination step 300B of a transition image starts from the image of the foreground elements 202. Images 402 of connected components are calculated from the image of the foreground elements 202 in step 401. Corresponding images 404 of components connected in the operating mask are calculated from the operating mask 253 at step 403. The determining step 300B of a transition image comprises a loop 450 on the images 404 of connected components in the operating mask that includes a series of 400 completed steps for any connected component 404 in the operating mask. The series 400 of steps begins with the creation 490 of an image 491 of empty pixels for the connected component considered. The image 491 is the same size as the associated component image 404 connected in the operating mask 253. A pixel in the empty pixel image 491 corresponds to a single pixel in the associated component image 404 connected in the mask operating 253.

The series 400 of steps comprises a loop 420 on the black pixels of the image 404 considered component connected in the operating mask 253. The loop 420 replaces, in the image 491 of empty pixels for the connected component considered, any pixel which is black in the image 404 of the component connected in the operating mask 253 by a color pixel 408. Each black pixel 405 of the considered image 404 is considered in turn. A color averaging step 406 uses as inputs said black pixel 405, a color averaging factor m 407 and the corresponding connected component image 402 in the foreground image, and generates the color pixel. 408 in the connected component considered in the transition image at the location corresponding to the black pixel 404. The color pixel 408 replaces an empty pixel of the image 491 of empty pixels. Then, another black pixel 405 of the considered image 404 of the connected component is taken into consideration and so on for all the black pixels of the considered image 404 of the connected component. At the end of the loop 420, a connected component 412 in the transition image 203 is formed by the color pixels 408 which are located at the locations of the black pixels 405 of the corresponding connected component 404 in the operating mask 253 and by pixels blanks which are located at the white pixel locations of the corresponding connected component 404 in the operating mask 253 and come from the empty image 491. Therefore, the connected component 412 in the transition image 203 has the same scale as the component connected 404 corresponding in the operating mask 253.

Then, another connected component 404 in the operating mask is considered in the loop 450 with the creation 490 of an image 491 of empty pixels for the connected component considered and a loop 420 on its black pixels. At the end of the determination step 300B of a transition image, all the connected components 412 of the transition image form the transition image 203.

Due to the way in which it was constructed at the selection step 201 of the pixels of the foreground elements in the full size image, the image of the foreground elements 202 is made of color pixels and empty pixels. In step 401 of creating component images connected from the foreground element image 202, the connected components of said image 202 are cut into individual digital images 402 which are smaller images than said image 202. Any image 402 of a connected component contains a single connected component. For example, a connected component 402 in the foreground image may comprise a textual element, a character, a lineart element, a letter, a symbol, or a punctuation mark.

In one embodiment of the invention, an image 402 of a connected component corresponds to a bounding rectangle of the connected component, i.e., the smallest rectangle that includes the connected component. In one embodiment of the invention, the upper-left coordinate of the bounding rectangle is retained to match the image of the foreground elements 202. The enclosing rectangles of two connected components may overlap. For example, if the image represents the letters "VA", the bounding box of 'V' can overlap the bounding box of 'A'. In step 403 of creating images 404 of connected components in the operating mask 253, each connected component image 404 of the operating mask is selected to correspond to a connected component image 402 in the image of the precursor elements. plan. A connected component image 402 in the image of the foreground elements and the corresponding image of the connected component 404 in the mask corresponds to the same set of pixels in the full-size image 11.

The color averaging 406 for a black pixel 405 of the connected component image 404 of the mask is a color averaging of the colors of a group of pixels of the connected component image 402 in the image of the elements. foreground which corresponds to the connected component image 404 in the operating mask, said pixels of said group having two characteristics. First, they do not have the value "empty" in the connected component image 402 in the foreground image, which means that they correspond to the black pixels in the initial mask 104. Second, they are in a window of mxm pixels in the connected component image 402 in the image of the foreground elements which is centered on the given black pixel 405, where m, the color averaging factor, is greater than n, the reduction factor of the resolution of the operating mask. If the window goes beyond the pixels of the connected component image 402, that portion beyond the pixels of the connected component image is not taken into account in the averaging 406.

Therefore, said black pixel 405 corresponds as to its location to a scaling window of mxm pixels in the full-size image 11, but is the color average over a window of mxm pixels in a single connected component image of the full-size image 11 (where the pixels that are white in the initial mask 104 are not considered). This method where the color averaging is performed on a window larger than the scaling window creates more homogeneous colors in the transition image 203, and thus in the foreground layer 12, a method where color averaging is performed on the scaling window. Due to the connected component image separation, pixel colors of different connected components can not mix during averaging of colors 406. If a blue character is followed by a red character, blue and red do not mix not.

If the centering of the averaging window on the location of said black pixel 405 is not possible, for example because n is equal to 2 and m is equal to 3, the center of the averaging window can be moved relative to in the center of the scaling window.

The averaging window can have any shape, any size, as long as it includes more pixels than the scaling window and pixels of a single connected component. If the operating mask 253 has the same scale as the full-size image 11, the color averaging 406 is performed on a window of at least two pixels in the connected component image 402 of the image of the before elements In one embodiment of the invention, the averaging window varies in size and shape according to a predetermined rule. For example, the averaging window may be large for a connected component corresponding to a large and narrow line for a connected component corresponding to a narrow line.

Color averaging 406 for black pixel 405 provides a color pixel 408 at the location of said black pixel 405 in the connected component image 412 in the transition image. Then a next black pixel 405 is considered and the corresponding color pixel 408 in the connected component image 412 in the transition image is provided by the averaging 406 until the color of all the black pixels 301 of the connected component image 404 of the mask has been calculated.

The color pixels 408 and the empty pixels set in step 490 form the connected component image 412 in the transition image. Then the next connected component 404 of the operating mask is considered in the loop 450. Therefore, the connected component image 412 in the transition image has the same scale as the connected component image 404 in the mask and the transition image 203 has the same scale as the operating mask 253.

In one embodiment of the invention, the color averaging 406 is performed from the pixel values of an averaging window in the full-size image 11, taking into consideration only pixels that are black in the image. initial mask 104. This provides the same transition image 203 as the method heretofore described which uses the foreground element image 202 without requiring this intermediate image of the foreground elements 202.

Fig. 9 provides an illustration of the determination of the transition image 300B described with reference to Fig. 8.

A matrix 901 supplies the value of the pixels in an original grayscale image 1, the value of a pixel indicating the gray level of said pixel. The original image represents a first character 902 and a second character 903 which are darker (less in the gray scale) than the background of said original image. A matrix 904 provides the pixel value determined by scanning the original grayscale image 1. Matrix 904 is the full-size image 11 referred to in the description of the present invention. The two characters are called 905 and 906 in 904. The set of pixels in columns 1 to 6 has been assigned reference number 960 and the set of pixels in columns 7 to 10 has been assigned reference number 970. A matrix 907 is the initial mask 104 determined from the full size image 11 (matrix 904) by the binarization steps 101 and the determination of the foreground elements 103. The two characters are referred to as 908 and 909 in 907. A matrix 910 is the image of the foreground elements 202. The two characters are called 911 and 912 in 910. A matrix 913 is the operating mask 253, determined from the initial mask 104 by the reduction of the resolution 105. two characters are called 914 and 915 in 913,

Two matrices, 917 and 919, illustrate the connected component images 402 in the foreground image. The matrix 917 comprises a character 918. The matrix 919 comprises a character 920. Two matrices, 922 and 924, illustrate the images of connected components 404 in the mask. The matrix 922 comprises a character 923. The matrix 924 comprises a character 925. Two matrices, 927 and 929, illustrate the connected component images 412 in the transition image. The matrix 927 comprises a character 928. The matrix 929 comprises a character 930. One of the pixels of the matrix 927 has received a reference number 931. A matrix 932 is the transition image 203 resulting from the determination of the image of transition 300B according to an embodiment of the present invention. The two characters 933 and 934 are present in the matrix 932. Two matrices, 936 and 937, show the connected component images 402 in the foreground image that are indicated in the scaling window. 938 and the averaging window 939 corresponding to the pixel 931 of the corresponding connected component image 412 of the transition image (matrix 927). The full size image (matrix 904) is not exactly equal to the original image (matrix 901), mainly because the color at the edge between the light parts and the dark part is a mixture between the dark and light parts. The initial mask (matrix 907) is determined by considering that all the pixels with a value lower than 100 are black and the others are white. The image of the foreground elements 202 (matrix 910) is determined from the full size image (matrix 904) and the initial mask (matrix 907) by setting the white pixels of the initial mask as empty and assigning to the black pixels of the initial mask their value in the full size image (matrix 904). The mask (matrix 913) is determined from the initial mask (matrix 907) by scaling by a reduction factor of resolution n equal to 2.

The connected component images 402 in the foreground element image (matrices 917 and 919) are pixel arrays determined such that any connected component corresponds to an image and each image corresponds to each connected component. '

The character 918 is the connected component of the image 917 and corresponds to the line 905 in the full size image (matrix 904) and the character 920 is the connected component of the image 919 and corresponds to the line 906 of the full image size (matrix 904). The connected component images 404 of the mask (matrices 922 and 924) correspond to the connected component images 402 in the foreground element image (matrices 917 and 919). 917 and 922 correspond to the same set 960 pixels in the full size image 11 (matrix 904) and 918 and 923 therefore have a base in the same line 905 in the full size image 11. 919 and 924 correspond to the same set 970 pixels in the full-size image 11 (matrix 904) and 920 and 925 therefore have a base in the same line 906 in the full-size image 11,

The connected component images 412 in the transition image (matrices 927 and 929) have the same scale as the component images connected in the mask (matrices 922 and 924). The white pixels of the component images connected in the mask (matrices 922 and 924) are empty pixels in the connected component images 412 in the transition image (matrices 927 and 929). Each black pixel of a connected component image of the mask (matrices 922 and 924) is computed from an average of the non-empty pixels of a window of 16 (four out of four) pixels (m = 4) of the corresponding connected component image in the image of the foreground elements (matrices 917 and 919) centered on the black pixels, the part of the window outside the matrix not being taken into account in the establishment of the average. For example, the position of the pixel 931 corresponds to the window 938 in the component images connected in the foreground element image (matrix 936) and the value (or color) of the pixel 931 is calculated as the average of the pixels. values of the non-empty pixels of the averaging window 939. The pixels of the other connected component image (matrix 937) are not taken into account in the calculation of the value of the pixel 931. The determination step 300 of a transition image according to another embodiment of the present invention is illustrated in FIG. 10. In order to distinguish between several embodiments of the present invention, the step 300 of determining a transition image of FIG. 10 is designated 30Q, The step 300C of determining a transition image starts from the image of the foreground elements 202. Images 402 of connected components are calculated from the image of the elements. foreground 202 at step 401. Corresponding images 404 of components connected in the operating mask are calculated from the operating mask 253 at step 403. The determining step 300C of a transition image comprises a loop 550 on the Connected component images 402. Loop 550 includes a verification step 501 that checks whether the colors in the connected component image 402 taken into consideration are significantly different. If so, the same series 400 of steps as in the step 300B determination of a transition image according to an embodiment of the invention described above is performed on the pixels of the connected component image 404 of the mask which corresponds to the connected component image 402 considered for determining the color of the pixels of an image of the connected component 412 considered in the transition image. If not, a series of 500 steps is performed.

The 500 series of steps includes a color averaging 502 which uses as input the connected component image 402 considered in the foreground image and provides an average color 503.

The 500 series of steps begins with the creation 590 of an image 591 of empty pixels for the connected component under consideration. The image 591 is the same size as the connected component image 404 considered in the operating mask 253. A pixel in the empty pixel image 591 corresponds to a single pixel in the connected component image 404 considered in the mask operating 253.

The 500 series of steps comprises a loop 560 on the black pixels of the connected component image 404 considered in the mask that corresponds to the connected component image 402 considered in the foreground image. The loop 560 replaces, in the image 591 of empty pixels for the connected component considered, any pixel that is black in the image 404 of the component connected in the operating mask 253 by a color pixel 506.

Each black pixel 504 of the connected component image 404 in the mask is considered in turn. A step of setting the value of the pixel equal to the average color 505 uses as inputs the black pixel 504 of the connected component image 404 in the mask that corresponds to the connected component image 402 considered in the image of the foreground elements and the middle color 503, and generates a color pixel 506 at the pixel location 504 in the connected component image 507 in the transition image. The color pixel 506 replaces an empty pixel of the image 591 of empty pixels. Then, another black pixel 504 of the connected component image 404 in the mask that corresponds to the connected component image 402 considered in the foreground image is taken into consideration and so on for all the black pixels of the connected component image 404 in the mask corresponding to the connected component image 402 considered in the foreground image. At the end of the loop 560, a connected component 507 in the transition image 203 is formed by the color pixels 506 which are located at the locations of the black pixels 504 of the corresponding connected component 404 in the operating mask 253 and by pixels blanks which are located at the white pixel locations of the corresponding connected component 404 in the operating mask 253 and come from the empty image 591. Therefore, the connected component 412 in the transition image 203 has the same scale as the component connected 404 corresponding in the operating mask 253.

After either the 400 series of steps or the 500 series of steps, another connected component image 402 in the image of the foreground elements is considered in the loop 550 and the verification step 501 is performed. At the end of the determination step 300C of a transition image, all connected component images 412 (generated in the 400 series of steps) and 507 in the transition image form the transition image 203.

For each connected component image 402 in the foreground element image, the verification step 501 checks whether said connected component image 402 in the foreground image comprises color pixels clearly. different. One method for performing such verification is to determine the standard deviation of the color of the non-empty pixels of the connected component image 402 and to consider that the colors are significantly different if said standard deviation is larger than a threshold. If the colors of a connected component image 402 are not significantly different, an overall color average over the entire connected component (not taking into account empty pixels) provides a satisfactory estimate of all the colors of the component. connected component image while leading to the possibility of a higher compression ratio than local averages because a region with the same color values requires fewer bits for compression than when the values are different. This corresponds to the 500 series of steps that results in 507 connected single-color components in the transition image. If the colors of a connected component image 402 are significantly different, averaging over the entire connected component image would not provide a satisfactory estimate of the colors of the connected component and would not provide at the end of has a good foreground layer 12. It is therefore preferable to perform a local average of the colors. This corresponds to the 400 series of steps.

The color averaging 502 for a connected component image 402 given in the foreground image is a color averaging of the colors of all the non-empty pixels of the connected component image 402 in the image. image of the foreground elements that corresponds to the connected component image 404 in the mask. Said pixels correspond to the black pixels in the initial mask 104 at the location of said connected component image. The color averaging 502 and the color setting step 505 of the pixels being equal to the mean 503 are equivalent to assuming that the averaging window of all the pixels 504 is the connected component image 402 complete in the image of the foreground elements. The color averaging 502 on the connected component image 402 complete provides the average color 503 that is assigned in step 505 to any pixel 504 of the connected component image that is black in the operating mask 253. Any pixel 409 of the connected component image that is white in the operating mask 253 is set as empty in step 410. The collection of color pixels 506 and empty pixels 411 form the connected component image 507 in the transition image if the 500 series of steps is used. The connected component image 412 provided by the 400 series of steps, which may include multiple colors, and the image of connected components 507 with connected single-color components provided by the 500 series of steps form the transition image 203 provided by the 300C determination of the transition image.

The present document comprises the description of two algorithms or methods 200A and 200B for constructing the foreground layer, each of them including a step of determining 300 of a transition image. This document also includes the description of three transition image determination algorithms or method 300A, 300B and 300C. Embodiments of the present invention include the combination of any of methods 200A and 200B with any of methods 300A, 300B and 300C,

Claims (15)

claims A method for creating, from an initial digital image (11), a foreground image (12) to be used in image compression, comprising the steps of a / creating an image of binary mask (253) from the initial digital image (11), each pixel of the binary mask image (253) being created from a corresponding first pixel window in the initial digital image (11) , and said bitmap image (253) comprising a first set of pixels which are of a first value and a second set of pixels which are of a second value; bf generating a transition image (203) that has an amount of pixels equal to the amount of pixels of the bitmask image (253) by determining for each pixel of the transition image (203) corresponding to a pixel of the first set of pixels in the bit mask image (253) a corresponding second pixel window in the initial digital image (11), the second pixel window comprising more pixels than the first pixel window, determining a color for each pixel of the transition image (203) corresponding to a pixel of the first set of pixels in the bitmask image (253) by a calculation based on the colors of the pixels in the corresponding second window, and establishing each a pixel of the transition image (203) corresponding to a pixel of the second set of pixels in the binary mask image (253) being empty; and cl creating the foreground image (12) from the transition image (203) by keeping the non-empty pixels color and determining a color for the empty pixels using the non-empty pixels neighbors. The method of claim 1, wherein the first pixel window comprises a pixel. 3. Method according to claim 1, wherein the first pixel window comprises a window η x n of pixels, n being at least 2. The method of any of the preceding claims, wherein the color-based calculation of the pixels in the corresponding second window is averaging. A method according to any one of the preceding claims, wherein the location of the center of a pixel in the transition image corresponds to the location of the center of the corresponding second pixel window in the initial digital image, The method of any of the preceding claims, further comprising a step (205) of reducing the resolution of the transition image (203) by step b) and step c). A method according to any one of the preceding claims, wherein in step c) the color of the empty pixels is determined from the non-empty pixels using a two-step sequential algorithm. The method of any of the preceding claims, further comprising the step of creating from the initial digital image (11) groups (402) of pixels, wherein the color of each pixel in the image. transition (203) is determined independently for each group (402) of pixels. The method of claim 8, further comprising the step of clipping the bitmap image (253) into groups (404) of pixels corresponding to the groups (402) of pixels in the initial digital image (11). . The method of claim 8 or 9, wherein each group (402) of pixels in the initial image (11) contains at least one of: a single connected component, a single textual element, a single lineart element or a single element character. The method according to any one of claims 8 to 10, further comprising the steps of detecting whether, in a group (402) of pixels in the initial digital image (11), the pixels corresponding to the pixels of the first set of pixels of the binary mask image have colors that are distinctly different, and, to be determined, if the pixels in the initial digital image (11) do not have significantly different colors, the average color (503) of pixels in the initial digital image (11) and to apply the average color (503) on each pixel in the transition image (203) corresponding to the group of pixels in the initial digital image (11) and corresponding to the pixels in the first set of pixels of the bitmap image (253). A compression method (10) for compressing an initial digital image (11) of a scanned document, said compression method (10) comprises the step of segmenting the initial digital image (11) into multiple layers image comprising a foreground image (12) containing color information for the foreground elements of said document, a background image (14) containing color information for elements of the foreground background of said document and a binary image (13) for selecting between pixels in said foreground image (12) and said background image (14), wherein said foreground image (12) ) is created according to any of the methods of claims 1 to 11. A computer-readable non-temporary medium storing a program causing a computer to execute a method according to any one of the methods of claims 1 to 12. A system comprising a processor and a non-temporary computer readable medium storing a program causing the processor to execute any of the methods of claims 1 to 12. The system of claim 14 further comprising an imaging device for capturing the initial digital image.