FR3138551A1 - Process for recording a digital image - Google Patents

Abstract

The invention relates to a method for recording a digital image in a computer format, said image or at least part of the image being encoded and recorded in the form of a matrix of pixels in which each pixel considered (2 ) among at least one or more pixels of the matrix comprises: - at least one separation line (3) within the pixel considered (2), separating the pixel considered (2) at least into two areas (31, 32), and/or capable of making this line visible and/or - at least one spatial gradient (4) of color and/or hue and/or luminosity within the pixel considered (2) or at least one of the domains (31 , 32) of the pixel considered (2). Figure for abstract: Fig. 1b

Description

Process for recording a digital image

The present invention relates to a method of recording a digital image in a computer format, as well as a method of displaying, pixelating, rasterizing, compressing and/or decompressing such an image and a program product computer comprising computer code instructions for carrying out any of these methods.

The field of the invention is more particularly but not limited to that of images acquired by a digital sensor, within a smartphone or touchscreen tablet or webcam or digital camera or digital reflex or bridge camera, or other.

State of the prior art

We know the image formats known as represented by pixels, or bitmap, or raster image, of which a pixel is illustrated in .

We know the so-called vectorized or vector image formats.

A vector image is composed of geometric objects such as straight segments, arcs of circles, or more generally curves, which define lines of a certain thickness, as well as surfaces of various shapes delimited in a general way by polygons, or by other shapes resulting from curves.

These objects can be superimposed. In this case, an order of representation is defined to know which object is represented first and partially hidden by another, and so on. The advantage of this representation is that in the event of enlargement, it is possible to maintain a precise definition, for example of the contours of objects, or of filled areas, which is not possible with so-called matrix images for which the dimensions are perceived. pixels beyond a certain enlargement.

Images represented in matrix format therefore do not tolerate enlargement. Indeed, the latter reveals the drawing of pixels which become large in size when the image is enlarged. These enlarged images are unpleasant to look at because they contain a mosaic of pixels. In addition, pixel vision can cause you to lose perception of the information contained in the image, or blur it. We can convince ourselves of this by enlarging an image of writing to the pixel resolution limit. By enlarging it, we often lose the meaning of what is written.

Vectorized images generally have an “artificial” character, too perfect. Indeed, they can result from drawings of primitives such as the drawing of a rectangle, a circle, or even a closed polygon filled with a certain color. Images of the real world almost systematically include gradients in luminosity linked, among other things, to their variable inclinations in relation to the light sources, and to the non-homogeneity of these sources. In addition, objects in the real world are very often covered with defects such as scratches, dust, more or less fine, discolorations linked to light, weave of fabrics, veins of tree leaves, strands of rope threads, hair, etc. And living beings and animals generally have more roughness than objects not coming directly from nature, such as wrinkles, hair, hairs, redness, wood veins... which are all very different elements from a pure geometric primitive. , except to give it a very reduced size, which can multiply its number to the extreme.

Also certain vectorized image languages include the possibility of adding filters, such as the VML language, which have the effect of blurring the edges of the images, making them less sharp and a little more consistent with real images. They also make it possible to apply color gradients, making these primitives less uniform in their texture, which corresponds to notions of color gradients added to the properties of a surface delimited by an object in order to better represent the effects of reality .

We therefore notice effects of vectorization, which for example cause the details of the light bands rolled up on a hat to be lost, which contain a weft or gather of the fabric perpendicular to the bands, which disappears with vectorization. Likewise, fur loses its wire details, becoming only continuous masses, the same goes for hair. Note that despite these losses, the vectorized image can contain a notion of color gradient. The presence of these gradients is not enough to restore the notions of hair and the details of the strips of fabric. This example confirms that vectorizing an image does not generally make it possible to reproduce all its nuances, unless you use extremely fine objects, for example a curve for each hair. In the latter case, we would end up with an extreme quantity of objects, making the representation heavier.

Therefore, despite the improvements in taking gradients into account, the images obtained by vectorization generally retain a perceptible artificial character, unless a large number of objects are used, or many effects are applied, as can be done in software rendering images. digital games, or digital synthesis of film images, at the cost of numerous calculations.

Vectorized images are therefore not a format suitable for the representation of photographs, which by nature contain an extreme diversity of textures, shadows, gradients and micro-defects. Also, software tools exporting images from camera modules (for taking single photographic images, or multiple images constituting a film), work intrinsically on fields or matrices of pixels, and compress them in formats compatible with this pixel field representation (such as .jpeg, .png), to reduce the file size of these images or videos.

But these raster images do not support enlargement, unless they significantly lose quality during visualization, on contours, lines, textures with significant spatial variation, unless they are initially generated with very fine spatial sampling in accordance with the maximum predictable enlargement, which generates a very large number of pixels and very large associated files, even with compression.

Finally, in this state of the art, we will notice that the raster image is potentially much faster to restore because it is enough to go through the data table of the pixels of the image in a double loop. For a vectorized image, it is appropriate to go through each object in a list of objects in steps of the desired restitution density, for example to create the output of this image towards a matrix representation compatible with a display screen, therefore according to the pitch of these pixels. If the number of objects is very high, to represent all the richness of a real image, for example to obtain details such as texture, these calculations can represent a much greater quantity of calculations than raster image processing, at a given quality of restitution.

• De diminuer l’impression de caractère artificiel de l’image, et/ou
• D’être plus adapté à la représentation de photographies et/ou d’images comportant une extrême diversité de textures, d’ombres, de dégradés, de micro-défauts, et/ou
• De mieux supporter l’agrandissement, sans perdre sensiblement en qualité lors de la visualisation, sur les contours, les lignes, les textures à variation spatiale significative, et/ou
• De faciliter des traitements de compression/décompression, ou affichage, et/ou
• de nécessiter moins de données, à qualité équivalente, et/ou
• d’avoir une meilleure qualité, à quantité de données équivalente, et/ou
• de diminuer son temps de traitement.

The aim of the present invention is to propose a method of recording a digital image in a computer format, allowing, compared to the state of the art:

• To reduce the impression of artificiality of the image, and/or
• To be more suitable for the representation of photographs and/or images comprising an extreme diversity of textures, shadows, gradients, micro-defects, and/or
• To better support enlargement, without significantly losing quality during visualization, on contours, lines, textures with significant spatial variation, and/or
• To facilitate compression/decompression treatments, or display, and/or
• to require less data, of equivalent quality, and/or
• to have better quality, for an equivalent quantity of data, and/or
• to reduce processing time.

• au moins une ligne de séparation au sein du pixel considéré, séparant le pixel considéré au moins en deux domaines, et/ou
• au moins un gradient spatial de couleur et/ou de teinte et/ou de luminosité au sein du pixel considéré ou d’au moins un des domaines du pixel considéré.

This objective is achieved with a method (typically implemented by computer) of recording a digital image in a computer format, said image or at least part of the image being encoded and recorded in the form of a matrix of pixels in which each pixel considered among at least one or more pixels of the matrix comprises:

• at least one separation line within the pixel considered, separating the pixel considered at least into two areas, and/or
• at least one spatial gradient of color and/or hue and/or luminosity within the pixel considered or at least one of the domains of the pixel considered.

• Deux points sur un pourtour du pixel considéré, et/ou
• Une distance, de préférence par rapport au centre du pixel considéré, et un angle.

Each separation line within the pixel considered can be coded by two pieces of data including:

• Two points on the edge of the pixel considered, and/or
• A distance, preferably relative to the center of the pixel considered, and an angle.

Separation lines of several neighboring pixels can be assembled so as to form a continuous line crossing these several neighboring pixels.

• Un gradient de luminosité, et/ou
• Un ou plusieurs gradients correspondant respectivement à une ou plusieurs couleur(s) et/ou teinte(s).

The at least one gradient within the pixel considered or at least one of the domains of the pixel considered may include:

• A brightness gradient, and/or
• One or more gradients corresponding respectively to one or more color(s) and/or shade(s).

The at least one gradient within the pixel considered or at least one of the domains of the pixel considered can be coded by:

• Une seule composante spatiale par rapport à l’au moins une ligne de séparation au sein du pixel considéré.

– two spatial components, or

• A single spatial component relative to the at least one separation line within the pixel considered.

At least one brightness and/or color and/or hue within the pixel considered or at least one of the domains of the pixel considered may depend on a brightness and/or color and/or hue of at least one neighboring pixel of the pixel considered or at least one domain of a pixel neighboring the pixel considered.

At least one gradient of brightness and/or color and/or hue within the pixel considered or at least one of the domains of the pixel considered may depend on a gradient of brightness and/or color and/or hue of at least one pixel neighboring the pixel considered or of at least one domain of a pixel neighboring the pixel considered.

All pixels in the pixel array can have the same dimensions.

The at least one separation line may have a viewable state with a viewing surface.

The at least one separation line may itself include a spatial gradient making it possible to modulate, when it is visible or visualized, its brightness(es) and/or colors and/or hues.

The at least one dividing line can be switched between its viewable state and a non-viewable state.

The at least one separation line can be switched between an active state and an inactive state.

The method according to the invention may include deactivation of the at least one separation line of the pixel considered if this pixel considered is sampled below a certain spatial sampling threshold.

The method according to the invention can be implemented embedded in a software module which is added to or integrated into a Web browser, for taking into account the recorded images, possibly their compression or decompression, and/or their visualization in part or all of a screen interacting with the web browser.

In certain embodiments:

- textual information represented by pointing symbols in a character font, and/or

- drawings of objects embedded in the form of raster or vectorized sub-images

can be added directly to the recorded image or added when viewing it, in order to be viewable, possibly according to a particular selection which only shows all or part of it, when the recorded image is played back.

Metadata can be added to the saved image.

According to yet another aspect of the invention, a method is proposed:

- displaying an initial image in a so-called final image on a screen comprising groups of photogenerators, or

- pixelation or rasterization of said initial image into a final image comprising final pixels devoid of separation line and/or spatial gradient of color and/or hue and/or brightness within these final pixels,

said initial image being recorded according to a recording method according to the invention,

• Pour chaque pixel final considéré ou chaque groupe de photogénérateurs considéré, une transcription d’au moins pixel initial ou d’au moins un domaine de pixel initial en ce pixel final ou groupe de photogénérateurs considéré:
  • à partir des composantes de luminosité et/ou couleur et/ou teinte de cet au moins un pixel initial ou au moins un domaine de pixel initial
  • vers ce pixel final ou groupe de photogénérateurs considéré,

characterized in that it includes:

• For each final pixel considered or each group of photogenerators considered, a transcription of at least initial pixel or at least one initial pixel domain into this final pixel or group of photogenerators considered:
  • from the brightness and/or color and/or hue components of this at least one initial pixel or at least one initial pixel domain
  • towards this final pixel or group of photogenerators considered,

• cet au moins pixel initial ou au moins un domaine de pixel initial et
• le pixel final ou groupe de photogénérateur considéré.

the transcription being carried out in proportion to this at least initial pixel or at least one initial pixel domain within a common surface facing between:

• this at least initial pixel or at least one initial pixel domain and
• the final pixel or photogenerator group considered.

For transcription, if the final pixel or group of photogenerators considered is opposite different initial pixels and/or domains of the initial pixels, its brightness and/or color and/or hue components can be a weighted average in proportion to the respective surfaces. of these different initial pixels and/or domains of the initial pixels facing the final pixel or group of photogenerators considered.

For transcription, the brightness and/or color and/or hue components of an initial pixel can be modulated by the spatial gradient values of each part of the initial pixel domain with respect to the final pixel or the group of photogenerators considered.

A separation line of an initial pixel can be used to be visualized and contribute to the luminosities and/or colors and/or hues of at least the final pixel or the group of photogenerators considered, in proportion to a surface area of visualization of at least part of the separation line facing the final pixel or the group of photogenerators considered.

According to yet another aspect of the invention, a method of compressing or decompressing an initial image recorded according to a recording method according to the invention is proposed, characterized in that the image is respectively compressed or decompressed. initial image into a respectively compressed or decompressed image respectively by reducing the volume of information representing the compressed image thanks to compression, or by increasing the volume of information representing the decompressed image thanks to decompression , compression or decompression carrying on at least one component among the color(s) and/or the hue(s) and/or the luminosity and/or the at least one separation line and/or the at least one spatial gradient at within each initial pixel considered.

We can respectively compress or decompress the initial image so as to make its content directly viewable according to a display or pixelation or rasterization method according to the invention.

We can respectively compress or decompress the initial image, by means of a wavelet transformation of each component to be compressed.

We can respectively compress or decompress the initial image, by means of a transformation using Discrete Cosine Transformation (DCT).

The method according to the invention may comprise, after compression or decompression, a search for identical sequences, and an indexing of these sequences and replacement of each repeated sequence by its index in a flow of data or image information.

The method according to the invention can use lossy compression on at least one component among the color(s) and/or the hue(s) and/or the brightness and/or the at least one separation line and/or or the at least one spatial gradient of each initial pixel considered or the at least one separation within each initial pixel considered, and without loss on at least one other component among activation of the at least one separation, and/ or visualization of the at least one separation and/or width of the at least one separation.

According to yet another aspect of the invention, a computer program product or a computer program is proposed comprising instructions:

- computer code to execute any of the methods according to the invention previously described, and/or

- which, when the program is executed by a computer and/or a processor and/or any computing means, lead it to implement any of the methods according to the invention previously described.

According to yet another aspect of the invention, a (computer-readable) data carrier is proposed, on which a computer program product or a computer program according to the invention.

Description of the figures and embodiments

Other advantages and particularities of the invention will appear on reading the detailed description of implementations and embodiments in no way limiting, and the following appended drawings:

there illustrates a pixel according to the prior art, having three colors R, G, B as a property. Its position is that of its position X, Y represented by the indexing to access it, which translates its position in the image,

• un nombre α (par exemple 1 octet), référencé 62, représentant un angle (par exemple par rapport à l’horizontale) de la droite passant par le centre du pixel 2 et perpendiculaire à une ligne de séparation 3 au sein de ce pixel 2,
• une distance d, algébrique (donc signée), référencée 61, qui repère la distance entre le centre du pixel et la séparation 3 dans le sgpixel 2, et repère donc, avec α, la position de la séparation 3 dans le sgpixel 2,
• Un gradient 4 s’appliquant par exemple sur la luminosité, et donc les 3 couleurs, pour définir une variation de luminosité à l’intérieur du sgpixel 2, plus exactement à l’intérieur d’un des domaines 31 du sgpixel 2. La direction du gradient 4 est supposé parallèle au segment de la distance 61 et donc perpendiculaire à la séparation 3 (donc la direction du gradient 4 est positionnée par l’angle 62). Dans un autre mode, on représenter le gradient 4 par deux coordonnées dans le repère X, Y.

there illustrates a pixel 2 according to the invention, called sgpixel, comprising:

• a number α (for example 1 byte), referenced 62, representing an angle (for example relative to the horizontal) of the line passing through the center of pixel 2 and perpendicular to a separation line 3 within this pixel 2 ,
• a distance d, algebraic (therefore signed), referenced 61, which identifies the distance between the center of the pixel and separation 3 in sgpixel 2, and therefore identifies, with α, the position of separation 3 in sgpixel 2,
• A gradient 4 applying for example to the brightness, and therefore the 3 colors, to define a variation of brightness inside the sgpixel 2, more precisely inside one of the domains 31 of the sgpixel 2. The direction of gradient 4 is assumed to be parallel to the segment of distance 61 and therefore perpendicular to separation 3 (therefore the direction of gradient 4 is positioned by angle 62). In another mode, we represent the gradient 4 by two coordinates in the coordinate system X, Y.

there illustrates a conversion of an image of sgpixels 2 recorded according to method 1 according to the invention and to be converted to a 4 x 4 matrix of final pixels 7 of a conventional bitmap image or groups 7 of photogenerators (filled squares) d 'a screen which shows the image of sgpixels 2,

there illustrates a variation of the which takes into account the exact position of the generators R, G, B in each final pixel 7 of the screen,

there illustrates different steps 11, 12, 13 of process 1 which can follow one another several times and in any order,

there illustrates a raster image according to the state of the art,

there illustrates an image according to the invention with sgpixels according to the invention.

These embodiments being in no way limiting, we may in particular consider variants of the invention comprising only a selection of characteristics described or illustrated subsequently isolated from the other characteristics described or illustrated (even if this selection is isolated within a sentence including these other characteristics), if this selection of characteristics is sufficient to confer a technical advantage or to differentiate the invention from the prior art. This selection includes at least one preferably functional characteristic without structural details, and/or with only part of the structural details if this part only is sufficient to confer a technical advantage or to differentiate the invention from the state of the art. anterior.

We will first describe, with reference to Figures 1b and 2 to 4, examples of implementation of a method 1 according to the invention.

• au moins une ligne de séparation 3 (droite ou courbe, de préférence droite) au sein du pixel considéré 2, séparant le pixel considéré 2 au moins en deux domaines 30 (distingués par des références 31 et 32 sur la , 2 et 3 correspondant au cas d’une seule ligne 3 séparant en deux domaines 30 ; dans des variantes, on peut avoir deux, trois, quatre ou plus lignes 3 séparant un sgpixel 2 en trois, quatre, cinq ou plus domaines 30, ces lignes pouvant se croiser ou se toucher), et/ou
• au moins un gradient spatial 4 de couleur (par exemple un gradient 4 par couleur) et/ou de teinte (par exemple un gradient 4 par teinte) et/ou de luminosité au sein du pixel considéré 2 ou au sein d’au moins un des domaines 30, 31, 32 du pixel considéré 2.

These examples therefore relate to a method 1 of recording 11 (in a RAM and/or in a read-only memory and/or a computer or any other computer means or system) of a digital image in a computer format, said image or least part of the image being encoded and recorded 11 in the form of a matrix of pixels in which each pixel considered 2 (called sgpixel as explained subsequently) among at least one or more or all of the pixels of the matrix includes:

• at least one separation line 3 (straight or curved, preferably straight) within the pixel considered 2, separating the pixel considered 2 at least into two areas 30 (distinguished by references 31 and 32 on the , 2 and 3 corresponding to the case of a single line 3 separating into two domains 30; in variants, we can have two, three, four or more lines 3 separating a sgpixel 2 into three, four, five or more domains 30, these lines being able to cross or touch), and/or
• at least one spatial gradient 4 of color (for example a gradient 4 per color) and/or hue (for example a gradient 4 per hue) and/or luminosity within the considered pixel 2 or within at least one domains 30, 31, 32 of the pixel considered 2.

The terms “gradient 4” and “variability 4” and “variation 4” and “spatial variation 4” are used equivalently in this description.

The terms “separation line 3” and “separation 3” are used equivalently in this description.

Each sgpixel comprises at least one separation line 3 and/or at least one gradient 4 in that it includes data relating to this at least one separation line 3 and/or at least one gradient 4, even if only a few would not have non-zero separation 3 or gradient 4 properties. In fact, this makes it possible to normalize the memory representation; otherwise, this creates the obligation to make fields of variable size, but on the other hand, these fields of variable size can take up less memory space, overall, if few pixels are concerned, and the fact, does not not assigning memory to the zero separation 3 and/or gradient 4 fields can already be seen as being a form of compression.

LZ77 type coding is a way of replacing repeating sequences with a reference to a previous identical sequence, which avoids re-writing the sequence, just replaced by its location and its length, which takes up much less space. In a similar way, if we reduce the notion of sequence to the 3 components R, G, B of a pixel, the TIFF format - is also a way of compressing an image by replacing identical codes repeated for pixels of a line having the same color, by a sequence which says at once: “there are n fields of this color in the n pixels which follow”, for example.

Thus, the implementation of method 1 is done by adding properties to each sgpixel 2 among all or part of the pixels of the representation image of an image:

- through at least one gradient 4, we can add a property of spatial variation of color(s) and/or brightness and/or hue(s) within the sgpixel 2. Different implementations are possible.

- by the at least one separation line 3, we can add a separation property within the sgpixel 2.

Preferably, in order to obtain the best advantages of the invention, it is recommended to add the possibility of these two properties 3, 4 together for the sgpixels 2. However, the addition of only one of these properties 3, 4 already brings a significant improvement. For the at least one separation line 3, the outline of the objects is significantly improved upon enlargement. For at least one gradient 4, the texture of the objects is significantly sharper, with the absence of pixel slots on gradients of color(s) and/or brightness and/or hue(s) for objects with sharp contours not parallel to the main axes of the image (therefore different, from 0, 90°, and also 45° and -45°) which are the least well restored by the prior art.

By default, all angle values in this description are given relative to the horizontal X axis of pixel array 2 or 7.

We call the pixels modified 2 by method 1 sgpixels 2 in the rest of the presentation, "sg" as "separable and gradable" (to indicate the notion of rate of spatial variation of one or more components of colors, hues or pixel brightness).

Preferably, the pixel matrix comprises or consists of a matrix of sgpixels 2.

The axes of the image plane, perpendicular to each other, are called X and Y. The Z axis if used is the axis is perpendicular to X and Y.

Method 1 can be applied to images comprising a transparency channel (generally called Alpha channel), or any other complementary channel.

• Deux points 51, 52 sur un pourtour 53 du sgpixel 2 considéré, et/ou
• Une distance 61, de préférence par rapport au centre du sgpixel considéré 2, et un angle 62.

Each separation line 3 within the sgpixel 2 considered is coded by two pieces of data comprising:

• Two points 51, 52 on a perimeter 53 of the sgpixel 2 considered, and/or
• A distance 61, preferably relative to the center of the sgpixel considered 2, and an angle 62.

• préférentiellement, ces sgpixels 2 sont rangés selon un ordre prédéfini en lignes et colonnes (ou l’inverse), à partir d’une définition globale du nombre de lignes et colonnes, et globale de leur rapport de dimension s’il n’est pas égal à 1. Cela permet d’économiser l’écriture d’une propriété position et/ou contour de chaque sgpixel 2 qui est implicitement déduite de la position dans la table. Aussi, un traitement élémentaire itératif en est facilité. Cet arrangement est déjà connu de l’art antérieur pour la représentation ‘bitmap’ d’une image ;
• mais si l’image à représenter est dotée de zones très peu denses et significatives en surface, on peut opter pour une réalisation qui ajoute des propriétés de position et/ou dimension variables de chaque sgpixels 2, par exemple obtenues par un code qui pointe sur une table ou une convention globale. On peut par exemple décider de 2 types de sgpixels 2, des carrés élémentaires de dimension u x u, et l’autre type de carrés de 2u x 2u. Il suffit de consacrer un bit d’information pour savoir de quel type de sgpixel 2 il s’agit, et de mettre une place une règle pour convertir la position dans la table en position dans l’image en tenant compte du cumul des largeurs des sgpixels 2 précédemment rencontrés.

Preferably all the sgpixels 2 and/or generally all the pixels of the pixel matrix have the same dimensions, but a variant with different dimensions is also possible. So :

• preferably, these sgpixels 2 are arranged according to a predefined order in rows and columns (or the reverse), from an overall definition of the number of lines and columns, and global definition of their dimension ratio if it is not equal to 1. This saves writing a position and/or contour property of each sgpixel 2 which is implicitly deduced from the position in the table. Also, elementary iterative processing is facilitated. This arrangement is already known from the prior art for the 'bitmap' representation of an image;
• but if the image to be represented has very sparse and significant areas on the surface, we can opt for a realization which adds variable position and/or dimension properties of each sgpixels 2, for example obtained by a code which points to a global table or convention. We can for example decide on 2 types of sgpixels 2, elementary squares of dimension uxu, and the other type of squares of 2u x 2u. It is enough to dedicate a bit of information to know what type of sgpixel 2 it is, and to put a place a rule to convert the position in the table into position in the image taking into account the accumulation of the widths of the sgpixels 2 previously encountered.

Method 1 is preferably implemented embedded in a software module which is added to or integrated into a Web browser, for taking into account the images recorded according to method 1, possibly their compression or decompression 12 as described subsequently, and/or their visualization or display 13 in part or all of a screen interacting with the web browser.

Likewise :

- textual information represented by pointing symbols in a character font, and/or

- drawings of objects embedded in the form of raster or vectorized sub-images

can be added directly to the image recorded according to method 1 or added during its visualization 13, in order to be viewable, possibly according to a particular selection which only shows all or part of it, during a restitution of the recorded image.

Thus, process 1 can be applied:

- to images which receive (from another channel) textual information encoded by character codes, activating symbols on the image (such as subtitling for example, or an overload with characters), and /Or

- to images which receive vectorized elements in addition to pixels.

Likewise, metadata can be added to the image recorded using method 1.

Implementation of the property of gradient 4 (or spatial variation) of color(s) and/or hue(s) and/or brightness

In reference to the , we see that separation lines 3 of several neighboring sgpixels 2 can be assembled by juxtaposition so as to form a continuous line crossing these several neighboring sgpixels 2.

• en haut de ce sgpixel considéré : un pixel ou sgpixel voisin NORD, pour lequel on pourra utiliser par la suite un indice N
• en bas de ce sgpixel considéré : un pixel ou sgpixel voisin SUD, pour lequel on pourra utiliser par la suite un indice S
• à gauche de ce sgpixel considéré : un pixel ou sgpixel voisin OUEST, pour lequel on pourra utiliser par la suite un indice O
• à droite de ce sgpixel considéré : un pixel ou sgpixel voisin EST, pour lequel on pourra utiliser par la suite un indice E
• en haut à gauche de ce sgpixel considéré : un pixel ou sgpixel voisin NORD OUEST, pour lequel on pourra utiliser par la suite un indice NO
• en haut à droite de ce sgpixel considéré : un pixel ou sgpixel voisin NORD EST, pour lequel on pourra utiliser par la suite un indice NE
• en bas à gauche de ce sgpixel considéré : un pixel ou sgpixel voisin SUD OUEST, pour lequel on pourra utiliser par la suite un indice SO
• en bas à droite de ce sgpixel considéré : un pixel ou sgpixel voisin SUD EST, pour lequel on pourra utiliser par la suite un indice SE.

Each sgpixel 2 considered can have 8 neighboring pixels or sgpixels (if it is not located on an edge of the pixel matrix):

• at the top of this sgpixel considered: a neighboring NORTH pixel or sgpixel, for which we can subsequently use an index N
• at the bottom of this sgpixel considered: a neighboring pixel or sgpixel SUD, for which we can subsequently use an index S
• to the left of this sgpixel considered: a neighboring WEST pixel or sgpixel, for which we can subsequently use an index O
• to the right of this sgpixel considered: a neighboring pixel or sgpixel EST, for which we can subsequently use an index E
• at the top left of this sgpixel considered: a neighboring NORTH WEST pixel or sgpixel, for which we can subsequently use an index NO
• at the top right of this sgpixel considered: a neighboring pixel or sgpixel NORTH EAST, for which we can subsequently use an index NE
• at the bottom left of this sgpixel considered: a neighboring SOUTH WEST pixel or sgpixel, for which we can subsequently use an index SO
• at the bottom right of this sgpixel considered: a neighboring SOUTH EAST pixel or sgpixel, for which we can subsequently use an SE index.

• Un gradient 4 de luminosité, et/ou
• Un ou plusieurs gradients 4 correspondant respectivement à une ou plusieurs couleur(s) et/ou teinte(s).

The at least one gradient 4 within the considered sgpixel 2 or within at least one of the domains 30, 31, 32 of the considered sgpixel 2 or, as we will see later, within the at least a separation line 3 of the sgpixel 2 considered comprises:

• A gradient 4 of brightness, and/or
• One or more gradients 4 corresponding respectively to one or more color(s) and/or shade(s).

Each gradient 4 among the at least one gradient 4 within the sgpixel 2 considered or at least one of the domains 30, 31, 32 of the sgpixel considered 2 or the at least one separation line 3 of the sgpixel 2 considered is coded by:

• Une seule composante spatiale (qui correspond donc à la norme du vecteur du gradient 4) définie (de préférence perpendiculairement à une ligne de séparation 3 par défaut, ce qui évite à avoir à mémoriser cette donnée) par rapport à l’au moins une ligne de séparation 3 au sein du sgpixel 2 considéré. Cette ligne de séparation 3 permet ainsi de définir le sens du vecteur du gradient 4, en considérant par exemple toujours par défaut que le vecteur du gradient 4 est toujours perpendiculaire à cette ligne 3 qui par défaut est une ligne droite.

– two spatial components preferably in the X and Y directions to define the direction and the module of the vector of this gradient 4, or

• A single spatial component (which therefore corresponds to the norm of the gradient vector 4) defined (preferably perpendicular to a separation line 3 by default, which avoids having to memorize this data) in relation to the at least one line separation 3 within the sgpixel 2 considered. This separation line 3 thus makes it possible to define the direction of the gradient vector 4, for example always considering by default that the gradient vector 4 is always perpendicular to this line 3 which by default is a straight line.

• selon 2 coordonnées X et Y, ou
• juste par une valeur (de préférence signée), son angle étant par définition perpendiculaire à la séparation 3 (même inactive), i.e. un gradient spatial 4 selon une seule coordonnée.

We can thus represent the spatial gradient 4:

• according to 2 coordinates X and Y, or
• just by a value (preferably signed), its angle being by definition perpendicular to the separation 3 (even inactive), ie a spatial gradient 4 according to a single coordinate.

• à toutes les composantes (de couleur) à la fois, donc à une combinaison de ces dernières, que l’on peut assimiler à une notion de luminosité par exemple, ou
• de façon différente aux différentes composantes de couleur ou teinte, ce qui donne plus de données à utiliser qu’un seul vecteur gradient (à 2 coordonnées) : 4 ou 6 coordonnées X Y pour chaque couleur, ou bien 2 à 3 coordonnées d’amplitude signée de variation, si l’angle est connu à partir de la séparation 3 (active ou pas).

Gradient 4 can be applied:

• to all the components (of color) at the same time, therefore to a combination of these, which can be assimilated to a notion of luminosity for example, or
• differently to the different color or hue components, which gives more data to use than a single gradient vector (with 2 coordinates): 4 or 6 XY coordinates for each color, or 2 to 3 coordinates of signed amplitude variation, if the angle is known from separation 3 (active or not).

The notion of spatial gradient 4 (first derivative(s)) can be supplemented by one or more terms of order higher than one (second derivatives, third derivatives, etc.), in order to better connect the jumps in slope of brightness and/or color(s) and/or hue(s) between sgpixels 2.

Implementing separation property 3

Each separation 3 is a segment type object, or even curved line, placed according to a set of parameters on the sgpixel 2 considered, separating it into several distinct domains 30, in particular each domain 30 can have a definition of color(s) and/or or hue(s) and/or luminosity which is specific to it and independent of the other domain(s) of this same sgpixel considered 2.

Eventually, there may be several (more than one) separations 3 per sgpixel 2, that is to say more than two domains 30 per sgpixel 2.

As explained previously, in the case of a separation 3 which is a segment, this separation 3 defines its relative position in the sgpixel 2 either by the position of 2 points A (reference 51) and B (reference 52) on the contours 53 of sgpixel 2 or by knowing a distance 61 to the center of sgpixel 2 and an angle 62 (polar coordinates).

Each separation 3 of a sgpixel 2 separates this sgpixel 2 into at least two domains 30 (for example, for two intersecting separations 3, each separation 3 will have two domains 30 on each of its sides, i.e. 4 domains in all within of one sgpixel 2) and:

- each domain 30, 31, 32 has a set of properties of color(s), hue(s), brightness, and spatial variation(s) 4 if this variation 4 is implemented), or

- at least one domain 30, 31 is described by a property set of color(s), hue(s), brightness and/or gradient(s) 4, the at least one other domain 30, 32 inherits color (s), hue(s), brightness and/or gradient(s) 4 of neighboring pixels or sgpixel(s). This solution is preferred compared to the solution of defining in the sgpixel 2 a set of color(s), hue(s), brightness and/or gradient(s) 4 for each domain 30, 31, 32. effect, the memory footprint to describe the image is reduced accordingly, and the inheritance of color(s), hue(s), brightness and/or gradient(s) 4 from the neighboring pixel(s) or sgpixels creates naturally an extension of color(s), hue(s), brightness and/or gradient(s) 4 in continuity with these sgpixels 2. The disadvantages being that it is a question of establishing some rules on how to read the neighboring pixels or sgpixels, and where appropriate to mix the information of several contiguous or close pixels or sgpixels between them and in relation to the region of sgpixel 2 to which it is necessary to inherit properties of color(s), tint(s), brightness and/or gradient(s) 4 as described in a little more detail later.

As illustrated in Figures 2 and 3, each separation line 3 has a displayable state with a display surface (which corresponds for example to the product of the length by the width of the line allowing this separation line 3 to be visualized) with its or its own brightness and/or color(s) and/or hue(s).

By displayable, we mean a state of the separation line 3 for which this line 3 is or will be displayed and visualized with its display surface when it is (if a display is in progress) or will be (if a display is not not in progress) displayed on a screen.

This display surface is in arbitrary units, for example defined in relation to the surface of pixel 2 comprising this separation line 3.

In addition, the at least one separation line 3 may itself include a spatial gradient making it possible to modulate, when it is displayed or viewable, its brightness(es) and/or color(s) and/or hue(s).

In method 1, the at least one separation line 3 can be switched between its displayable or visible state (with its display surface) and a non-displayable or invisible state.

Thus, each separation 3 can be made visible or invisible.

In the case where a separation 3 is made visible, a set of thickness or width and length or viewing surface is added to it, and possibly a set of color(s) and/or tint(s) and/or brightness , and/or spatial variation(s) 4 of color(s) and/or hue(s) and/or brightness, with all or part of these properties specific to this separation, and/or all or part of these properties defined globally for all separations 3 of all sgpixels of the image.

In the case where a separation 3 is made invisible, it no longer has a viewing surface (but is still coded and recorded by method 1, in particular for its position within its sgpixel 2, but under a status or state " invisible ")

The at least one separation line 3 can also be switched between an active state and an inactive state. Thus each separation 3 can be active or inactive. When it is inactive (and still coded and recorded by method 1, in particular for its position within its sgpixel 2, but under an “inactive” status or state) and it is the only separation 3 of an sgpixel 2, this sgpixel 2 is only considered as a single domain, in particular to uniformly apply its information of color(s), hue(s), and/or brightness and/or its gradient(s) 4 (if used and not zero) of color(s), hue(s), and/or brightness. If separation 3 is inactive and variability 4 of color(s), hue(s), and/or brightness is at zero value, we therefore find the visual rendering of a 'classic' pixel of the prior art, even if method 1 will be quite distinct from the prior art due to specific coding or recording of these zero values.

• Actif et visualisable ou visualisée : elle est active, donc elle marque une séparation entre plusieurs domaines 30, et on la voit avec sa surface de visualisation
• Actif et non visualisable ou non visualisée : elle est active, donc elle marque une séparation entre plusieurs domaines 30, mais on ne peut pas la voir, autrement dit cette ligne 3 est une ligne avec une longueur droite ou curviligne mais a une épaisseur ou largeur nulle,
• Non Actif et non visualisable ou non visualisée : elle est non active, donc elle ne marque pas de séparation entre plusieurs domaines 30, et on ne peut pas la voir, autrement dit cette ligne 3 est une ligne avec une longueur droite ou curviligne mais a une épaisseur ou largeur nulle.
• Non Actif et visualisable ou visualisée : elle est non active, donc elle ne marque pas de séparation entre plusieurs domaines 30, mais on la voit avec sa surface de visualisation.

Thus, each separation line 3 can have a state:

• Active and viewable or visualized: it is active, therefore it marks a separation between several domains 30, and we see it with its viewing surface
• Active and not visible or not visualized: it is active, therefore it marks a separation between several domains 30, but we cannot see it, in other words this line 3 is a line with a straight or curvilinear length but has a thickness or width nothing,
• Not Active and not viewable or not visualized: it is not active, therefore it does not mark a separation between several domains 30, and we cannot see it, in other words this line 3 is a line with a straight or curvilinear length but has zero thickness or width.
• Not Active and viewable or visualized: it is not active, so it does not mark a separation between several domains 30, but we see it with its viewing surface.

Typically, the method comprises a deactivation of the at least one separation line 3 of the sgpixel considered 2 if this sgpixel considered 2 is sampled below a certain spatial sampling threshold, to avoid erroneous interpretation of areas that are too poorly sampled. . This allows you to read a text at the sampling limit, but avoids the risk of no longer recognizing what is written if you zoom in. Indeed, keeping separations 3 activated could amount to deciding incorrectly on the meaning of pixels, which it may then be more prudent not to do so as not to encode false information in the image.

The visibility and activation of a separation 3 can therefore be independent, for example:

- separation 3 inactive (so color(s), hue(s), brightness(es), gradient(s) 4 identical on either side of separation 3), but

- visibility of the active separation to represent a thread, a hair, a very fine object... superimposed on the background colors of the sgpixel 2.

We can also add a partial transparency property to the visibility of at least one separation line 3.

Inheritance rules from neighboring pixels or sgpixels

• au moins une luminosité et/ou couleur et/ou teinte au sein du sgpixel considéré 2 ou de l’au moins un des domaines 30, 31, 32 du sgpixel considéré 2 dépend d’une luminosité et/ou couleur et/ou teinte d’au moins un pixel voisin (qui peut aussi être un sgpixel) du sgpixel considéré 2 ou d’au moins un domaine 30, 31, 32 d’un pixel voisin (qui peut aussi être un sgpixel) du sgpixel considéré 2, et/ou
• au moins un gradient 4 de luminosité et/ou de couleur et/ou de teinte au sein du sgpixel 2 considéré ou de l’au moins un des domaines 30, 31, 32 du sgpixel considéré 2 dépend d’un gradient 4 de luminosité et/ou de couleur et/ou de teinte d’au moins un pixel voisin (qui peut aussi être un sgpixel) du sgpixel considéré 2 ou d’au moins un domaine 30, 31, 32 d’un pixel voisin (qui peut aussi être un sgpixel) du sgpixel considéré 2.

Preferably, for at least one or each sgpixel 2 considered:

• at least one brightness and/or color and/or hue within the sgpixel considered 2 or at least one of the domains 30, 31, 32 of the sgpixel considered 2 depends on a brightness and/or color and/or hue d 'at least one neighboring pixel (which can also be an sgpixel) of the considered sgpixel 2 or at least one domain 30, 31, 32 of a neighboring pixel (which can also be an sgpixel) of the considered sgpixel 2, and/ Or
• at least one gradient 4 of brightness and/or color and/or hue within the sgpixel 2 considered or at least one of the domains 30, 31, 32 of the sgpixel considered 2 depends on a gradient 4 of brightness and /or color and/or hue of at least one neighboring pixel (which can also be an sgpixel) of the sgpixel considered 2 or of at least one domain 30, 31, 32 of a neighboring pixel (which can also be one sgpixel) of the considered sgpixel 2.

Typically, the color component(s), hue(s), and/or brightness and/or its gradient(s) 4 of the neighboring sgpixel are spatially extended to obtain the color component(s), hue(s), and/or brightness and/or its gradient(s) 4 (If variability 4 is implemented in method 1) of the domain 30, 32 not defined in its own right.

Depending on the angles 62 of the at least one separation 3 of the sgpixel 2 inheriting at least one component, there may be inheritance of one or more neighboring pixels with average or weighted average from neighboring pixels.

Example 1: in the case of vertical separation 3 (90° angle) which defines a right side, method 1 takes into account for domain 32 the colors of the sgpixel on the right (or EST) (with consideration of at least variability 4).

Example 2: in the case of oblique separation 3 at an angle 135°, method 1 takes into account for domain 32 the neighboring pixel in the North-East position.

Example 3: in the case of oblique separation 3 at an angle between 90° and 135°, method 1 makes a pro rata weighting between its 2 sgpixels East and North-East, from the angular difference, which is written :

C _c (i) = ((135° - α) * Prol(C _E (i)) + (α - 90°) * Prol(C _NE (i)))/(135°-90°)

• α l’angle de la séparation (0° étant l’horizontale, l’héritage se faisant alors en dessous avec le pixel voisin SUD, 180° étant aussi l’horizontale avec héritage avec le pixel voisin Nord au-dessus, ce qui traduit le fait que le segment 3 est orienté et ce qui définit le domaine 31 en propre et celui 32 hérité.)
• C_c(i) sont les valeurs des couleurs i = 1 à 3 (pour respectivement 1 pour R (pour Rouge ou Red), 2 pour G (pour Green ou Vert), 3 pour B (pour Blue ou Bleu)) du sgpixel considéré 2 au centre.

With :

• α the angle of separation (0° being the horizontal, the inheritance then taking place below with the neighboring pixel SOUTH, 180° also being the horizontal with inheritance with the neighboring pixel North above, which translates the fact that segment 3 is oriented and what defines domain 31 itself and that 32 inherited.)
• C _c (i) are the values of the colors i = 1 to 3 (for respectively 1 for R (for Red or Red), 2 for G (for Green or Green), 3 for B (for Blue or Blue)) of the sgpixel considered 2 in the center.

• C_E(i) pour le sgpixel à l’Est (ou à droite) sont les valeurs des couleurs i = 1 à 3 (pour respectivement 1 pour R (pour Rouge ou Red), 2 pour G (pour Green ou Vert), 3 pour B(pour Bue ou Bleu)) du sgpixel voisin à droite du sgpixel considéré 2 au centre,
• C_NE(i) pour le sgpixel au Nord Est (ou en haut à droite) sont les valeurs des couleurs i = 1 à 3 (pour respectivement 1 pour R (pour Rouge ou Red), 2 pour G (pour Green ou Vert), 3 pour B (pour Bue ou Bleu)) du sgpixel voisin en haut à droite du sgpixel considéré 2 au centre,
• Prol() est l’opérateur de prolongement qui tient compte de la variabilité spatiale 4 définie dans le sgpixel en argument, vers le lieu de l’héritage.

The index c is for the sgsgpixel 2 'at the center', from which a domain 32 inherits from its neighbors,

• C _E (i) for the sgpixel to the East (or to the right) are the values of the colors i = 1 to 3 (for respectively 1 for R (for Red or Red), 2 for G (for Green or Green), 3 for B (for Bue or Blue)) of the neighboring sgpixel to the right of the considered sgpixel 2 in the center,
• C _NE (i) for the sgpixel in the North East (or top right) are the values of the colors i = 1 to 3 (for respectively 1 for R (for Red or Red), 2 for G (for Green or Green) , 3 for B (for Bue or Blue)) of the neighboring sgpixel at the top right of the sgpixel considered 2 in the center,
• Prol() is the extension operator which takes into account the spatial variability 4 defined in the sgpixel argument, towards the place of inheritance.

This rule can be expressed similarly for the alpha angle varying from 0 to 360°, by invoking the different sgpixels at the cardinal points South, South-East, East, North-East, North, North-West, West, South- West.

This provides a non-limiting example of a rule providing color continuity between the part of the sgpixel that inherits and its neighbor(s). Other methods are also possible, for example by involving more than 1 or 2 neighboring sgpixels in the rules defined above.

Coding of properties or components

• Les composantes (aussi appelés canaux) du sgpixel, 2 qui comprennent :
  • couleur(s), teinte(s), et/ou luminosité du sgpixel 2, et/ou
  • couleur(s), teinte(s), et/ou luminosité de chaque domaine 30 (de préférence distincts pour chaque domaine 30) au sein du sgpixel 2, et/ou
  • couleur(s), teinte(s), et/ou luminosité de chaque séparation 3 (de préférence distincts pour chaque séparation 3) au sein du sgpixel 2, et/ou
  • gradient(s) de couleur(s), teinte(s), et/ou luminosité du sgpixel 2, et/ou
  • gradient(s) de couleur(s), teinte(s), et/ou luminosité de chaque domaine 30 (de préférence distincts pour chaque domaine 30) au sein du sgpixel 2, et/ou
  • gradient(s) de couleur(s), teinte(s), et/ou luminosité de chaque séparation 3 (de préférence distincts pour chaque séparation 3) au sein du sgpixel 2
• état actif ou état inactif de chaque séparation 3 (de préférence distincts pour chaque séparation 3) au sein du sgpixel 2, et/ou
• état visualisée ou état non visualisée (ou état visualisable ou état non visualisable) de chaque séparation 3 (de préférence distincts pour séparation 3) au sein du sgpixel 2, et/ou
• position de chaque séparation 3 (dont l’orientation qui peut définir un domaine 31 de propriétés de couleur(s), et/ou luminosité et/ou teinte définie par celles du pixel et/ou gradients, et une zone 32 héritant de propriétés de pixels voisins) au sein du sgpixel 2, et/ou
• surface ou largeur de visualisation de chaque séparation 3 (de préférence distincts pour séparation 3) au sein du sgpixel 2.

In the method 1 just described, the properties coded and recorded by the method 1 of at least one or each sgpixel 2 therefore include:

• The components (also called channels) of the sgpixel, 2 which include:
  • color(s), hue(s), and/or brightness of sgpixel 2, and/or
  • color(s), hue(s), and/or brightness of each domain 30 (preferably distinct for each domain 30) within the sgpixel 2, and/or
  • color(s), hue(s), and/or brightness of each separation 3 (preferably distinct for each separation 3) within the sgpixel 2, and/or
  • color gradient(s), hue(s), and/or brightness of sgpixel 2, and/or
  • gradient(s) of color(s), hue(s), and/or brightness of each domain 30 (preferably distinct for each domain 30) within the sgpixel 2, and/or
  • gradient(s) of color(s), hue(s), and/or brightness of each separation 3 (preferably distinct for each separation 3) within the sgpixel 2
• active state or inactive state of each separation 3 (preferably distinct for each separation 3) within the sgpixel 2, and/or
• visualized state or non-visualized state (or displayable state or non-displayable state) of each separation 3 (preferably distinct for separation 3) within sgpixel 2, and/or
• position of each separation 3 (including the orientation which can define a domain 31 of color properties, and/or brightness and/or hue defined by those of the pixel and/or gradients, and a zone 32 inheriting properties of neighboring pixels) within sgpixel 2, and/or
• surface or width of visualization of each separation 3 (preferably distinct for separation 3) within the sgpixel 2.

• peut être définie par un jeu d’information numériques appartenant en propre à chaque sgpixel 2, ou
• dans un souci de réduction de l’emprunte mémoire, peut être obtenue par référence via un code en propre au sgpixel 2 adressant une table externe. Cette table externe réalise un dictionnaire de valeurs de propriétés ou composantes, afin d’obtenir une description plus large et paramétrable, sans augmenter significativement l’emprunte mémoire au niveau du sgpixel 2.

In method 1, any of the components or properties previously described (ie at least one of these components or properties, all or part):

• can be defined by a set of digital information belonging specifically to each sgpixel 2, or
• in order to reduce the memory footprint, can be obtained by reference via code specific to sgpixel 2 addressing an external table. This external table creates a dictionary of property or component values, in order to obtain a broader and configurable description, without significantly increasing the memory footprint at the sgpixel 2 level.

For example, this implementation by table pointed from an index in sgpixel 2 will be useful if we assume that the 'visibility' property is only rarely activated in an image. For example, visibility will be activated to restore the presence of hair on a person, the presence of fine threads emanating from a fabric. As these elements are rarely present, and they generally have a rather homogeneous color, or a similar width, but a variable color, we can choose not to endow the sgpixels 2 with all the properties specific to these separations, properties such than colors, color gradient along the sgpixel, width of the separator. For example, we will choose to represent the brightness of the visible separations in each sgpixel, considering that it can vary a lot on the image, but the hue, the saturation (of color), the width will be defined in a table thus having few elements, considering that these elements vary little, or even include non-disturbing quantification, for example 15 entries maximum. Thus, each sgpixel 2 has for example a code for activating the visibility of the separation on a 4-bit integer (i.e. 2 ⁴ = 16 values), the value 0 being invisibility, the values 1 to 15 pointing in said previously described table. We can possibly code the brightness of these objects on 4 to 8 other bits (i.e. 16 to 256 values), or not and place this information also in said table.

In another example of representation of the position of separation 3 in sgpixel 2, conversely, still as an example, suppose that the position property of separation 3 is often activated, or even according to a variety of angles and positions that would make a significant number of combinations. So, we represent for example by two one-byte numbers (i.e. 256 values) the two points of the separation segment 3 in contact with the contour of the sgpixel 2. We can keep one of the values of these two numbers to make the separation 3 inactive. Rather than coding these elements in a table, which would have as many entries as these two bytes, which would not save memory space, having a two-byte code addressing the table, and 2 bytes in the table, to compare to the only two bytes of the property itself in sgpixel 2.

Transcription

When viewing an image obtained by method 1, it is appropriate to transform the information coded by this method into values of the photoemitters 71, 72, 73 generating the image (recorded by method 1) on a screen, which amounts to pixelizing (or 'to rasterize' in English) the image according to the pitch of the display screen.

Thus, method 1 can include a conversion of an initial image into a final image for:

- a display 13 of the initial image in the so-called final image on a screen comprising groups 7 of photogenerators 71, 72, 73 (each group 7 comprising for example three photogenerators R referenced 71, V referenced 72 and B referenced 73) , Or

- a pixelization 13 or rasterization 13 of said initial image into the final image comprising final pixels 7 devoid of separation line and/or spatial gradient of color(s) and/or hue(s) and/or brightness within these final pixels 7 (in other words no final pixel 7 is an sgpixel),

said initial image being recorded according to the invention with at least one sgpixel 2 as described previously.

The reference 7 is used for each final pixel 7 or for each group 7 of photogenerators, in particular on the , because the principle of transcription is the same.

Each initial pixel can be an sgpixel 2, the initial pixels comprising at least one, preferably several, sgpixel(s) 2. We will preferably consider that each initial pixel is an sgpixel 2.

• Pour chaque pixel final 7 considéré ou chaque groupe 7 de photogénérateurs 71, 72, 73 considéré, une transcription 13 d’au moins pixel initial 2 ou d’au moins un domaine 30 de pixel initial 2 en ce pixel final 7 ou groupe 7 de photogénérateurs considéré:
  • à partir des composantes de luminosité(s) et/ou couleur(s) et/ou teinte(s) de cet au moins un pixel initial 2 ou au moins un domaine 30 de pixel initial
  • vers ce pixel final 7 ou groupe 7 de photogénérateurs 71, 72, 73 considéré,

Thus, method 1 comprises:

• For each final pixel 7 considered or each group 7 of photogenerators 71, 72, 73 considered, a transcription 13 of at least initial pixel 2 or of at least one domain 30 of initial pixel 2 in this final pixel 7 or group 7 of photogenerators considered:
  • from the components of brightness(s) and/or color(s) and/or hue(s) of this at least one initial pixel 2 or at least one domain 30 of initial pixel
  • towards this final pixel 7 or group 7 of photogenerators 71, 72, 73 considered,

• cet au moins pixel initial ou au moins un domaine 30 de pixel initial et
• le pixel final 7 ou groupe 7 de photogénérateur considéré.

the transcription being carried out in proportion to this at least initial pixel or at least one domain 2 of initial pixel 2 within a common surface facing between:

• this at least initial pixel or at least one domain 30 of initial pixel and
• the final pixel 7 or group 7 of the photogenerator considered.

In the present description, it is considered that two parts, on the one hand of the initial image and on the other hand of the final image, face each other or are facing each other if they are superimposed in a common representation on the one hand of the the initial image and on the other hand the final image in a common conversion reference illustrating the conversion or transcription of the initial image to the final image.

For transcription, if the final pixel 7 or group 7 of photogenerators considered is opposite different initial or initial pixel(s) and/or domain(s) 30 of initial or initial pixel(s) and/or separation 3 of initial or initial pixel(s), the components of brightness and/or color(s) and/or tint(s) of the final pixel 7 or of the group 7 of photogenerators considered are a weighted average in proportion to the respective surfaces of these different pixels initial or initial pixel(s) and/or domain(s) 30 of initial or initial pixel(s) and/or separation 3 of initial or initial pixel(s) facing the final pixel 7 or group 7 of photogenerators considered.

For example, for the calculation of the proportion of surfaces, in addition to Figures 2 and 3 which illustrate in particular cases which will be described subsequently, let P _i be the initial pixels, an initial pixel of index k.

Let P _f be the final pixels, a final pixel of index r.

We notice a color component value,

where c is the color index, for example 1..3,

d the index of the domain, for example 1..2 for each side of the separation 3. We add the value 3 if the separation is visible, to describe this new domain, domain described by the axis of the line of the separation 3 , thickened by half its width on each side. This thickening is to be deduced from the surface of the adjacent domains 31, 32.

is the common facing surface (hence the intersection symbol) between the initial pixel (sgpixel) index k, domain d, and the final pixel (f as final).

is the total surface area of the final pixel, also denoted S _f if all the final pixels have identical surfaces.

The color contribution for the final pixel r, coming from the initial pixel k, is:

with :

K(r) the set of indices k for index r for which there is a match.

D(k) the set of domain(s) of pixel k (previously chosen from K(r))

and which must be understood as: on the left, the color component c, for the final pixel of index r (r which can also be written according to two indices ixr, and iyr which describe the 2 dimensions, for example concatenated in a table described by the index r) is equal to the sum over all the indices k of initial pixels, facing the pixel r, all the domains d of this pixel k - concerned by the intersection, or all the domains , those taken in excess being assigned a zero surface, they naturally disappear from the sum -, from the color component c, times the common surface opposite (hence the intersection symbol), sum divided by the total surface of the pixel restored.

The domain d can take the values 1 and/or 2 depending on the side or domain 31, 32 of the (sg)pixel 2 in question, if there is only one separation 3 active. If separation 3 is inactive, there is only one domain d=1 which occupies the entire sgpixel, unless separation 3 is visible, see below.

In the case where separation 3 is made visible, we define a new domain d=3. The calculation of the surface can be explained as the product of the length L _sep of the separation 3 of the initial sgpixel 2 which is opposite the final pixel 7 (also called restored pixel), times the width W of the separation 3 of this initial pixel k (which can be preferentially defined as a property for each initial pixel k, or global to use less memory space), which is written:

For transcription, the components of brightness and/or color(s) and/or hue(s) of an initial pixel and/or domain 30 of an initial pixel and/or separation 3 of an initial pixel are modulated by the values of gradient(s) 4 of each part of initial pixel and/or domain 30 of initial pixel and/or separation 3 of initial pixel facing the final pixel 7 or the group 7 of photogenerators considered.

For example, for the calculation of the proportion of surfaces with gradients 4, we therefore have:

where we add a vector (therefore which has 2 coordinates) in Col, to take into account that Col is modified by the gradient 4, inside the sgpixel. We agree on all points belonging to the intersecting surface.

The dependence of color on position in the initial pixel can be written:

Or is the value of the gradient 4 of all the components at once, - which is then written - or a potentially different gradient for each component c. We make a scalar product, denoted '.', between this gradient and the displacement vector , where C is the color reference point , and M the point which describes the initial pixel. This point C is for example the center of the initial pixel, or a corner.

Depending on the representations chosen, c can take for example three values 1 to 3, Col() represents the values of the colors Red, Green, Blue. Or again, in a YUV system, with Y the relative luminance and U and V for the chrominance, and:

Y ≃ R + G + B

U ≃ B–Y

V ≃ R – Y

then Col(1) can represent the brightness, Col(2) the U component, Col(3) the V component. Other types of representation are obviously possible, such as the TSL representation for hue saturation brightness (in English HSL Hue Saturation Lightness).

A separation line 3 of an initial pixel can be used to be visualized and contribute to the luminosities and/or colors and/or hues of at least one final pixel 7 or group 7 of photogenerators considered, in proportion to any or all of the part, of the viewing surface of the separation line 3, which is opposite the final pixel 7 or the group 7 of photogenerators considered.

In the preferred and illustrated case of a separation line 3 in a straight line, this display surface is obtained from the product of a property of width of the separation line 3 by the length of said separation line 3.

The visualization surface of the separation line 3 separating different domains 30 accordingly reduces the surface taken into account of the adjacent domain(s) 30 of the initial sgpixel 2 for the other surface proportions.

• Si la densité de pixel l’image initiale de sgpixels est telle que l’on a un pixel initial en face d’un groupe 7 photogénérateur R, G, B de l’écran, le procédé 1 calcule la valeur moyenne sur chaque composante de couleur de l’écran, au prorata de la surface de chaque couleur de chaque domaine 30 du sgpixel 2 initial de l’image initiale, éventuellement complétée au prorata de la surface de visualisation de la séparation 3 (sa largeur x sa longueur) fois ses composantes de couleurs, respectivement.
• Si ce rapport 1 :1 n’est pas obtenu, il conviendra de considérer un prorata de la surface relative occupée par le dit photogénérateur 71, 72 ou 73 (surface ramenée à la surface totale du pas de réplication) et de calculer la moyenne des propriétés des domaines 30 des pixels initiaux en superposition de cette surface. De la sorte, les valeurs moyennes de lumière transmise ou générée par l’écran seront spatialement au plus proches des informations représentées localement dans chaque domaine 30 des sgpixels 2 initiaux, pour une meilleure fidélité de la représentation de l’image par l’écran.

For example, with reference to Figures 2 and 3, for pixelization (or rasterization) for viewing the initial image on a screen:

• If the pixel density of the initial image of sgpixels is such that we have an initial pixel facing a group 7 photogenerator R, G, B of the screen, method 1 calculates the average value on each component of color of the screen, in proportion to the surface of each color of each domain 30 of the initial sgpixel 2 of the initial image, possibly supplemented in proportion to the viewing surface of the separation 3 (its width x its length) times its color components, respectively.
• If this 1:1 ratio is not obtained, it will be appropriate to consider a proportion of the relative surface occupied by said photogenerator 71, 72 or 73 (surface reduced to the total surface of the replication step) and to calculate the average of the properties of the domains 30 of the initial pixels superimposed on this surface. In this way, the average values of light transmitted or generated by the screen will be spatially closest to the information represented locally in each domain 30 of the initial sgpixels 2, for better fidelity of the representation of the image by the screen.

This type of rule does not prevent the application of gamma (linearity) type corrections more generally, or color conversion between the initial image of the sgpixels and its materialization produced on the screen.

Of course, in the event of enlargement, only part of the initial image of sgpixels can be represented, on the basis of the same correspondence rules applied to pixels in or straddling the area of the screen.

Figures 2 and 3 illustrate how to apply this type of rule, with a variation depending on whether we limit ourselves to considering the final pixel 7 or group 7 for its overall size R+G+B as indivisible, or whether we cut it into contrary in 3 distinct entities 71, 72, 73 present in their respective and different positions.

In reference to the , for pixelization (or rasterization) for conversion to a classic bitmap format image with final pixels without sgpixel, method 1 proceeds with the same correspondence rules for those exposed to materialize the initial image of sgpixels towards a screen, if we wish to convert it to a classic matrix format object of the prior art.

There illustrates a conversion of an initial image of sgpixels 2 recorded according to method 1 according to the invention and to be converted into a final image towards a 4 x 4 matrix of final pixels 7 or groups 7 of photogenerators 7 (filled squares) of a screen that shows the image of 2 sgpixels.

Among the sgpixels 2 of the initial image, we notice in particular three sgpixels referenced SGA, SGB and SBC successively along a line of sgpixels.

We assume each final pixel 7 capable of restoring a Red (R), Green (G), Blue (B) color by lighting emitters 71, 72, 73 proportionally to the values R, G, B of these final pixels.

The central sgpixel 2 is a so-called SGB sgpixel. For this SGB, the separation line (line) 3 is the separation of the sgpixel SGB, active, and visible (for example green in color, at the width represented directly by that of the line) separating SGB into two domains 31 and 32. D The other 2 sgpixels are represented in finer lines around.

A final pixel is called by its row and column number in this way:

(end pixel matrix row number, end pixel matrix column number)

We notice that by its total overlap with SGB, the final pixel in row 2, column 3 (final pixel (2, 3)) receives its colors only from the SGB sgpixel.

Let's assume domain 31 of the SGB sgpixel (called sg B-1) is red at 100% (the V and B at 0%).

Let's assume domain 32 of the SGB sgpixel (called sg B-2) is 100% blue (and 0% R and G). Line 3 of SGB is assumed to be 100% green (and 0% R and B).

Suppose SGB feature 3 represents 5% of the SGB surface, sg B-1 domain 65%, sg B-2 domain 30%.

Then the final pixel (2, 3) receives the values R, G, B at:

R = 100% * 65% = 65%

V = 100% * 5% = 5%

B = 100% * 30% = 30%

Let's assume the SGA sgpixel is only Red at 100% (no separation or other color).

The final pixel (2, 2) then receives:

R = 100% (mixture of the part of SGA and SGB in proportion to the respective surfaces, which here makes 100% red). V and B are at 0%

The final pixel (2, 1) also receives R = 100%, G and B = 0% directly from SGA.

Assume the blue SGC sgpixel at 95% (its other components at 0%).

Then the final pixel (2, 4) receives R = 0%, G = 0%, V = 15% * 100% + 85% * 95% = 95.75% (because it is assumed to have 15% overlap with SGB, 85 % with SGC).

These numerical examples translate a possible surface protorata rule to restore the intensities of final pixels to be displayed from the information of the sgpixel 2 image.

We notice that line 3 is restored here by mixing because the screen resolution does not allow for better. On the contrary, if the final pixels of the screen are extremely fine, which occurs when the initial image is enlarged by sgpixels 2 around the pattern SGA, SGB, SGC, then the line 3 of green color could correspond to a very green pixel below this green line with a little Red and Blue from neighboring SG B-1 and SG B-2. If we enlarge further, line 3 can be represented by an oblique row of completely green final pixels.

By way of illustration, the sgpixel SGE under the sgpixel SGB is represented with its active separation 3, but in a less saturated color (for example green), with a more vertical angle, connecting continuously with that above (this continuity which is not obligatory, but can be quite common if the two visible separations represent a real element such as a hair). The width of separation 3 of SGE is assumed to be reduced.

Note that we have significantly changed the color of the SGB between sg B-1 and sg B-2, and made separation 3 visible, in order to illustrate the rule of surface proportions, but in the majority of cases, on a natural image , it is frequent that if the visibility of a line 3 is active, the domains 30 will remain of a similar color, because a fine object, such as a hair on a face rarely corresponds to a change in background color (for example the skin under the hair). So the example here would rather be that of an artificial drawing with materialization of contours.

Spatial variability 4 has not been enabled to simplify the description here. It could have been on separations 3 to connect without a visible jump from slightly saturated green color to very saturated (which would show all the quality of restitution at very high magnification). It could have been between SGB part sg B-2 and SGC to take into account the transition from 100% to 95% on Blue. Note also that we can assume SGC with a blue gradient increasing towards the left, at 95% in the center, which can make an inheritance for the sgB-2 part taking a value of 100% thanks to this SGC gradient which enhances the blue from SGC, according to the Prol() extension operation previously mentioned. SgB-2 also inherits this gradient.

There is an illustration of a variant of conversion from sgpixels 2, which takes into account the exact position of the generators (also called emitters or photogenerators) red (R, referenced 71, and all obliquely hatched), green (G, referenced 72, and all hatched horizontally), and blue (B, referenced 73, and all hatched vertically) in each final pixel 7 of the screen. This pixelation is potentially fairer in terms of rendering, but requires knowing the details of the shape of the screen, which may not be the case, and that we would prefer the method illustrated in the .

So the is a representation similar to that of the , shown with the same sgpixel values SGA, SGB, SGC, etc. The R, G, B emitters of the final pixels are shown. The surface proration rules are applied taking into account whatever color of each pixel emitter is overlapping with the 2 sgpixels.

So, as the R of the final pixel (2, 3) is not completely in sg B-1 but mostly, it will receive its value here at 95% and not 65%, (assuming that the 3 green separation of SGB and the sg B-2 domain represents 5% of the surface facing the Red emitter of the final pixel (2, 3)). Likewise, the green of the final pixel (2, 4) will receive B=95% because it is completely under the SGC Blue sgpixel at 95% (and not B=95.75% as in the ).

Compression or decompression

The method 1 can also include a step of compressing or decompressing an initial image recorded according to the method previously described (and therefore comprising one or more sgpixel(s) 2): the initial image is respectively compressed or decompressed into an image respectively compressed or decompressed respectively by reducing the volume of information representing the compressed image thanks to compression or by increasing the volume of information representing the decompressed image thanks to decompression , the compression or decompression relating to at least one component or property of sgpixel 2 previously listed, preferably among the color(s) and/or the hue(s) and/or the brightness (of sgpixel 2, domain 30, and/or separation 3) and/ or the at least one separation line and/or the at least one spatial gradient (of color(s), hue(s) and/or luminosity for each sgpixel 2, domain 30, and/or separation 3) within each initial pixel considered.

We respectively compress or decompress the initial image so as to make its content directly viewable according to a method according to the visualization or display steps previously described, without additional step of transcription of the final image.

For the decompressed image, the property of separation 3 or gradient 4 is ‘directly’ usable, in the sense of figures 3 and 4.

Whereas for the compressed image, for example by wavelet, we do not directly have the separation value 3 or gradient 4 for each sgpixel, because it is obtained as a frequency component applying to several pixels.

Typically, we respectively compress or decompress the initial image, by means of:

• d’une transformation par Transformation en Cosinus Discret (DCT).

- a wavelet transformation of each component to be compressed, and/or

• of a transformation by Discrete Cosine Transform (DCT).

After compression or decompression, method 1 may include a search for identical sequences, and an indexing of these sequences and replacement of each repeated sequence by its index in a flow of data or image information.

Preferably, method 1 uses lossy compression on at least one (less critical) component among the color(s) and/or the hue(s) and/or the brightness and/or the at least one line of separation and/or the at least one spatial gradient of each initial pixel considered or the at least one separation within each initial pixel considered, and without loss on at least one other (more critical) component among activation or deactivation of the at least one separation 3, and/or visualization of the at least one separation 3 and/or width or viewing surface of the at least one separation 3.

In compression, in general, it is a question of considering as many additional channels to the spatial information for example R, G, B, and Alpha, the separation information, spatial gradient, and if applied, separation widths , its colors, its gradient 4 (Alpha being the transparency channel if existing).

Of course, for the object or file which contains the image recorded according to the invention and compressed, a global description field may be desirable to define the use or not of this or that element among separation, spatial variation, properties visualization of the separation, or any other method allowing you to know if these elements are part of the compressed image object. We can only consider the separation and the spatial variation, without the visibility of the separations, which will eliminate the need to have the color, width and gradient channels of the separations.

Except in special cases, we can preferably fill with a single value homogeneous throughout the image the positions/angles for separations 3 inactivated, in order not to create false information which would harm the compression and add unnecessary spatial frequencies in the image to be compressed, and the same goes for other unused channels.

Generally speaking, we can apply the compression techniques known for the R, G, B, Alpha channels to these additional channels which are separation 3 and spatial variation 4, separation visibility information 3, activation separation 3, etc.

Let us cite some compression techniques for this purpose. In the state of the art, we know image compressions by transformation generally by sub-blocks of the image, by projection of said block on DCT type bases (for “Discrete Cosine Transformation”), or by compressions by wavelets. The sequence produced by concatenating these projection coefficients on these bases and the different pixels is possibly further compressed by a compression type LZ77 or LZ78 or equivalent (LZ77 consists schematically of identifying and indexing identical sequences in position and length rather than of repeat and write explicitly, see https://fr.wikipedia.org/wiki/LZ77_et_LZ78). We also know entropic compressions, by replacing frequent symbols with shorter symbols.

LZ77 type compression can also be applied directly to the initial image without DCT or wavelet type compression.

Those skilled in the art may refer to the descriptions of the compression formats JPEG 2000, JPEG XR, JPEG XL, PNG, among others for the implementation of these types of compression.

These transformations can be lossless or lossy. When lossy compressions are implemented, high-frequency components can be modified for example by thresholding: if the amplitude of these components is too low, they can be reduced to zero in order to be able to ignore them (or compress them further). advantage). Likewise, the coefficients obtained can be quantized, that is to say we can voluntarily round the low binary weights to 0 and delete these bits, which reduces the number of bits of the coefficients resulting for example from DCT transformations. , or in wavelet, or directly from the initial image, and this means fewer bits to transmit, or it produces more identical sequences in an additional LZ77 type compression.

For method 1 according to the invention for which it is desired to apply compression to the image produced, in the case of using lossy compression, it is very much preferable to separate each coordinate of each piece of information (separation , spatial variation, and optional visibility of separations, etc.) in channels which will be compressed independently.

Indeed, for lossy compressions which can generally lead to modifying values by loss of low-order bits (the cancellation to zero of weak high-frequency components can also be assimilated), it is appropriate to ensure that these rounding of values would not excessively affect one of the properties added to sgpixels 2 by method 1. For example for a binary concatenation X:Y into a pseudo single number, the value Y could become significantly false if it underwent the setting to 0 bits considered to be of low order, more significantly than the component X which would be spared by these roundings. Thus, we recommend considering X and Y as 2 different channels, so that rounding of values affects, for example, the X and Y directions in a similar way, and the same for the sgpixel 2. Likewise for a polar representation distance 61, and angle 62, we will preferentially consider this information in 2 different channels, not concatenated in a binary manner.

If particular position values are used to inactivate separation 3, with lossy compression, it will be appropriate to represent this state of activity by another channel applied to compression, for example a Boolean field (with 2 values active/not active), in order to avoid any loss of this information following rounding coding, unless special precautions are taken to be robust.

Likewise, if separation visibility 3 is used, via code encoding that points to a table, and that code is presented directly at the compression input (rather than the property values it represents) , it is generally necessary to ensure that this code will never be corrupted by compression/decompression, unless special arrangements are made to be robust. We can, to better resist a change in the value of the code, store information close in value to boxes in the table of close indices, in order to obtain a small change in the values of the properties coming from the table at a change in value. code, therefore box in the table.

Of course, in the desirable case of compressing the channel of these codes for representing the visibility of separations, the values of the pointed table should preferably be recorded in global fields of the file or object of the image in order to be able to restore them with the decompressed image, later (and not assuming them to be universally known for all images).

Among the additional channels, if each sgpixel contains 2 or more sets of color properties, for example to define the colors on each side of the separation, they will of course also be treated as additional channels to be compressed / then decompressed .

The method 1 can include a change of reference for the information of separation 3, spatial variation 4, visibility of separation 3, etc.

Possibly, if the compression allows it due to the absence of risks of jumps linked to the loss of information, combinations between the separation channels, spatial variation, and/or visibility of the separation, etc. can be performed, particularly, but not necessarily, if lossless compression is used. These combinations can use correlations and thus reduce the amount of information to be compressed.

Let us recall for example that for the prior art, for R, G, B, we know for example the change of reference Y, U, V consisting of calculating for example Y = R+G+B (with possibly other weightings than 1, 1, 1 between colors), which provides a so-called Y luminance channel, and U and V chrominance channels. As the human eye is less sensitive to the spatial frequencies of U and V than those of Y, high spatial frequency information of these U and V components can be attenuated or removed in order to reduce the information to be compressed, upstream transformation by DCT or by wavelet, for example.

In the case of method 1 according to the invention, such combinations can therefore be made, if necessary, especially to remove redundancies of information.

For example, if the spatial variation 4 is represented in each sgpixel 2 by two components X, Y, we can consider that it is often perpendicular to the separation line 3, in the sgpixels 2 where it is active. So, we can simply present for compression the amplitude of the spatial variation 4 (preferably signed, the luminosity being able to increase or decrease in the direction of the separation 3), and not its angle, since it is known by that of separation 3. On decompression, the angle of gradient 4 will be obtained by perpendicularity to separation 3. Even if it is inactive, the angular information 62 of separation 3 can thus be interpreted to orient gradient 4 in the plane XY - so in sgpixel 2. This saves a direction field. The X and Y information of gradient 4 represents the spatial variability of sgpixel 4 and can then be replaced by .

Upon decompression, process 1 restores the different property channels of the sgpixels 2, in addition to the restitution of the color channels and possibly transparency, in order to fill the properties of all the sgpixels 2 of the image. If the compression/decompression is lossy, it is necessary to ensure that the rounding does not significantly modify the decompressed image obtained. For example, moderate inaccuracy of spatial variation 4 may be acceptable. But a movement of activation of separation 3 to a neighboring pixel would not necessarily be so. Likewise if an unintentional duplication of the activation of separation 3 occurs at pixels neighboring those where it is actually active.

Method 1 repeats for several or all the blocks of pixels of the initial image, if necessary, the decompression operations of each sub-block of pixels.

The method 1 according to the invention is implemented by a computer program product comprising computer code instructions for executing the method 1.

In general, each of the steps of method 1 is implemented solely by technical means, which may for example include at least one computer, a central or calculation unit, an analog electronic circuit (preferably dedicated), a digital electronic circuit (preferably dedicated), and/or a microprocessor (preferably dedicated), and/or software means, and/or the computer program product previously described.

Examples of areas of application of process 1

Process 1 applies more particularly to images from the real world, therefore to photographs or films. But it can of course also bring its advantages for images from synthetic objects, for example from the translation of vector image primitives in order, for example, to distribute them on a website in a format that does not require de-vectorization. locally, while allowing a zoom of better quality, than the translation of these vector images into bitmap allowed before the process of the invention, as was sometimes the case in the prior art.

Web images generally do not use vectorization as standard to avoid the risks of potentially variable rendering times depending on the complexity of stacking or density of objects. The sgpixel format according to the invention, by its conservation of the notion of matrix avoids this pitfall, which allows this format to be used for the Web, while guaranteeing restitution times controlled in advance, and this with the possibility better enlargements of parts of the images.

It is applicable for example to a computer or a smartphone (or “computer” or “smart phone”), to provide image enlargements of better quality than the prior art. To this end, a pixelization software module can translate the parts of the enlarged image into the pixels of the screen.

Likewise, a 'plug-in' specific to this task could be added to web browsing software to render and zoom an image of a web page according to method 1.

Example images

There illustrates a raster image according to the state of the art with little sampling.

There illustrates a raster image according to the invention with sgpixels according to the invention, better restored than that of the thanks to separations 3, gradients 4, and visualizations of separations 3 at certain locations in the image.

On the , we notice :

- in 91, the absence of the notion of gradient leads to jumps in the texture,

- in 92, the shadow of the metallic object makes crenellations appear at the border of this shadow,

- in 93, the very fine object (hair) reveals niches to represent it,

- in 94 (on the paperclip), the pixelation reveals unsightly jumps in texture, linked to the very fine luminous zone at the top of the wire.

On the obtained by sgpixels:

- gradients 4 make it possible to avoid color jumps in 91, the limit on the appearance of slots in 92 is no longer perceptible, and the thin object in 93 keeps a very linear shape without the appearance of slots,

- In 92, the separation of pixels makes it possible to keep a sharp edge (non-aliased) even at low pixel density,

- In 93, the visualization of the separation on several pixels allows this very linear, non-aliased form to appear,

- In 94, the texture brightness is more uniform. The very bright area at the top of the wire can be represented by a separation of a certain width,

- the edge of the thread gains sharpness thanks to the activation of the separation in the pixels which straddle the background fabric and the paperclip,

- the gradient allows a more homogeneous texture between the top of the wire and its edges.

• des propriétés de séparation 3, activable ou non dans le sgpixel 2, définissant au moins deux domaines 30 de couleur dans chaque sgpixel de l’image.
• des propriétés de variation spatiale 4, au sein de chaque sgpixel 2 ou domaine, des propriétés de couleur(s) et/ou teinte(s) et/ou luminosité.
• des propriétés de rendre visualisable ladite séparation 3.

Thus, to summarize method 1 according to the invention, for an image described by a two-dimensional table of pixels, describing for example a color field, for example R, G, B values, the method adds:

• separation properties 3, which can be activated or not in sgpixel 2, defining at least two color domains 30 in each sgpixel of the image.
• properties of spatial variation 4, within each sgpixel 2 or domain, properties of color(s) and/or hue(s) and/or brightness.
• properties of making said separation visualizable 3.

The proposed method 1 provides a method for representing an image digitally and restoring, recording and transmitting it. It allows the image to be enlarged to observe only part of it, without significant loss of quality due to the visible appearance of spatial discretization.

It allows a description of the image to be preserved by juxtaposition of pixels, and not by vectorization in the sense of the prior art. The recommended density of these (sg)pixels can remain moderate, for example equal to that of the sensors which made it possible to produce this image. It applies to photographs (single images) or films (sequential images). In particular, it is not necessary to increase the density of these sgpixels, whereas the prior art would need it for equivalent perceived quality to obtain a quality compatible with the possible enlargements.

Maintaining a notion of pixel juxtaposition allows for easier compression/decompression or display processing. Indeed, while the state of the art may require conversion of large surface objects before being able to carry out processing (such as display), method 1 allows display algorithms that can be parallelized more easily because the actions remain more local as for the 'bitmaps' of the prior art.

The quality of restitution of multiple details in textures is much higher than what the vectorized formats of the prior art allow intrinsically. Indeed, for multiple gradients in an object outline, a vectorized object does not allow a fine description, whereas sgpixels 2 allow it while keeping the sharp contours that the vectorized format allows.

In the sense of reducing the pixel density of the image, method 1 makes it possible to choose a density of sgpixels below that of the image sensor, or of the digitization grid, while maintaining a restitution of the good quality image, and while encoding in the image a significant part of the resolution provided by the density of photosites of the sensor, via the properties of the sgpixels. This already achieves a form of image compression compared to prior art. A method of compressing/decompressing the image of method 1 is also described in order to be able to transmit and store it at lower cost.

The sgpixel compression process allows compression ratios similar to those of the compression techniques applied to the prior art. A compressed sgpixel image is therefore more advantageous than one of the prior art thanks to its better quality.

This process 1 applies more particularly to images from the real world, therefore to photographs or films. But it can of course also bring its advantages for images from synthetic objects, for example from the translation of vector image primitives in order, for example, to distribute them on a website in a format that does not require de-vectorization. locally, while allowing a zoom of better quality, than the translation of these vector images into bitmap allowed before method 1 of the invention, as was sometimes the case in the prior art.

At equivalent matrix density compared to the state of the art, the details are finer in method 1 according to the invention. So by reducing the density of the sgpixel matrix, the process has less data in total. For example, for:

1/ ^5th of sgpixels in X in method 1 compared to the number of pixels in X of the state of the art,

1/ ^5th of pixels in Y in method 1 compared to the number of pixels in Y of the state of the art

i.e. 25 times fewer pixels, but:

2 to 3 times more information per sgpixel than per pixel,

which makes the ratio 2/25 to 3/25, therefore showing that the image of method 1 with sgpixels is intrinsically compressed compared to its equivalent quality in pixels of the state of the art.

Starting from this example of an image with 25 times fewer sgpixels, we can still have better definition in process 1 by zooming. We can convince ourselves of this by imagining a zoom on hair or fabric texture, for example: the hair remains hair at all zooms, whereas it becomes a juxtaposition of pixels for the classic raster image.

Compared to vectorization, method 1 controls a processing time for these sgpixel images of the same order as the raster image, while this time and quantity of calculations can be much higher for vectorization.

In addition to the natural compression effect of method 1 as demonstrated previously, this sgpixel format supports compression techniques, for example by wavelets and deletion of redundant patterns which are simply referenced, which further allows the size of the data to be reduced to quality. equivalent compared to the classic matrix equivalent.

Finally, process 1 allows a better rendering of the natural complexity of photographic images which can be restored with many 'flattening' effects than vectorization, or to put it another way, with all the finesse, and above all even better than raster images allow this.

Method 1 supports enlargement better than a state-of-the-art raster image, particularly for an image of writing.

Of course, the invention is not limited to the examples which have just been described and numerous adjustments can be made to these examples without departing from the scope of the invention.

Of course, the different characteristics, shapes, variants and embodiments of the invention can be associated with each other in various combinations as long as they are not incompatible or exclusive of each other. In particular, all the variants and embodiments described above can be combined with each other.

Claims (27)

Method for recording a digital image in a computer format, said image or at least part of the image being encoded and recorded in the form of a matrix of pixels in which each pixel considered among at least one or more pixels of the matrix includes:

• at least one separation line within the pixel considered, separating the pixel considered at least into two areas, and/or
• at least one spatial gradient of color and/or hue and/or luminosity within the pixel considered or at least one of the domains of the pixel considered.

Method according to claim 1, characterized in that each separation line within the pixel considered is coded by two pieces of data comprising:

• Two points on the edge of the pixel considered, and/or
• A distance, preferably relative to the center of the pixel considered, and an angle.

Method according to any one of the preceding claims, characterized in that separation lines of several neighboring pixels are assembled so as to form a continuous line crossing these several neighboring pixels. Method according to any one of the preceding claims, characterized in that the at least one gradient within the pixel considered or at least one of the domains of the pixel considered comprises:

• A brightness gradient, and/or
• One or more gradients corresponding respectively to one or more color(s) and/or shade(s).

Method according to any one of the preceding claims, characterized in that the at least one gradient within the pixel considered or at least one of the domains of the pixel considered is coded by:
– two spatial components, or

• A single spatial component relative to the at least one separation line within the pixel considered.

Method according to any one of the preceding claims, characterized in that at least one brightness and/or color and/or hue within the pixel considered or at least one of the domains of the pixel considered depends on a brightness and /or color and/or hue of at least one pixel neighboring the pixel considered or of at least one domain of a pixel neighboring the pixel considered. Method according to any one of the preceding claims, characterized in that at least one gradient of brightness and/or color and/or hue within the pixel considered or at least one of the domains of the pixel considered depends on 'a gradient of brightness and/or color and/or hue of at least one pixel neighboring the pixel considered or of at least one domain of a pixel neighboring the pixel considered. Method according to any one of the preceding claims, characterized in that all the pixels of the pixel matrix have the same dimensions. Method according to any one of the preceding claims, characterized in that the at least one separation line has a state that can be viewed with a viewing surface. Method according to claim 9, characterized in that the at least one separation line itself comprises a spatial gradient making it possible to modulate, when it is visible, its luminosity(s) and/or colors and/or hues. Method according to claim 9 or 10, characterized in that the at least one separation line is switched between its displayable state and a non-displayable state. Method according to any one of the preceding claims, characterized in that the at least one separation line is switched between an active state and an inactive state. Method according to claim 12, characterized in that it comprises a deactivation of the at least one separation line of the pixel considered if this pixel considered is sampled below a certain spatial sampling threshold. Method according to any one of the preceding claims, characterized in that it is implemented in an embedded manner in a software module which is added to or integrated into a Web browser, for taking into account the recorded images, possibly their compression or decompression, and/or their visualization in part or all of a screen interacting with the web browser. Method according to any one of the preceding claims, characterized in that:
- textual information represented by pointing symbols in a character font, and/or
- drawings of objects embedded in the form of raster or vectorized sub-images
are added directly to the recorded image or added during its visualization, in order to be viewable, possibly according to a particular selection which only shows all or part of it, during a restitution of the recorded image. Method according to any one of the preceding claims, characterized in that metadata is added to the recorded image. Process :
- displaying an initial image in a so-called final image on a screen comprising groups of photogenerators, or
- pixelation or rasterization of said initial image into a final image comprising final pixels devoid of separation line and/or spatial gradient of color and/or hue and/or brightness within these final pixels,
said initial image being recorded according to any one of claims 1 to 16,
characterized in that it includes:

• For each final pixel considered or each group of photogenerators considered, a transcription of at least initial pixel or at least one initial pixel domain into this final pixel or group of photogenerators considered:
  • from the brightness and/or color and/or hue components of this at least one initial pixel or at least one initial pixel domain
  • towards this final pixel or group of photogenerators considered,

the transcription being carried out in proportion to this at least initial pixel or at least one initial pixel domain within a common surface facing between:

• this at least initial pixel or at least one initial pixel domain and
• the final pixel or photogenerator group considered.

Method according to claim 17, characterized in that, for transcription, if the final pixel or group of photogenerators considered is opposite different initial pixels and/or domains of the initial pixels, its components of brightness and/or color and/or hue are a weighted average in proportion to the respective surfaces of these different initial pixels and/or domains of the initial pixels with respect to the final pixel or group of photogenerators considered. Method according to claim 17 or 18, characterized in that, for transcription, the brightness and/or color and/or hue components of an initial pixel are modulated by the spatial gradient values of each part of the initial pixel domain next to the final pixel or the group of photogenerators considered. Method according to any one of claims 17 to 19, characterized in that a separation line of an initial pixel is used to be visualized and contribute to the brightness and/or colors and/or hues of at least final pixel or group of photogenerators considered, in proportion to a viewing surface of at least part of the separation line facing the final pixel or group of photogenerators considered. Method of compressing or decompressing an initial recorded image according to any one of claims 1 to 16, characterized in that the initial image is respectively compressed or decompressed into a respectively compressed or decompressed image respectively by reducing the volume of information representing the compressed image thanks to compression, or by increasing the volume of information representing the decompressed image thanks to decompression , the compression or decompression relating to at least one component among the color(s) and /or the hue(s) and/or the brightness and/or the at least one separation line and/or the at least one spatial gradient within each initial pixel considered. Method according to claim 21, characterized in that the initial image is respectively compressed or decompressed so as to make its content directly viewable according to a method according to any one of claims 17 to 20. Method according to claim 21 or 22, characterized in that the initial image is respectively compressed or decompressed, by means of a wavelet transformation of each component to be compressed. Method according to any one of claims 21 to 23, characterized in that the initial image is respectively compressed or decompressed, by means of a transformation by Discrete Cosine Transformation (DCT). Method according to any one of claims 21 to 24, characterized in that it comprises, after compression or decompression, a search for identical sequences, and an indexing of these sequences and replacement of each repeated sequence by its index in a stream data or information from the image. Method according to any one of claims 21 to 25, characterized in that it uses lossy compression on at least one component among the color(s) and/or the hue(s) and/or the brightness and/ or the at least one separation line and/or the at least one spatial gradient of each initial pixel considered or the at least one separation within each initial pixel considered, and without loss on at least one other component among activation of the at least one separation, and/or visualization of the at least one separation and/or width of the at least one separation. Computer program product comprising computer code instructions for executing a method according to one of claims 1 to 26 when said program is executed on a computer.