RU2504840C1 - Automatic photograph retouching method - Google Patents

Abstract

FIELD: information technology.SUBSTANCE: automatic photograph retouching method involves creating a data array from photographs of different themes of classes, creating a database of references therefrom by interactive processing, based on Photoshop CS2, of predefined textures which give a comfortable perception of images of objects on the photographs, constructing a function of a photometric correction signal between the original and reference photographs, determining barcodes of the original photograph and the reference photograph by decoding brightness I(x,y) of matrices of images with the size |m×n| of elements in the matrix of intensities of tonal transitions with dimensions ||k×k|| of elements of the original and reference retouched photographs, algebraic subtraction of the matrix of barcodes while setting a threshold for positive identification of the reference, creating a reference address with barcode extension therefrom, a threshold difference and a photometric correction signal function, automatic search of the reference for the analysed photograph and retouching thereof based on the calculated barcode at the address in a reference database and the photometric correction signal function.EFFECT: automatic search for the reference for a processed photograph by creating code features of each reference located in a reference database, and subsequent automatic photometric correction of the processed photograph on a reference texture mask stored in the reference database.4 dwg

Description

The invention relates to computer science and can find application in systems for the collection, conversion, processing of information in various fields of human activity: forensics, astronautics.

Remote sensing of the Earth from space in the interests of mineral exploration, forestry, ecology, monitoring of soil cover is carried out by obtaining digital images of the underlying surface. Selective features of objects in the images are color, tone, texture, topology. Environmental monitoring is carried out, as a rule, by obtaining spectrozonal images on a three-layer photocarrier with sensitive layers of R, G, B ranges.

The well-known "Method for assessing atmospheric pollution", Patent RU No. 2117286, 1998, A01G 23/00 - analogue. The analogue method includes converting the spectral brightness of the image I (x, y) into digital matrices | m × n | elements in the G, R sections of the visible spectrum, elementwise logical sorting of pixels in both matrices in accordance with the algorithm: if R> G, then R, and if R <G, then R = R _max - | k | · G, where k - the correlation coefficient of chromatic coefficients r, g, get the resulting matrix of the same size, calculate the numerical characteristics of the electrical signal of the resulting matrix: mathematical expectation, variance, envelope of the spatial spectrum, calculate the histogram of the distribution of pixels by brightness, carry out the binding of the obtained relative Foot distribution law to the absolute values of the state of the atmosphere in the region of the index from its values and Square reference sites.

The disadvantage of the analogue is the dependence of the correlation coefficient of chromatic coefficients r, g on the shooting conditions and the inadequacy of the algorithm for logical sorting of pixels to the physical process, which distorts the final histogram of the distribution of pixels by brightness and the accuracy of the estimated parameter estimation.

The closest analogue of the claimed technical solution is a method of retouching images interactively [see, for example, P.A. Minko, “Processing of Graphics in Photoshop GS2, ed. Eksmo, 2007, chap. 7. “Image Adjustment”, pp. 170-186]. In the closest analogue method, the retouching area is set by pre-selecting it, then the selected area is filled with a visually selected operator, a separate shade from the set of predefined textures offered by Photoshop in the standard BGR color palette or in other CMYK, Lab palettes, giving a comfortable perception of the image of the object.

The disadvantages of the closest analogue can be considered:

- the absence of a code feature that allows you to automatically select a reference image for the processed image;

- the subjectivity of the comfortable perception of the image of the object, depending on the operator.

The problem solved by the claimed invention is to automatically search for a reference for the processed image by creating code features of each reference located in the reference database, and subsequent automatic photometric correction of the processed image on the mask texture mask stored in the reference database.

The technical solution to the problem is achieved by the fact that the method of automatic retouching of images includes creating an array of data from images of various subjects and classes, forming a database of standards from them by processing "in interactive mode" based on specialized software Photoshop CS2, pre-installed textures that give a comfortable perception of images of objects in the pictures, the construction of the photometric correction signal function between the source and reference pictures, the definition of "barcodes" of the original picture and this bosom by recoding the brightness I (x, y) of image matrices of size | mxn | elements in the matrix of intensities of tonal transitions of dimension || kxk || elements of the initial and reference retouched images, algebraic subtraction of “barcode” matrices with a threshold for reliable identification of the standard, the formation of the standard address with the extension of its “bar code”, the threshold difference and the function of the photometric correction signal, automatic search for the standard for the analyzed image and its retouching based on the calculated “barcode” at the address in the base of standards and the function of the photometric correction signal.

The invention is illustrated by drawings, where:

figure 1 - the functions of the signals of photometric correction:

a) histograms of brightness of pixels of the current 1 and retouched 2 images;

b) the standard (optimal) curve for increasing the contrast of most pictures of the closest analogue;

figure 2 - visualized images of the matrix of "barcodes":

a) a reference retouched image, b) an initial image, c) algebraic subtraction of the “barcode” matrices;

figure 3 is a functional diagram of a device that implements the method;

figure 4 - the result of automatic retouching of pictures:

a) the analyzed image; b) automatically retouched image.

The technical essence of the method is as follows.

Interactive photo image enhancement is widely used in various photo editors [see, for example, P.A. Minko, “Graphics Processing in Photoshop CS2, ed. Eksmo, 2007, pp. 71-89, pp. 145-151]. However, the level of automation of photo-correction processes is still very low: the person who is the photo editor is forced to choose the optimal parameters for transforming photo images in accordance with his artistic taste. Automation consists in writing special programs (scripts and actions) in languages perceived by photo editors, which can carry out a fixed sequence of transformations when the corresponding program is launched.

While there are no universal methods for automatic optimization of the basic parameters of photographs, since the choice of algorithms and parameters for such optimization depends on the plot of the images. For example, the menu for general and selective contrasting of images using low-pass and high-pass filtering of the closest analogue is represented by a gallery including more than 80 filters (see the closest analogue, pp. 225-237).

The tools for interactive retouching of pictures of the closest analogue are:

- correction of histograms of the distribution of pixels by brightness, Histogram command, Photoshop CS2, p.171, illustrated in figure 1;

- increasing contrast, general dimming or lightening of the image, Levels command, p.172;

- Selective contrast adjustment using the filter gallery, Curves command, Adjustment group, Filter menu, pp. 174, 225-229.

In the claimed method, for automatic retouching of images, a code conversion of the matrix of the original image, the brightness function I (x, y) of dimension | m × n | elements into a “barcode” matrix (matrix of frames) of size || k × k || elements, where k is the maximum brightness value. The frame matrix is not a reversible transformation of the original image, but it carries complete information for recognizing the plot of the image.

The transcoding algorithm includes the following procedures:

- define the matrix | k × k | barcode elements, where k is the maximum brightness value;

- choose a fixed aperture window with a dimension of 2 elements per line;

- set the scan cycle of the matrix within 1≤i≤m; 1≤j≤n, where i is the row number, j is the column number;

- cyclically select two adjacent elements x (i, j) and x (i, j + 1) of the original matrix, and if the value x (i, j) = a, x (i, j + l) = b, then the element barcode matrices with indices (a, b) are increased by 1;

- display the elements of the matrix | k × k | that are not equal to zero, in the form of nodes of the "barcode" matrix, the sizes of which are proportional to the accumulated values of the kth brightness;

- calculate the diagonal element of the matrix of "barcodes"

The above procedures, as well as the algorithm for subtracting the “barcode” matrix of the current image from the standard, are implemented on the basis of specialized programs in the Turbo-Pascal language.

The program for obtaining matrices of "barcodes".

Matrix difference (difference matrix) can be obtained using a simple algorithm

Based on the results of the above operations, the address of the standard is formed with the extension of its “bar code”, threshold difference and the function of the signal of photometric correction for this class and plot of the analyzed images. The procedure for implementing automatic retouching is considered in the example of a specific implementation.

An example implementation of the method.

The claimed method can be implemented according to the scheme of figure 3. Functional diagram of the device of Fig. 3 contains a flash card 1, a set of client digital photo images in one of the commonly used formats (JPG, TIFF, RAW, etc.), a hardware-software system of the Internet 2, implemented on a microprocessor base, a digital converter 3, forming digital photo images (client or reference) special matrix of image frames, RAM 4, containing a matrix of frames of client photo images, ROM 5, containing reference digital photo images in one of the commonly used formats, ROM 6, containing m Assist of matrix frames of reference photo images, a digital analyzer-discriminator 7 implemented on a microprocessor base, which determines, by pairwise comparison of frame matrices for each client photo image, the nearest reference photo image and the measure of difference between the client photo image and the reference image, a digital converter 8 implemented on the microprocessor base that corrects client digital images taking into account the differences between each client’s photo image and the corresponding reference image, RAM 9, containing adjusted client digital photographs in one of the commonly used formats. All elements of the device are implemented on the existing technical base, Intel-type PCs. Previously, in ROM 5, specialized Photoshop QS2 software is installed with global access to the involved commands of the photometric correction signal functions: Histogram, Levels, Curves. In RAM 4 install a specialized program for calculating the matrix of "bar codes" of images.

The pixel intensities (0..255) of the image are in the A1.txt text file. The intensities are read from the file line by line in pairs. A pair of "neighboring" intensity values, divided entirely by 16, determines the coordinates of a 16x16 matrix, called a barcode. The value of the element of the barcode matrix is an integer corresponding to the number of horizontal pairs of image pixels with intensities whose values coincide with the coordinates of the element. The barcode matrix is written line-by-line to the SCod.txt text file.

The matrix "barcode" obtained from the image shown in figa.

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 123 3 170 one 0 0 0 0 0 0 0 0 0 0 0 0 2 0 eleven 0 0 0 0 0 0 0 0 0 0 0 0 0 158 7 7199 1151 720 0 3 0 0 0 0 0 0 0 0 0 2 one 1027 2081 1508 one 13 0 2 0 0 0 0 one 0 0 3 one 638 1301 42416 2187 2449 13 32 0 13 2 0 four 0 0 one 0 one four 2032 3814 2181 four 7 0 one one 0 2 0 0 four 0 56 22 2127 2039 47525 2397 189 one 112 28 0 108 0 0 one 0 10 8 36 0 2197 19244 2898 0 32 8 0 51 0 0 one one 9 6 33 3 102 2768 59833 8 180 29th 0 165 0 0 0 0 0 0 0 0 0 0 four one four 0 0 one 0 0 0 0 16 10 thirty one fourteen four 115 0 156 fourteen 0 221 0 0 one 0 52 fifteen eleven 0 12 3 2 0 7 one 0 82 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 one 0 49 37 146 3 112 52 56 0 76 103 0 56002

A visualized image of the matrix "barcode" of the original image (Fig. 2A) is illustrated in fig.2b.

The barcode matrix obtained from the original image by using the Curves optimal photometric correction command.

471 0 four 0 55 12 8 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 four 0 107 3 152 one 0 0 0 0 0 0 0 0 0 0 0 0 2 0 10 0 0 0 0 0 0 0 0 0 0 0 56 0 140 6 6740 1121 718 0 3 0 0 0 0 0 0 0 12 0 2 one 1004 2048 1498 one 13 0 2 0 0 0 0 one 7 0 3 one 630 1298 42409 2187 2449 13 32 0 13 2 0 four 0 0 one 0 one four 2032 3814 2181 four 7 0 one one 0 2 0 0 four 0 56 22 2127 2039 47525 2397 189 one 112 28 0 108 0 0 one 0 10 8 36 0 2197 19244 2898 0 32 8 0 51 0 0 one one 9 6 33 3 102 2768 59833 8 180 29th 0 165 0 0 0 0 0 0 0 0 0 0 four one four 0 0 one 0 0 0 0 16 10 thirty one fourteen four 115 0 156 fourteen 0 221 0 0 one 0 52 fifteen eleven 0 12 3 2 0 7 one 0 82 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 one 0 49 37 146 3 112 52 56 0 76 103 0 56002

A visualized image of the matrix of the "bar code" of the retouched image is illustrated in figa.

To establish the "correspondence" of the original image to the standard, algebraic subtraction of matrices is performed to obtain a difference matrix. The compliance measure is chosen by setting a certain threshold according to the criterion N <E, where N is the accepted norm of the difference matrix, for example, the maximum absolute value, E is the number set by the expert. Obviously, for pictures of different classes and scenes, the number (E) can vary within certain limits and should be established experimentally. For the analyzed source image of FIG. 2a, the difference matrix took the following form.

-471 0 -four 0 -55 -12 -8 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 -four 0 16 0 eighteen 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 one 0 0 0 0 0 0 0 0 0 0 0 -56 0 eighteen one 459 thirty 2 0 0 0 0 0 0 0 0 0 -12 0 0 0 23 33 10 0 0 0 0 0 0 0 0 0 -7 0 0 0 8 3 7 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

The visualized image of the difference matrix of "barcodes" is illustrated in figv. The database of standards can be a relational database, the fields of records of which consist of a pointer (address of the standard) as part of a “barcode” matrix, a classifier (measure of coincidence of a difference matrix), and extensions (Curves commands — photometric correction signal functions).

The effectiveness of the method is characterized by the ability to automatically retouch images with a different plot with high reliability of identification and quality of processing.

Claims (1)

The method of automatic retouching of images includes creating an array of data from images of various subjects and classes, forming a database of standards from them by processing "in interactive mode" based on specialized software Photoshop CS2, predefined textures that provide a comfortable perception of images of objects in images, building a photometric signal function corrections between the source and reference images, the definition of "barcodes" of the original image and the standard by recoding the brightness I (x, y) of the matrices from size expressions | m × n | elements in the matrix of intensities of tonal transitions with dimension || k × k || elements of the initial and reference retouched images, algebraic subtraction of “barcode” matrices with a threshold for reliable identification of the standard, the formation of the standard address with the extension of its “bar code”, the threshold difference and the function of the photometric correction signal, automatic search for the standard for the analyzed image and its retouching based on the calculated “barcode” at the address in the base of standards and the function of the photometric correction signal.

Images