RU2015561C1 - Device for aperture correction of images - Google Patents

Abstract

FIELD: automatics; computer technology. SUBSTANCE: device may be used for preliminary image processing, when images have strongly pronounced spatial nonuniformity. Device has delay units 1, 2, adders 3, 4, subtracters 6, 7. Accumulator 5, subtracter 8 and multiplication unit 9 are brought into the device additionally. Image may be adaptively converted in conveyer mode with local statistic characteristics allowed for. EFFECT: improved precision of the device. 3 dwg

Description

 The invention relates to automation and computer technology, in particular to devices for image processing, and can be used for pre-processing in order to increase their sharpness and contrast.

 A device for converting images of objects [1], comprising a counter, a first switch, a pulse distributor, a second switch, a first memory block, a first register, a first adder, a second memory block, a second register, a second adder.

 This device performs gradational image correction using the histogram alignment method. In this case, the brightness range of the original image is subjected to non-linear transformation in such a way as to bring the histogram of the processed image to a uniform form (i.e., the distribution of brightness levels in the processed image should have a uniform appearance).

 The disadvantages of the device are the high level of hardware costs and low accuracy when processing images that have a significantly pronounced spatial heterogeneity, since it does not allow for local static connections in the image when processing it.

 Closest to the proposed is a device for converting images of objects [2], containing two registers, a shift register block, an extreme filtering unit, two average value calculation units, an adaptive filtering unit, and the average value calculation unit contains a first delay unit, the first group of outputs of which connected to the first group of inputs of the first subtractor, the outputs of which are connected to the first group of inputs of the first adder, the second delay unit, the first group of outputs of which is connected to howling group of inputs of the second subtractor, a second adder, the clock inputs of the delay units are combined.

 The prototype device implements the aperture correction of the image in order to increase its local contrasts (detail) by non-linear suppression of the background component of the video signal and subsequent linear amplification of the previously corrected signal. For this, at each point in the image, estimates of the upper and lower boundaries of the signal (signal span) are formed in some of its local neighborhoods. The filtering parameters at each point in the image (the degree of suppression of the background component in the video signal and its gain) are a function of the magnitude of the values of the video signal in some local neighborhood of the current point. The prototype device allows to increase local contrasts of spatially uniform images.

 The disadvantage of the prototype device is the low accuracy of image conversion with pronounced spatial heterogeneity. This is explained by the following. First, estimates of the upper and lower boundaries of the video signal in the local neighborhood of the current image point are formed on the basis of the results of the extreme filtering operation. The operation of extreme filtering has a strong spatial inertia, which does not allow you to quickly adjust the parameters of the nonlinear filter. Using the results of the extreme filtering operation to adjust the parameters of a nonlinear filter leads to a significant decrease in filtering efficiency in the presence of high-amplitude pulsed noise, or a larger number of small-sized objects. Secondly, in non-linear image filtering, only the magnitude of the video signal is taken into account and its local average value is completely not taken into account. Image sections containing various objects characterized by different levels of the local average signal (background) and having the same magnitude of the video signal are processed by the filter with the same parameters, which leads to low accuracy of the generated results for images that have significant spatial heterogeneity.

 The aim of the invention is to improve the accuracy of the device when processing images with significantly pronounced spatial heterogeneity.

 The goal is achieved in that in the device for aperture correction of images containing the first delay unit, the first group of outputs of which is connected to the first group of inputs of the first subtractor, the outputs of which are connected to the first group of inputs of the first adder, the second delay unit, the first group of outputs of which is connected to the first the group of inputs of the second subtractor, the second adder, and the clock inputs of the delay units are combined, introduced the third subtractor, accumulating the adder and the multiplication unit, the clock input of which is combined the clock inputs of the accumulating adder and the first delay block is the device’s clock input, while the information inputs of the first delay block are combined with the corresponding inputs of the second group of the first subtractor and are the information inputs of the device, the second group of outputs of the first delay block is connected to the first group of inputs of the third subtractor, the outputs which are connected to the information inputs of the multiplication unit, the inputs of which are connected to the first group of inputs of the second adder, the outputs of the first sum The ators are connected to the information inputs of the second delay unit, the second and third groups of outputs of which are connected to the second groups of inputs of the first adder and the second subtracter, the outputs of which are connected to the information inputs of the accumulating adder, the outputs of which are connected to the second groups of inputs of the third subtractor and the second adder, outputs which are the outputs of the device.

 In FIG. 1 shows a diagram of a device; in FIG. 2 shows the relative position of the elements of the current aperture NxM and the elements of the previous aperture (shaded), differing from each other by two columns of N elements in each; in FIG. 3a shows an example of the original signal X (t), in FIG. 3b is a smoothed signal S (t), in FIG. 3c shows the difference between the original signal X (t) and the smoothed signal S (t), in FIG. 3d - the result of Y (t) aperture correction operation.

 The device comprises delay units 1 and 2, adders 3 and 4, an accumulating adder 5, subtractors 6, 7, and 8, a multiplication unit 9, information input 10, clock input 11, output 12.

 The information input 10 of the device is connected to the second group of inputs of the subtractor 6 and to the information inputs of the delay unit 1. The first group of outputs of delay unit 1 is connected to the first group of inputs of subtractor 6. The outputs of subtractor 6 are connected to the first group of inputs of adder 3. The second group of inputs of adder 3 is connected to the second group of outputs of delay unit 2. The outputs of the adder 3 are connected to the inputs of the delay unit 2. The first and third groups of outputs of the delay unit 2 are connected respectively to the first and second groups of inputs of the subtractor 7. The outputs of the subtractor 7 are connected to the inputs of the accumulating adder 5. The outputs of the accumulating adder are connected to the second group of inputs of the subtractor 8 and the second group of inputs of the adder 4. The second group of outputs of the block 1 delay connected to the first group of inputs of the subtractor 8. The outputs of the subtractor 8 are connected to the inputs of the block 9 multiplication. The outputs of the multiplication unit are connected to the first group of inputs of the adder 4. The outputs of the adder 4 are the outputs 12 of the device.

 The delay unit 1 has a length of N˙L, with its first and second outputs connected to the outputs of the N˙L and (N / 2˙L + + M / 2 + 1) -th delay elements for one clock cycle (N˙M - size moving aperture, L is the number of samples in the image line). The delay unit 2 has a length L, and its first, second and third outputs are connected to the outputs of the (M + 1) th, Lth and first delay elements for one clock cycle.

 The proposed device implements aperture correction of the image based on the analysis of the values of the source image at the current point and the local average signal value in its certain neighborhood. In this case, the original image (Fig. 3a) is smoothed (Fig. 3b), i.e. an image is formed containing only the low-frequency component corresponding to the background. Then, the obtained smoothed image is subtracted from the original one (Fig. 3c), i.e. the background component is subtracted from the original image. The result is an image containing only objects without a background, which is then amplified and summed with a smoothed image (Fig. 3d). As a result, the contrast and sharpness of the processed image increase, which leads to an increase in its visual quality. The described method (the method of unsharp masking) is widely used in photography and can improve the efficiency of processing images with significantly pronounced spatial heterogeneity.

X_ij-S

, (1) где X_ij - значение исходного изображения в точке (i,j) изображения;
Y_ij - значение преобразованного изображения в точке (i,j) изображения;
S_i+N/2;j+M/2 - локальное среднее значение в точках локальной окрестности размерами N.M точек с центральной точкой (i,j) (нижний правый угол которой находится в точке (i+N/2; j+M/2));
α - весовой коэффициент, задается априори в пределах от 1,5 до 5.The device implements a function calculation
Y _ij = S _{i + N / 2; j + M / 2} +

X _ij -S

, (1) where X _ij is the value of the original image at the point (i, j) of the image;
Y _ij is the value of the transformed image at the point (i, j) of the image;
S _{i + N / 2; j + M / 2} - local average value at points of a local neighborhood with the size of NM points with a central point (i, j) (whose lower right corner is at the point (i + N / 2; j + M / 2));
α - weight coefficient, is set a priori in the range from 1.5 to 5.

The calculation of the local average value S _ij over a moving rectangular neighborhood (aperture) with dimensions NxM is carried out recursively.

b_β , (4) где b_β - сумма значений элементов текущей скользящей апертуры размерами N.M и центром в точке (i-N/2;j-M/2), (β = i˙L+j);
a_β - сумма значений элементов текущего столбика высотой N и нижней точкой (i,j);
L - количество точек в строке изображения.Suppose that the image has L points in each row and the sum of b _β-1 values of the previous neighborhood of a _β-1 with the dimensions of NM elements is known (see Fig. 2, elements of the neighborhood of a _{β-1 are} shaded). The current neighborhood of a _β differs from the neighborhood of a _{β-1 by} two columns of N elements in each. At the same time, the current β-th column differs from the column located one point above, also two points. The following conclusions can be drawn from here.
b _β = b _β-1 + a _β -a _β-M ; (2)
a _β = a _β-L + X _ij - X _{iN, j} ; (3)
S _ij =

b _β , (4) where b _β is the sum of the values of the elements of the current moving aperture with dimensions of NM and the center at the point (iN / 2; jM / 2), (β = i˙L + j);
a _β is the sum of the values of the elements of the current column with a height N and a lower point (i, j);
L is the number of points in the image line.

 This approach to the formation of the values of the processed image allows you to take into account the statistical features of the original image and to increase the accuracy of processing images with significant spatial heterogeneity.

 The device operates as follows.

 At the initial time, the delay blocks 1 and 2 and the accumulating adder 5 are reset.

In the current k-th operation cycle, a parallel code of the current sample X _{ij of the} image is received at the information input 10 of the device, which then goes to the second group of inputs of the subtractor 6 and the inputs of the delay unit 1. On the first group of outputs of block 1, a count value X _iN delayed by N˙M clock cycles is formed, which then goes to the first group of inputs of the subtractor 6. At the outputs of the subtractor 6, a value of the value X _ij - X _iN is generated, which goes to the first group of inputs of the adder 3 , the second group of inputs of which receives the value of a _kL from the second group of outputs of block 2 of the delay.

At the outputs of adder 3, the code a _k = a _kL + X _ij - X _{iM, j} (according to expression (3)) is generated, which then goes to the inputs of delay unit 2. In parallel, the values of a _k-1 and a _kM-1 from the third and first groups of outputs of the delay unit 2 are sent to the second and first groups of inputs of the subtractor 7. Code a _k-1 - a _kM-1 from the outputs of the subtractor 7 goes to the inputs of the accumulating adder 5, in which the value is formed
b _k-1 = b _k-2 + a _k-1 - a _kM-1 (according to
expression (2)).

b_γ с выходов накапливающего сумматора 5 поступает на вторую группу входов сумматора 4 и вторую группу входов вычитателя 8, на первую группу входов которого поступает значение отсчета X_i-N/2,j-M/2 с второй группы выходов блока 1 задержки. Код X_i-N/2,j-M/2 - S_ij с выходов вычитателя 8 поступает на входы блока 9 умножения, в котором осуществляется ее умножение на коэффициент α. Значение α (X_i-N/2,j-M/2 - S_ij) с выходов блока 9 умножения поступает на первую группу входов сумматора 4, на выходах которого формируется согласно выражению значение
Y_i-N/2,j-M/2 = S_ij + α (X_i-N/2,j-M/2 - S_ij), которое затем поступает на выход 12 устройства. При поступлении очередного тактового импульса с тактового входа 11 устройства в блоках 1 и 2 задержки осуществляется прием и сдвиг информации, срабатывают накапливающий сумматор 5 и блок 9 умножения и начинается новый такт работы устройства.Thus, in each clock cycle in the accumulating adder 5, the value of the corresponding sum b _{γ of the} elements of the current sliding aperture is formed. The calculation of the value S _{ij of the} local average over the current moving aperture is carried out by dividing the value of b _γ by the value M˙N, which is carried out by working out the τ least significant bits from the output of the accumulating adder (with τ = log ₂ (MN) = log ₂ M + log ₂ N ) The value of S _ij =

b _γ from the outputs of the accumulating adder 5 enters the second group of inputs of the adder 4 and the second group of inputs of the subtractor 8, the first group of inputs of which receives the reference value X _{iN / 2, jM / 2} from the second group of outputs of the delay unit 1. The code X _{iN / 2, jM / 2} - S _ij from the outputs of the subtractor 8 is fed to the inputs of the multiplication unit 9, in which it is multiplied by the coefficient α. The value α (X _{iN / 2, jM / 2} - S _ij ) from the outputs of the multiplication unit 9 is fed to the first group of inputs of the adder 4, the outputs of which are formed according to the expression value
Y _{iN / 2, jM / 2} = S _ij + α (X _{iN / 2, jM / 2} - S _ij ), which then goes to the output 12 of the device. Upon receipt of the next clock pulse from the clock input 11 of the device in blocks 1 and 2 of the delay, information is received and shifted, the accumulating adder 5 and the multiplication block 9 are triggered, and a new clock cycle of the device starts.

 The proposed device implements aperture correction of the image by adaptive suppression of the low-frequency component of the signal corresponding to the background, and amplification of the high-frequency component corresponding to the details of objects in the image. The conversion of the original image is carried out adaptively taking into account its local statistical characteristics. This allows to increase the accuracy of the device when processing images with pronounced spatial heterogeneity.

 The device operates in a conveyor mode and has high speed, which allows it to be effectively used in high-speed specialized systems, for example, in real-time image processing systems.

Claims (1)

 DEVICE FOR APERTURE CORRECTION OF IMAGES, containing the first delay unit, the first group of outputs of which is connected to the first group of inputs of the first subtractor, the outputs of which are connected to the first group of inputs of the first adder, the second delay unit, the first group of outputs of which is connected to the first group of inputs of the second subtractor, second the adder, the clock inputs of the delay units are combined, characterized in that, in order to improve the accuracy of the device when processing images with a pronounced spatial heterogeneity by affinity, a third subtractor is introduced into the device, accumulating an adder and a multiplication unit, the clock input of which is combined with the clock inputs of the accumulating adder and the first delay unit and is a device clock signal, the information inputs of the first delay unit are combined with the corresponding inputs of the second group of the first subtractor and are information inputs devices, the second group of outputs of the first delay unit is connected to the first group of inputs of the third subtractor, the outputs of which are connected to information the inputs of the multiplication unit, the outputs of which are connected to the first group of inputs of the second adder, the outputs of the first adder are connected to the information inputs of the second delay unit, the second and third groups of outputs of which are connected to the second groups of inputs of the first adder and the second subtractor, the outputs of which are connected to information inputs accumulating adder, the outputs of which are connected to the second groups of inputs of the third subtractor and the second adder, the outputs of which are the outputs of the device.

Images