JP2008234315A - Gradation compression method and gradation compression device for digital image - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a gradation compression method and gradation compression device for digital image, to reproduce the texture and the curvature radius of shape of an object. <P>SOLUTION: The gradation compression method for a digital image comprises an image area division process 30 for dividing the area of the digital image to a high-luminance area and a low-luminance area according to the luminance of a pixel; and a retina response compression process 40 for gradation-compressing the high-luminance area and the luminance of the pixel by response characteristic of a retina cell which logarithmically senses stimulation of light. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a digital image gradation compression method and a gradation compression apparatus.

  A digital image is reproduced as an image by a display device such as a CRT display monitor, a liquid crystal display monitor, a plasma display monitor, or a print device. However, a display device such as a CRT display monitor and a print device each have a limit in the brightness that can be expressed. For example, each channel of a general-purpose CRT display monitor and a liquid crystal display monitor expresses a gradation with 8 bits. On the other hand, each channel of a digital image can record data of more than 8 bits such as 16 bits and 24 bits. Therefore, in order to reproduce a digital image with a gradation exceeding 8 bits (hereinafter referred to as HDR (High Dynamic Range) digital image) on a general-purpose CRT display monitor or the like, the HDR digital image is converted into a digital image with 8 bits or less (hereinafter referred to as “digital image”). Compression to LDR (Low Dynamic Range) digital image). In addition, the range of luminance that can be sensed by the human eye extends to a high-luminance region that far exceeds the luminance that can be reproduced by general-purpose display devices and printing devices. Therefore, when the video of the LDR digital image reproduced on a general-purpose CRT display is viewed, the user may feel uncomfortable.

The conventional HDR digital image gradation compression method is roughly divided into two types: a global operator that processes all the pixels of the HDR digital image using the same compression function, and a compression function for each pixel of the HDR digital image. There are local operators that change. The global operator creates a luminance histogram by dividing the luminance of the pixels of the HDR digital image into a predetermined luminance value range, and compresses the luminance of the pixels of the HDR digital image based on the luminance histogram (for example, see Non-Patent Document 1). ). On the other hand, the local operator obtains an effect similar to dodging and burning used in the analog photographic technique by using the luminance component of the HDR digital image and its low frequency component (see, for example, Non-Patent Document 2).
“A Visibility Matching Tone Reproduction Operator for High Dynamic Range Sciences” IEEE Transactions on Visualization and Computer graphics, 1997. 3, no. 4, P291-306 “Photographic Tone Production for Digital Images” ACM Transactions on Graphics, JULY 2002, Vol. 21, no. 3

  However, when an HDR digital image is tone-compressed by a global operator, the size of the highlight portion of the digital image (the portion where the angle formed between the regular reflection vector of incident light and the visual vector is around 0 ° to 10 °) varies. . The size of the highlight portion affects the texture of the object and the shape curvature. For example, when the size of the highlight portion is relatively small, the object is expressed as if it is rubber instead of metal. In addition, when the size of the highlight portion is relatively large, the pixels are blown out and the curvature of the shape of the object cannot be expressed. Further, when the HDR digital image is subjected to gradation compression by a global operator, the luminance gradation of the shade portion of the digital image (the portion formed by the regular reflection vector of incident light and the visual vector is 30 ° or more) varies. The luminance gradation of the shade portion affects the curvature of the shape of the object. For example, when the luminance gradation becomes gentle, the shape curvature gives a gentle impression. Further, when the luminance gradation becomes steep, the curvature of the shape gives a steep impression.

  When the HDR digital image is tone-compressed by a local operator, the balance of the brightness of the highlight portion is lost, and the curvature of the shape of the object is affected. In addition, the luminance gradation of the shade portion as a whole fluctuates and affects the curvature of the shape of the object.

  The present invention has been made in view of the above, and an object of the present invention is to provide a gradation compression method and a gradation compression apparatus for a digital image capable of reproducing the material texture and shape curvature of an object.

In order to achieve the above object, a first invention provides a gradation compression method for a digital image,
An image region dividing step for dividing the region of the digital image into a high luminance region and a low luminance region according to the luminance of the pixel of the digital image;
A retinal response compression step for gradation-compressing the luminance of the pixels in the high-luminance region according to the response characteristic of retinal cells that logarithmically sense light stimulation;
It is characterized by providing.

According to a second aspect of the present invention, in the gradation compression method for a digital image according to the first aspect, the image region dividing step is configured such that the region of the digital image is divided into a high luminance region and a medium luminance region according to the luminance of the pixel of the digital image. And divided into low brightness areas,
In the retinal response compression step, the luminance of the pixels in the high-brightness region and the medium-brightness region is grayscale-compressed according to the response characteristics of retinal cells that logarithmically sense light stimulation,
Pixels in the medium luminance region are obtained by weighted averaging the first luminance before gradation compression by the retinal response compression step and the second luminance after gradation compression by the retinal response compression step. And a medium luminance region weighted average process for correcting the luminance of the medium.

  In this way, the first luminance before gradation compression by the retinal response compression step and the second luminance after gradation compression by the retinal response compression step are weighted and averaged, and the medium luminance region By correcting the luminance of these pixels, the luminance gradation after gradation compression can be continuously connected.

  According to a third aspect of the present invention, in the gradation compression method for a digital image according to the first or second aspect, the luminance of the pixel of the digital image is linearly corrected before the region of the digital image is divided by the image region dividing step. The method further includes a linear correction step.

  According to a fourth aspect of the present invention, in the gradation compression method for a digital image according to any one of the first to third aspects of the invention, the logarithmic value of the first luminance before the gradation compression by the retinal response compression step and the retina The slope of the input / output response curve indicating the input / output relationship with the logarithmic value of the second luminance after gradation compression in the response compression step is the pair of the logarithmic value of the first luminance and the second luminance. The input / output invariant straight line substantially coincides with the slope of the input / output invariant straight line at a point where the numerical value is the same.

  According to a fifth aspect of the present invention, in the gradation compression method for a digital image according to any one of the first to fourth aspects of the invention, the luminance is maintained in the XYZ color space while maintaining the color vector direction in the L * a * b * color space. The tone compression is performed by the color tone, and the color tone is relatively maintained in the L * a * b * color space.

  Thus, while maintaining the direction of the color vector in the L * a * b * color space, the luminance is tone-compressed in the XYZ color space, and the color tone is relatively retained in the L * a * b * color space. Thus, the color impression of the object can be reproduced. Further, the luminance (Y = 100) of the complete diffuse reflection surface can be used as a threshold value.

According to a sixth aspect of the present invention, there is provided a digital image gradation compression apparatus, wherein the digital image region is divided into a high luminance region and a low luminance region in accordance with the luminance of the pixel of the digital image;
Retinal response compression means for gradation-compressing the luminance of the pixels in the high-luminance region based on response characteristics of retinal cells that logarithmically sense light stimulation;
It is characterized by providing.

  ADVANTAGE OF THE INVENTION According to this invention, the gradation compression method and gradation compression of a digital image which can reproduce the material texture and the curvature of a shape of an object by applying the process in which human eyes feel light to the gradation compression of a digital image. The costume is obtained.

  The best mode for carrying out the present invention will be described below with reference to the drawings.

  FIG. 1 is a process diagram showing the steps of an embodiment of an image gradation compression method according to the present invention. These processes are realized by, for example, a microcomputer. As shown in FIG. 1, the image gradation compression method of the present embodiment includes a pre-processing step 10, a linear correction step 20, an image region dividing step 30, a retinal response compression step 40, and a medium luminance region weighted average step. 50 and a tone gradation compression process 60. The image gradation compression method of the present embodiment applies the process in which human eyes feel light to the digital image gradation compression method. Hereinafter, each process will be described in detail. First, the correspondence between the process of detecting light by the human eye and each process of this embodiment will be described.

  The human eye has a small pupil to reduce the amount of light entering the eye in a bright environment, and the pupil is large to increase the amount of light entering the eye in a dark environment. The linear correction step 20 corresponds to a process in which the human eye adjusts the diameter of the pupil according to the environment and adjusts the amount of light entering the eye.

  In addition, when the human eye has a high-brightness area with high brightness and a low-brightness area with low brightness in the visual field, the sensitivity to light brightness is reduced in the retinal cells where the high-brightness area appears. Happens. When partial adaptation occurs, the sensitivity to the luminance of light decreases not only in the portion of the retinal cell in which the high luminance region is reflected but also in the adjacent retinal cell. That is, the retinal cells in the portion where the medium luminance region that is the boundary between the high luminance region and the low luminance region is also affected by the partial adaptation. The image region dividing step 30 divides an image region into a high luminance region where partial adaptation occurs, a middle luminance region affected by partial adaptation, and a low luminance region where partial adaptation does not occur. Further, the retinal response compression step 40 performs gradation compression on the luminance of pixels in a high luminance region where partial adaptation occurs and a middle luminance region affected by partial adaptation, based on response characteristics of retinal cells that logarithmically detect light stimulation. To do. In addition, the medium luminance area weighted average process 50 mitigates the influence of partial adaptation on the medium luminance area. Hereinafter, each step will be described in detail.

In the preprocessing step 10, data necessary for each subsequent step is created. First, color data of the HDR digital image is prepared in both the XYZ color system and the L * a * b * color system. In the XYZ color system, the Y value represents luminance, the (X, Z) vector direction represents hue, and the (X, Z) vector size represents saturation. The XYZ color system is normalized with the luminance Y of a perfect diffusing reflective surface (Perfect Reflecting Diffuser) as 100. In the L * a * b * color system, the value of L * is lightness, the vector direction of (a *, b *) is hue, and the vector size of (a *, b *) is saturation. Express. As a color conversion formula between the XYZ color system and the L * a * b * color system, the formula defined by JISZ8729 is applied. Here, in order to simplify the following description, the color data of the XYZ color system and the color data of the L * a * b * color system created by the preprocessing step 10 are respectively (X ₀ , Y ₀ , Z ₀ ), (L * ₀ , a * ₀ , b * ₀ ).

  As described above, the linear correction step 20 corresponds to the process in which the human eye adjusts the diameter of the pupil according to the environment and adjusts the amount of light entering the eye.

When the human eye viewing the object, the diameter of the pupil changes from _{d 0} to _{d 1,} the area of the pupil varies _(d ^0/2) from 2π to ^{_{(d 1/2) 2 π}} . At this time, since the amount of light entering the human eye is proportional to the area of the pupil, it becomes d ₁ ² / d ₀ ² of the amount of light before the pupil diameter changes. The luminance Y ₁ of each pixel after the pupil diameter changes from d ₀ to d ₁ can be calculated by Equation 1.

線形補正工程２０は、数式１から明らかなように、瞳孔の直径ｄ_１をｄ_０より大きく設定すると、各画素の輝度を高く補正し、瞳孔の直径ｄ_１をｄ_０より小さく設定すると、各画素の輝度を低く補正する。尚、瞳孔の直径ｄ_１は、ユーザが適当な値を入力できる。

Linear correction process 20, as is clear from Equation 1, when the diameter d ₁ of the pupil is greater than d _0, the brightness of each pixel high corrected, if the diameter d ₁ of the pupil is set to be smaller than d _0, each The pixel brightness is corrected to be low. The diameter d ₁ of the pupil, the user can input an appropriate value.

The diameter d ₀ of the pupil, when the amount of previous HDR digital image to be corrected by the linear correction step 20 and L _bk, can be calculated from the Crawford equation showing the relationship between the pupil diameter and the light amount (see Equation 2) .

また、線形補正工程２０により補正される前のＨＤＲデジタル画像の光量Ｌ_ｂｋは、ＨＤＲデジタル画像の各画素の輝度Ｙ_０を算出平均（Ａｒｉｔｈｍｅｔｉｃ Ａｖｅｒａｇｅ）する式（数式３参照）、又はＨＤＲデジタル画像の各画素の輝度Ｙ_０を幾何平均（Ｇｅｏｍｅｔｒｉｃ Ａｖｅｒａｇｅ）する式（数式４参照）から算出できる。尚、数式３及び数式４において、ｎはＨＤＲデジタル画像の画素の総数、δは対数値が発散することを防ぐための微少な値である。

Further, the light amount L _bk of the HDR digital image before being corrected by the linear correction step 20 is an equation (see Equation 3) for calculating the luminance Y ₀ of each pixel of the HDR digital image (see Equation 3), or the HDR digital image. The luminance Y ₀ of each pixel can be calculated from an equation (see Equation 4) that geometrically averages (Geometric Average). In Equations 3 and 4, n is the total number of pixels of the HDR digital image, and δ is a minute value for preventing the logarithmic value from diverging.

数式４は、各画素の輝度Ｙ_０の対数を平均化するため、各画素の輝度Ｙ_０を直接平均化する数式３と比較して、輝度が異常に高い画素が存在する場合であっても、その影響を低減できる。

Equation 4, in order to average the logarithmic luminance Y ₀ of each pixel, as compared with Equation 3 to directly averaging the luminance Y ₀ of each pixel, even when the luminance is present abnormally high pixel , The effect can be reduced.

As described above, when the user inputs the pupil diameter d ₁ , the linear correction step 20 of the present embodiment corrects the luminance Y ₀ of each pixel of the HDR digital image to Y ₁ according to Equation 1. The user, instead of entering the diameter d ₁ of the pupil, that does not change the diameter of the pupil may choose to words not corrected luminance Y ₀ of each pixel. In the following description, for simplification of the description, the user even when you choose not to correct the luminance Y ₀ of each pixel, representing a and Y ₁ rather than Y ₀ luminance of each pixel.

  As described above, the image region dividing step 30 corresponds to a part of the process in which the retinal cells of the human eye partially adapt. That is, the image region dividing step 30 divides a high luminance region where partial adaptation occurs, a medium luminance region affected by partial adaptation, and a low luminance region where partial adaptation does not occur. Therefore, the image region dividing step 30 divides each pixel into a high luminance region, a medium luminance region, and a low luminance region by the following steps.

In the first step, a value L _H obtained by offsetting the luminance Y _{1 of} each pixel by the threshold value L _ext is calculated according to Equation 5. Here, L _ext is a parameter indicating the ratio of the low-luminance area to the entire image. As L _ext increases, the proportion of the low luminance region increases. Although the user can input an appropriate value for the value of L _ext , it is preferable to empirically set the luminance (100) of the complete diffuse reflection surface.

第２のステップでは、数式６に従って、Ｌ_ＨからＬ_Ａを算出する。尚、数式６において、σは、ユーザが適当な値を入力でき、高輝度領域及び中輝度領域の広がりを表現する。σの値が大きくなるほど、画像全体に対する低輝度領域の占める割合が減る。

In a second step, according to Equation 6, and calculates the _{L A} from _{L H.} In Equation 6, σ can input an appropriate value by the user, and expresses the spread of the high luminance region and the medium luminance region. As the value of σ increases, the ratio of the low-luminance area to the entire image decreases.

第３のステップでは、第２のステップで算出したＬ_Ａに応じて、高輝度領域、中輝度領域、低輝度領域を決める。具体的には、Ｌ_Ａの値が０の領域を低輝度領域、Ｌ_Ａの値が０超ｂ以下の領域を中輝度領域、Ｌ_Ａの値がｂ超の領域を高輝度領域とする。中輝度領域と高輝度領域とを分割する閾値ｂは、ユーザが適当な値を入力できるが、経験的にＬ_ｅｘｔの１／１０倍の値とすることが好ましい。つまり、Ｌ_ｅｘｔの値として完全拡散反射面の輝度（１００）を使用した場合、ｂの値は１０が好ましい。

In a third step, in accordance with the L _A calculated in the second step, the high luminance region, the middle luminance region, it determines the low-luminance region. Specifically, L _A low luminance region a region value is 0, the value of L _A is 0 super b following areas the middle luminance region, L _A value b than in the region of the high luminance region of. Although the user can input an appropriate value as the threshold value b for dividing the medium luminance region and the high luminance region, it is preferable that the threshold value b is empirically set to 1/10 times L _ext . That is, when the brightness (100) of the perfect diffuse reflection surface is used as the value of L _ext, the value of b is preferably 10.

The image region dividing step 30 can set a fourth region other than the high luminance region, medium luminance region, and low luminance region. For example, when the HDR digital image is a design design of an industrial product, the region other than the industrial product, that is, the background region is set to the fourth by setting the α channel in the HDR digital image separately from each channel expressing the color. Can be set in the area. Specifically, by recording auxiliary data in which the industrial product area is α = 1 and the background area is α = 0 in the HDR digital image, the area where α = 0 is set as the fourth area. The fourth region is a region that is not affected by the partial adaptation, the value of L _A and 0. The reason why the fourth area is not affected by the partial adaptation is that, in the design of industrial products, the user pays attention to the material texture and shape curvature of the industrial product, and the background material texture and shape curvature. Is not paying attention. In this way, by setting the α channel, it is possible to reduce the area to be a target of the retinal response compression process 40 described later and improve the calculation speed.

FIG. 2 is a diagram illustrating an example of a process of dividing an HDR digital image area by the image area dividing step 30. FIG. 2A is a design design of the industrial product of this example. 2 (b) is a diagram showing a distribution of _{L H.} 2 (c) is a diagram showing the distribution of _{L A.} FIG. 2D is a diagram illustrating a high luminance region, a medium luminance region, a low luminance region, and a fourth region.

  As described above, the image dividing region process 30 of the present embodiment divides each pixel into at least a high luminance region, a medium luminance region, and a low luminance region by the first to third steps.

As described above, the retinal response compression process 40 corresponds to a part of the process in which the retinal cells of the human eye partially adapt. That, retinal response compression step 40, the response characteristic of retinal cells for sensing the stimulation of light logarithmically, floors luminance Y ₁ of the pixel of the luminance regions in affected high-luminance region and partial adaptation of partial adaptation occurs Tone compression.

The response characteristic of the retinal cell of the human eye is that the response voltage of the retinal cell is R, the maximum and minimum values of the response voltage R of the retinal cell are R _max and R _min, and the stimulation intensity received by the retinal cell is I. Then, it can be approximately expressed by Equation 7. In Equation 7, σ is a semi-saturation constant that is a geometric mean of R _min and R _max, and n is an increase index of the response voltage R of the retinal cell to the stimulation intensity I. The value of the increase index n is generally 0.7 to 2.0, although there are individual differences and environmental differences.

網膜応答圧縮工程４０は、数式７を利用して、高輝度領域及び中輝度領域の画素の輝度Ｙ_１を圧縮する。すなわち、数式７における網膜細胞が受ける刺激強度Ｉを、網膜応答圧縮工程４０により階調圧縮される前の第１の輝度Ｙ_１に置き換え、数式７における網膜細胞の応答電圧Ｒを、網膜応答圧縮工程４０により階調圧縮された後の第２の輝度Ｙ_２に置き換える。このように数式７を利用することにより、高輝度領域及び中輝度領域の各画素の輝度Ｙ_１をＹ_２へ圧縮する数式８を得る。尚、数式８において、Ｙ_ｍａｘ、Ｙ_ｍｉｎは、それぞれ、網膜細胞の応答電圧Ｒの最大値Ｒ_ｍａｘ、最小値Ｒ_ｍｉｎに対応し、ユーザが適当な値を入力できる。また、増加指数ｎの値は、０．７〜２．０の範囲内でユーザが適当な値を選択できる。

Retinal response compression step 40, using Equation 7, compresses the luminance Y ₁ of the pixel in the high luminance region and the medium luminance area. That is, the stimulus intensity I received by the retinal cell in Equation 7 is replaced with the first luminance Y ₁ before gradation compression by the retinal response compression step 40, and the response voltage R of the retinal cell in Equation 7 is replaced with the retinal response compression. replaced by step 40 to the second luminance Y ₂ after being grayscale compression. By using Equation 7 in this way, Equation 8 is obtained that compresses the luminance Y ₁ of each pixel in the high luminance region and the medium luminance region into Y ₂ . In Equation 8, Y _max and Y _min correspond to the maximum value R _max and minimum value R _min of the response voltage R of the retinal cell, respectively, and the user can input appropriate values. Further, the value of the increase index n can be selected by the user within a range of 0.7 to 2.0.

図３は、網膜応答圧縮工程４０により階調圧縮される前の第１の輝度Ｙ_１の対数値と、網膜応答圧縮工程４０により階調圧縮された後の第２の輝度Ｙ_２の対数値との入出力関係を示す曲線（以下、入出力応答曲線）の一例を示す図である。図３では、Ｙ_ｍａｘ＝２００、Ｙ_ｍｉｎ＝１の入出力応答曲線を、増加指数ｎ＝０．５、ｎ＝０．８４８７、ｎ＝１．０の３つの場合について図示した。また、図３では、入出力応答曲線に加えて、第１の輝度Ｙ_１の対数値と第２の輝度Ｙ_２の対数値とが同一であるとする入出力不変直線を図示した。尚、Ｙ_ｍａｘ＝２００、Ｙ_ｍｉｎ＝１のとき、σの値は、数式７から１４．１４となる。

FIG. 3 shows a logarithmic value of the first luminance Y ₁ before gradation compression by the retinal response compression step 40 and a logarithmic value of the second luminance Y ₂ after gradation compression by the retinal response compression step 40. It is a figure which shows an example of the curve (henceforth an input / output response curve) which shows the input / output relationship. In FIG. 3, input / output response curves with Y _max = 200 and Y _min = 1 are shown for three cases of increasing exponents n = 0.5, n = 0.8487, and n = 1.0. Further, in FIG. 3, in addition to the input and output response curve illustrating the output invariant straight first and logarithm of luminance Y ₁ and the second logarithm of the luminance Y ₂ is assumed to be identical. When Y _max = 200 and Y _min = 1, the value of σ is expressed by Equation 7 to 14.14.

  In the retinal response compression step 40, the luminance is gradation-compressed based on the response characteristics of retinal cells that logarithmically sense light stimulation. At this time, the input / output response curve is preferably asymptotic to the input / output invariant straight line in order to maintain the material texture and the curvature of the shape of the object of the digital image before and after gradation compression. Therefore, the value of the increase index n is calculated by the steps described below.

In the first step, the slope g of the input / output response curve at the point where the input / output response curve intersects with the input / output invariant straight line (in the vicinity of the first luminance Y ₁ = σ) is calculated by the least square method using Equation 9. calculate. The slope g is a function of the increase index n, as is apparent from FIG. In Equation 9, S is the sampling number.

第２のステップでは、入出力応答曲線と入出力不変直線とが交差する点で両方の傾きが略一致するときの増加指数ｎの値を最急降下法により算出する。すなわち傾きｇが略１になるときの、増加指数ｎの値を、最急降下法により算出する。最急降下法による算出では、「ｇ−１」の絶対値が０．００１以下、より好ましくは０．０００１以下になるまで計算を繰り返す。

In the second step, the value of the increase index n is calculated by the steepest descent method when the slopes of the input / output response curve and the input / output invariant straight line substantially coincide with each other. That is, the value of the increase index n when the slope g becomes approximately 1 is calculated by the steepest descent method. In the calculation by the steepest descent method, the calculation is repeated until the absolute value of “g−1” is 0.001 or less, more preferably 0.0001 or less.

Thus, the value of the increase index n when the input / output response curve asymptotically approaches the input / output invariant straight line is calculated by the first to second steps. For example, when Y _max = 200 and Y _min = 1, the value of the increase index n is 0.8487, as is apparent from FIG.

As described above, the retinal response compression process 40 of this embodiment performs logarithmic compression of the luminance Y ₁ of each pixel in the high luminance region and the middle luminance region affected by the partial adaptation to Y ₂ according to Equation 8. To do.

  As described above, the medium luminance region weighted average process 50 corresponds to a part of the process in which the cells of the human eye partially adapt. That is, the medium luminance area weighted average process 50 corresponds to alleviating the influence of partial adaptation on the medium luminance area.

Specifically, for the middle luminance region, the first luminance Y ₁ before gradation compression by the retinal response compression step 40 and the luminance Y ₂ after gradation compression by the retinal response compression step 40 are weighted. Average to mitigate the effect of partial adaptation in the medium luminance region. The third luminance Y ₃ after mitigating the effect of partial adaptation can be calculated by Equation 10.

中輝度領域加重平均工程５０は、数式１０から明らかなように、同じ中輝度領域内の画素であっても、輝度が高い画素は、部分順応の影響を比較的大きく受け、輝度が低い画素は、部分順応の影響を比較的小さく受けるように修正する。このように、中輝度領域の画素の輝度を修正することにより、階調圧縮後の輝度の階調を連続的につなぐことができる。

As is clear from Equation 10, the medium luminance region weighted averaging step 50 is relatively affected by partial adaptation even if the pixels are in the same medium luminance region, and the pixels with low luminance are The correction is made so that the effect of partial adaptation is relatively small. Thus, by correcting the luminance of the pixels in the middle luminance region, the luminance gradation after gradation compression can be continuously connected.

  As described above, the medium luminance weighted average process 50 according to the present embodiment mitigates the influence of partial adaptation on the medium luminance region in accordance with Equation 10.

The tone gradation compression process 60 uses the luminance Y ₂ of each pixel in the high luminance area, the luminance Y ₃ of each pixel in the medium luminance area, and the luminance Y ₁ of each pixel in the low luminance area, and performs HDR in each step described later. Tone compresses the tone of digital images.

In the first step, the luminance Y ₂ of each pixel in the high luminance region, the luminance Y ₃ of each pixel in the middle luminance region, and the luminance Y ₁ of each pixel in the low luminance region are calculated from the XYZ color system by L * a * b. * Convert to color system. As a color conversion formula between the XYZ color system and the L * a * b * color system, the formula defined by JISZ8729 is applied. Here, in order to simplify the subsequent description, the luminance of the L * a * b * color system created in the first step is expressed as L * ₁ .

In the second step, the color tone after gradation compression is obtained from the color tone (a ₀ *, b ₀ *) of each pixel prepared in the preprocessing step 10. Here, using the fact that the L * a * b * color space is a spherical uniform color space, while maintaining the direction of the color vector of (L ₀ *, a ₀ *, b ₀ *), the saturation is increased. change. That is, with the gradation compression from L * ₀ to L * ₁ , the vector direction of (a ₀ *, b ₀ *), that is, the hue, while maintaining the hue, (a ₀ *, b ₀ *) ), That is, the saturation is changed to L * ₁ / L * ₀ times. Thus, while maintaining the direction of the color vector in the L * a * b * color space, the luminance is tone-compressed in the XYZ color space, and the color tone is relatively retained in the L * a * b * color space. Thus, the color impression of the object can be reproduced. Further, the luminance (Y = 100) of the complete diffuse reflection surface can be used as a threshold value.

  In the third step, the L * a * b * color system data created in the second step is converted into XYZ color system data. As a color conversion formula between the XYZ color system and the L * a * b * color system, the formula defined by JISZ8729 is applied. By converting to the XYZ color system, it can be used as LDR digital image data on a general-purpose CRT display or the like.

  FIG. 4 is a diagram comparing the LDR digital image created by the present embodiment and the LDR digital image created by the comparative example. Here, the gradation compression method described in Non-Patent Document 2 was used as the gradation compression method of the HDR digital image of the comparative example. FIG. 4A is a design design of an industrial product used for gradation compression in the present example and the comparative example. FIG. 4B is a diagram showing the luminance distribution of the pixels along the line A-A ′ in FIG. 4A of the LDR digital image created according to the present embodiment. FIG. 4C is a diagram showing the luminance distribution of the pixels on the line A-A ′ in FIG. 4A of the LDR digital image created by the comparative example.

  As apparent from FIG. 4, the luminance distribution of the LDR digital image created by the comparative example does not maintain the luminance balance in the highlight portion, and the luminance gradation generally increases in the shade portion. Yes. On the other hand, the luminance distribution of the LDR digital image created according to the present embodiment maintains the luminance balance in the highlight portion, and the luminance gradations match in the shade portion.

  As described above, if the digital image is tone-compressed by the tone compression method of the present embodiment, the texture of the object and the curvature of the shape can be reproduced as compared with the comparative example.

  The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the above-described embodiments, and various modifications and substitutions can be made to the above-described embodiments without departing from the scope of the present invention. Can be added.

  For example, the gradation compression of the digital image of the present embodiment is the gradation compression of the HDR digital image into the LDR digital image. However, as long as it is used to compress the gradation of the digital image, for example, the 24-bit HDR digital image is converted. It can also be used for gradation compression into a 16-bit HDR digital image.

Further, the linear correction step 20 of the present embodiment is set so that the user can input the pupil diameter value d ₁ , but the user can input the value of the pupil diameter square ratio d ₁ ² / d ₀ ^2. It is good also as a setting. By inputting the ratio d ₁ ² / d ₀ ² of the square of the pupil diameter, the step of calculating d ₀ can be omitted, and the calculation speed can be improved.

It is process drawing which shows the process of one Example of the image gradation compression method which concerns on this invention. FIG. 10 is a diagram illustrating an example of a process of dividing an HDR digital image area by an image area dividing step 30; Input / output of the logarithmic value of the first luminance Y ₁ before gradation compression by the retinal response compression step 40 and the logarithmic value of the second luminance Y ₂ after gradation compression by the retinal response compression step 40 It is a figure which shows an example of the curve which shows a relationship. It is the figure which compared the LDR digital image produced by the present Example, and the LDR digital image produced by the comparative example.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Pre-processing process 20 Linear correction process 30 Image area segmentation process 40 Retina response compression process 50 Intermediate area weighted average process 60 Color tone compression process

Claims (6)

In the gradation compression method for digital images,
An image region dividing step for dividing the region of the digital image into a high luminance region and a low luminance region according to the luminance of the pixel of the digital image;
A retinal response compression step for gradation-compressing the luminance of the pixels in the high-luminance region according to the response characteristic of retinal cells that logarithmically sense light stimulation;
A gradation compression method for a digital image, comprising: The image region dividing step divides the digital image region into a high luminance region, a medium luminance region, and a low luminance region according to the luminance of the pixel of the digital image,
In the retinal response compression step, the luminance of the pixels in the high-brightness region and the medium-brightness region is grayscale-compressed according to the response characteristics of retinal cells that logarithmically sense light stimulation,
Pixels in the medium luminance region are obtained by weighted averaging the first luminance before gradation compression by the retinal response compression step and the second luminance after gradation compression by the retinal response compression step. 2. The gradation compression method for a digital image according to claim 1, further comprising a medium luminance region weighted average process for correcting the luminance of the digital image.   The gradation of the digital image according to claim 1, further comprising a linear correction step of linearly correcting the luminance of the pixel of the digital image before dividing the region of the digital image by the image region dividing step. Compression method.   Input / output indicating the input / output relationship between the logarithmic value of the first luminance before gradation compression by the retinal response compression step and the logarithmic value of the second luminance after gradation compression by the retinal response compression step The slope of the response curve substantially coincides with the slope of the input / output invariant straight line at the point where it intersects with the input / output invariant straight line where the logarithmic value of the first luminance and the logarithmic value of the second luminance are the same. The gradation compression method for a digital image according to any one of claims 1 to 3.   While maintaining the direction of the color vector in the L * a * b * color space, the luminance is gradation-compressed in the XYZ color space, and the color tone is relatively retained in the L * a * b * color space. The gradation compression method of the digital image as described in any one of Claims 1-4. An image region dividing means for dividing the region of the digital image into a high luminance region and a low luminance region according to the luminance of the pixel of the digital image;
Retinal response compression means for gradation-compressing the luminance of the pixels in the high-luminance region based on response characteristics of retinal cells that logarithmically sense light stimulation;
A gradation compression apparatus for a digital image, comprising:

Images