TWI660708B - Method for reconstructing fundus image - Google Patents

Abstract

The invention provides a fundus image reconstruction method, including: receiving a fundus image; identifying a plurality of component image data as an input of an iterative operation; and calculating a respective weighting of the component image data based on a first scale. The iterative operation includes: performing an enhancement operation to calculate a plurality of enhanced component image data, and based on a second ratio, calculating respective enhancement weights of the enhanced component image data; determining whether to end the iterative operation, if not , The enhanced component image data of the operation is used as the input of the next iterative operation. After the iterative calculation is completed, the fundus image is reconstructed based on the enhanced component image data.

Description

Fundus image reconstruction method

The invention relates to an image reconstruction method, in particular to a method for reconstructing an eye fundus image.

The retinal hyperplasia membrane or epiretinal membrane (Epiretinal Membrane) is a disease that occurs at the fundus location, which seriously affects the vision of patients. The fundus hyperplasia film can be properly treated. As long as the correct location of the hyperplasia film can be diagnosed and removed completely, the lost vision can be improved.

The findings of fundus hyperplasia can be identified through fundus examination. A fundus lens can be used to capture the back of the eye to obtain a fundus image. Fundus images record the appearance of the patient's retina, which is used to check for abnormalities caused by eye diseases and to track the progress of eye diseases. Clinicians can track the patient's fundus image to understand the historical information of the patient's eye disease, and make appropriate diagnosis and treatment. In particular, the changes in the details displayed in the fundus image may be related to a certain disease. For example, in the case of a color fundus image, the abnormality displayed in the image or the particularly prominent transparent (white) area may be related to the fundus hyperplasia related. That is, the abnormal white area in the image is an optical characteristic caused by a thin film formed on the fundus position or on the retina. Clinicians can identify the hyperplasia membrane by regularly tracking the patient's fundus image records. A fundus image containing a hyperplasia film should include a brighter image area and a darker image area. The image area corresponds to the area of the fundus hyperplasia film. In practice, without special image processing, fundus images taken through general ophthalmoscopes will make it difficult for clinical staff to identify the brighter and darker areas of the image, and it will not be possible to give accurate diagnosis and treatment.

Therefore, in order to give more reliable and efficient fundus diagnosis to patients with such eye diseases, a tool is bound to improve the identification of image details and contrast.

The present invention uses image processing for identification of improving image details and contrast, and is suitable for fundus images, especially fundus images related to fundus hyperplasia films. The invention provides a method for image reconstruction, which converts an unprocessed original image into a reconstructed image. A fundus image (initial image) taken by a fundus mirror is processed by the method provided by the present invention and converted into a reconstructed image, which has the characteristics of contrast and / or pattern detail enhancement.

The fundus image reconstruction method is executed by at least one processor and includes: receiving a color image data related to a fundus image; performing an iterative operation to obtain a plurality of enhanced component image data; and based on the And other enhanced component image data to reconstruct the color image data.

Before performing the iterative operation, the method includes: obtaining a plurality of component image data according to the color image data as an input of the iterative operation. Generally, the fundus image contains color. The method provided by the present invention is to identify a plurality of component image data related to the image based on the color component of the received color image data, and each component image data corresponds to a color, and the color corresponding to the component image is different from Another color corresponding to another component image. For example, one component image may correspond to red, and the other component image may correspond to blue. These component images represent the components of each color of the color image. The present invention utilizes these component images As an input of the iterative operation, and processed separately.

The iterative operation includes: calculating a respective weight of the component image data based on a first ratio of the component image data. Each of the component image data has a different contribution in the color image, that is, each component image data may have a different pixel value on a corresponding pixel. This results in a proportional relationship between the component image data of the color image data. Based on such a proportional relationship, a weighting operation is performed on each of the component image data. Based on such a proportional relationship, a plurality of weight values are generated, each of the weight values corresponding to each of the component images, and the weighting operation is performed based on such a corresponding relationship.

The iterative operation includes performing an enhanced operation to calculate a plurality of enhanced component image data, wherein each enhanced component image data is based on the sum of the other component image data multiplied by a respective weight. Multiply the other component image data by a corresponding weight value. The sum of the weighted other component image data is the enhanced component image data. The color corresponding to the enhanced component image data is different from the colors corresponding to the other component image data, and the other image data also correspond to different colors. In other words, the component image data corresponding to one color is obtained by calculating the weighted sum of the component image data corresponding to other colors.

The enhancement operation further includes: calculating an enhancement weight of each of the enhanced component image data based on a second ratio of the enhanced component image data. As mentioned above, the enhanced component image data also has a proportional relationship with each other, so based on this relationship, a plurality of weight values (herein referred to as enhanced weight values) can be further generated for the enhanced component image data for subsequent processing.

The iterative operation includes: based on the corresponding enhanced weighting and the enhanced component shadow An average value of the image data is used to perform normalization operations on the respective enhanced component image data. The obtained enhanced component image data may cause the correspondingly rendered image intensity to be too high, so a normalization operation related to the image intensity is performed to obtain an appropriate image intensity.

The iterative operation includes: determining whether to end the iterative operation, and if not, the enhanced component image data of the normalized operation is used as the input of the iterative operation. When the iterative operation or the enhanced component images meet a specific condition, the iterative operation is ended. When the condition is not satisfied, the iterative operation is performed again, and the component image data is replaced with the normalized result as an input for subsequent iterative operations. The flow of subsequent iteration operations is basically the same as the flow of the previous iteration operation. End the iterative calculation and reconstruct the color image data based on the enhanced component image data. The image displayed by the reconstructed color image data has a color and detail performance different from the original image.

These and other features and advantages of the invention will be presented in more detail in the following description of the invention and the drawings illustrated by the principles of the invention.

2‧‧‧Image Acquisition Unit

4‧‧‧ imaging unit

6‧‧‧lighting unit

61‧‧‧light source

62‧‧‧optical element

8‧‧‧ camera unit

10‧‧‧Computer device

102‧‧‧ processor

104‧‧‧Storage unit

106‧‧‧Display unit

108‧‧‧ input interface

E‧‧‧ eyes

20‧‧‧ input image

22‧‧‧ reconstructed image

24‧‧‧ component image data

24r, 24g, 24b ‧‧‧ component image data

26‧‧‧ Intermediate reconstruction image

W _r , W _g , W _b ‧‧‧ weight value

300 to 320‧‧‧ steps

The first figure shows a system suitable for the fundus image reconstruction method of the present invention.

The second figure shows an input image and a reconstructed image (according to the method provided by the present invention).

The third figure shows the flow of the method of the present invention.

The invention will be described more fully with reference to the accompanying drawings, and specific examples and specific embodiments are shown by way of illustration. However, the subject matter of this claim can be embodied in many different forms, so The construction of the claimed or covered subject matter is not limited to any example embodiments disclosed in this specification; the example embodiments are merely examples. As such, the invention resides in providing a reasonably broad scope to the claimed subject matter that is claimed or covered. In addition, for example, it is claimed that the subject matter may be embodied as a method, apparatus, or system. Thus, specific embodiments may take the form of, for example, hardware, software, firmware, or any combination of these (known not to be software).

The term "in one embodiment" used in this specification does not necessarily refer to the same specific embodiment, and "in other embodiments" used in this specification does not necessarily refer to a different specific embodiment. Its purpose is, for example, that the claimed subject matter includes all or part of a combination of exemplary embodiments.

The first figure illustrates a system for obtaining a fundus image of an eye E and its processing. The system includes an image capture device, an image capture (recording) device, and a processing and application device. The image capturing device may be included in the image capturing device to capture a clear fundus image. The processing and using device carries program instructions, which are configured to execute the image reconstruction method of the present invention to process a received image. The processing and using device receives the initial image data obtained by the image capturing device, and outputs a reconstructed image through the special weighting process provided by the present invention. The image capture device and the image capture device are not necessarily combined into a single capture device, and the image capture device can also be combined into a processing and application device.

The photographing device may be a fundus camera dedicated to photographing a fundus image. As shown in the first figure, an image capture device of the system may include an image capture unit 2 and an imaging unit 4. The image capturing unit 2 includes a plurality of optical elements, such as an objective lens, a semitransparent mirror, a focusing lens, an aperture, and the like. The image capturing unit 2 is configured to form a Focusing surface such that the bottom surface of an eye E overlaps the focusing surface. The image capturing unit 2 can be matched with a lighting unit 6. The lighting device 6 includes a light source 61, a condensor, and other optical elements 62. A part of the illumination unit 6 is included in the image capturing unit 2, such as a common objective lens. The lighting unit 6 is configured to project light into the eyeball, illuminate the fundus area, and provide sufficient light for the image capture unit 2 to capture. According to the operation, the light source 61 can be selectively turned on. In the present invention, more lighting units can be included to meet different viewing needs.

The imaging unit 4 includes a plurality of optical elements, such as a reflector, a dichroic mirror, a relay lens, and the like. The imaging unit 4 is connected to the rear end of the image capturing unit 2, and is used for receiving the light captured by the former and projecting to a camera unit 8, which is included in the image capturing device. In other possible cases, the imaging unit 4 and the camera unit 8 may be included in the image capturing device to capture the fundus image and output a color image data. A captured fundus image is transmitted to the processing and application device. The processing and application unit may be a computer device 10 including at least a processor 102, a storage unit 104, a display unit 106, and an input interface 108. The captured fundus image is stored in the storage unit 104 in the form of data. The storage unit 104 also stores a plurality of instructions for being read by the at least one processor 102 to perform various operations, including operations for reconstructing fundus images provided by the present invention. The storage unit 104 also stores the results of each iteration operation as a source of subsequent iteration operations or reconstructed images. The storage unit 104 is a memory that stores the color image data and intermediate data related to image processing (such as component image data and enhanced component image data). The intermediate information will be explained later. One or more processed or unprocessed images may be displayed on the display unit 106 for viewing and comparison. The display unit 106 may be a display device, such as an LCD high-resolution screen, which displays the transparent area (i.e., shadow) of the fundus image in the color image data after reconstruction. The white part of the image is the area where the fundus hyperplasia film may form). The input interface 108 is composed of hardware and software, such as an input keyboard, a touch device, and an operating program that interacts with these hardware. The input interface 108 is configured to provide a user or an operator to interact with the computer device 10. Through the input interface 108, the computer device 10 can receive control parameters for executing the method of the present invention, such as a number of preset values or conditions related to the operation, which will be described in the subsequent content. According to the operation of the input interface 108, the processor 104 can cause the color image data reconstructed by different iteration times (that is, the intermediate reconstruction image) to be displayed on the display device.

The image reconstruction method of the present invention is not necessarily executed in the system of the first figure, and the color image data can also be transmitted to other computers or servers for processing through other methods, such as the network. There may be one or more processors executing the method of the present invention, and these processors may be stored in a computer device or distributed among different computing devices.

The second figure shows an input image 20 (initial image) converted into a reconstructed image 22 by the method of the present invention. The input image 20 is displayed based on a color image data. The color image data is obtained by an image capturing device. The color image data may be an output of the image capturing device, and is generally unprocessed. The format of the color image data can be jpg, jpeg, bmp or png. The color image data is presented in a matrix manner, for example, an input image 20 of 512 × 512 pixels, and the color image data is also a 512 × 512 matrix. Each element in the matrix represents a pixel value. . The color image data and the component image data and the enhanced component image data of subsequent calculations all include a plurality of pixel values. The color image data is composed of a plurality of component image data. The component image data is defined based on a color model, which may be selected from one of a RGB color model, a CMYK color model, and a HSI color model.

As shown in the second figure, first, the color image data related to the input image 20 (such as a fundus image) is received. The processor obtains a plurality of component image data 24 based on the color image data. The component image data 24 exemplified here is obtained based on the three primary color models, that is, the red part 24r, the blue part 24b, and the green part 24g of the color image data. Conversion between models to obtain other component image data based on the color mixing model or other component image data based on the HSI model. The component image data 24 is also represented in a matrix manner, such as a 512 × 512 matrix, and the elements of each matrix are pixel values related to a color. In other words, the color image data includes or consists of a superposition of the component image data 24. The initial image obtained by most image capturing devices generally forms corresponding data according to the components of its three primary colors, which is related to the configuration of the photosensitive element. Therefore, it is possible for those skilled in the art to identify the component image data of the three primary colors in the color image data.

The method of the present invention includes a weighting operation. The processor generates a plurality of weight values W _r , W _g , and W _b (initial weights) according to a proportional relationship (first ratio) between the component image data 24r, 24g, and 24b, each of which corresponds to a previously identified Component image data 24r, 24g, 24b. The proportional relationship (the first ratio) will be described later.

The method of the present invention includes an iterative operation to enhance the transparent area of the fundus image in the color image data. The method of the present invention also includes more iterative operations to further strengthen the area. The iterative operation includes an enhanced operation and an enhanced weight operation. Identification of these components has image data 24r, 24g, 24b and the weight calculation of such weight value W _r, W _g, W _b, respectively, may be used as an input to the iterative computation (the first iteration operation). The enhancement operation is performed to calculate a plurality of enhanced component image data (not shown), wherein each enhanced component image data is based on other component image data multiplied by a respective weight (weight values W _r , W _g , W _b ) the sum. The enhanced operation will be explained in detail in the following content. The data dimension of the enhanced component image is the same as the component image data before the enhancement, that is, a 512 × 512 matrix.

The iterative weighting operation included in the iterative operation is based on a proportional relationship (second ratio) of the enhanced component image data, and calculates the respective enhanced weighting of the enhanced component image data, that is, a plurality of enhanced weight values are generated. The enhancement weight here is different from the aforementioned initial weight. The enhancement weight is obtained based on the enhancement operation in the iterative operation; the initial weight is obtained based on the initial component image data.

The enhanced component image data may be an output of the iterative operation. The output is used as an input of the next iteration operation (second iteration operation), and the enhanced operation and the enhanced weight operation are repeated to obtain new plurality of enhanced component image data as the next iteration An input for the operation. These operations are repeated until the processor ends the iterative operation according to a judgment result. After the processor ends the iterative operation, the color image data is reconstructed based on the latest enhanced component image data. The enhanced component image data finally obtained and the corresponding component image data of the color image data are superimposed on each other to generate the reconstructed image 22. Compare the fundus images 20 and 22 before and after reconstruction in the second image. The reconstructed image 22 has a higher brightness contrast, which helps to identify a bright area and a relatively dark area of the image 22. In particular, the pattern details of the reconstructed image 22 are more distinct due to the enhancement.

Before the processor stops the continuous iteration operation, the output of one or more iteration operations, that is, the enhanced component image data, may be stored in the storage unit 104 in the first picture system. At the same time, one or more intermediary reconstructed images 26 (based on the stored It can be stored together. The intermediate reconstructed image 26 is different from the reconstructed image 22 of the second image. The intermediary reconstruction image 26 is produced before the iterative operation has not stopped. The intermediary reconstructed images 26 can be recalled and displayed via a display device. Each intermediate reconstruction image 26 may be generated based on the enhanced component image data of the last operation after every predetermined number of iteration operations. For example, an intermediate reconstruction image is generated every 1000 iteration operations. Therefore, as long as the iterative operation has not stopped, at least one intermediary reconstructed image 26 can be obtained. The characteristics such as contrast and brightness of the reconstructed images 26 exhibit a continuous change, and this change can be used for other image processing, for example, it can be used as whether to end or stop the continuous iterative calculation.

The third diagram is the flow of the method of the present invention, including steps 300 to 320, which are executed by at least one processor. For the convenience of explanation, n in the figure represents the initial processing without iteration operation, n + 1 represents the processing after the first iteration operation, and so on. The method starts at step 300 and receives an initial image. An arithmetic device or a computer receives the color image data, that is, the input image 20 of the second image, and the color image data is related to the fundus image. The initial image is taken from an image capturing device and usually the color or pixel value is not processed. However, in other embodiments, the initial image may also undergo preliminary processing, such as compression of the image or normalization of pixel values.

In step 302, the color components included in the color image data are identified by at least one processor, such as the aforementioned three primary colors. The three primary color components contained in each pixel value of the color image data are identified, and a plurality of component image data are obtained accordingly, as described by the following equation: X _initial = r _n + g _n + b _n ...... ( 1), where X _{initial is} the color image data, and r _n , g _n , b _n are component image data, as shown in the second figure, the component image data 24r, 24g, 24b based on the three primary color model. In other embodiments, other component image data may be identified from the color image data based on a CMYK model or an HSI model. These component image data have the same dimensions, such as 512 × 512. The component image data is also an input for subsequent iterative calculations.

Step 304: Generate an initial weight about the color component. A proportional relationship (first scale) of the component image data is used to calculate the respective weighting of the component image data. The first ratio is a ratio of respective average values of the component image data. According to the first ratio, a plurality of initial weight values are generated. The weighting of the respective component image data, that is, an initial weight value corresponding to each component image data, is the sum of the average value of the respective component image data divided by the average value of all the component image data, as described by the following equation: Among them, k represents one of the component image data, i = { r, g, b } represents all component image data, avg ( k ) represents the pixel average of a component image data, and Σ _i avg ( i ) represents all components The average pixel value of the image data, and W _k is a weight value related to the component image data k , such as W _r , W _g , and W _b calculated in the second figure. These weight values are used as an input for subsequent iterations.

Step 306: Perform an iterative operation. The iterative operation starts from an enhanced operation to generate a plurality of enhanced component image data (n + 1), wherein each enhanced component image data is based on other component image data multiplied by a respective one. Weighted sum. Each of the enhanced component image data is calculated based on at least two of the component image data that has been weighted, and wherein at least two of the component image data (n) used for the enhanced operation correspond respectively. Two different colors (for example, green and blue), and the enhanced component image data (n + 1) calculated based on at least two of the component image data (n) corresponds to another color (for example, red), the other One color is different from the two different colors. For example, the operation of the enhanced component image data (n + 1) based on the three primary colors is described by the following equation: Among them, r _n , g _n , b _{n are} the component image data (matrix), W _rn , W _gn , W _bn are weight values (constant) corresponding to the component image data, r _{n +1} , g _{n +1} and b _{n +1 are} the enhanced component image data (matrix).

The iterative operation also includes an enhanced weighting operation. In step 310, a weighted operation is performed on the enhanced image data obtained by the operation. The weighting operation in step 310 is the same as the weighting operation in step 304, but the basis of the calculation is different. Step 304 is calculated based on the initial component image data (n), and step 308 is calculated based on the enhanced component image data (n + 1). Based on a proportional relationship (second scale) of the enhanced component image data (n + 1), the respective enhancement weights of the enhanced component image data (n + 1) are calculated. Generate a plurality of enhanced weight values W k _{n + 1} , as described by the following equation: Where r _{n +1} , g _{n +1} , and b _{n +1 are} the enhanced component image data (matrix). The iterative operation can be completed after generating the enhanced component image data and corresponding enhanced weight values.

The iterative operation also includes a normalization operation, step 312. Based on a corresponding mean value of the enhancement weighting and the enhancement component image data, normalization operations are performed on the respective enhancement component image data. The normalization operation is performed on the respective enhanced component image data to calculate the original respective enhanced component image data minus the original average value of the respective enhanced component image data and 1 plus the product of the respective enhancement weights, and the calculation results are regarded as normalized respective Enhanced component image data, as described by the following equation: r _{n +1} (normalized) = r _{n +1} - avg ( r _{n +1} ) × (1+ Wr _{n +1} ) ...... (7), g _{n +1} (normalized) = g _{n +1} - avg ( g _{n +1} ) × (1+ Wg _{n +1} ) ...... (8), b _{n + 1} (normalized) = b _{n + 1} -avg (b _{n + 1} ) × (1 + Wb _{n + 1} ) ... (9). The enhanced component image data of the normalized operation can be used as an output of the iterative operation. In some embodiments, step 312 may be omitted, that is, the enhanced component image data is used as the output of the iterative operation.

Step 314: Determine whether to end or stop the iterative operation, that is, whether to continue to strengthen the enhanced component image data and its enhanced weighting operation. If the iterative operation is not ended or stopped, the enhanced component image data (step 312) of the normalized operation is used as the input of the iterative operation. Continue the iterative operation and return to step 306, and the input of the continuous iterative operation (n + 2) includes the enhanced component image data (n + 1) obtained from the foregoing step 308 and the enhanced weight obtained from step 310 Value (n + 1).

In one embodiment, the determining whether to end the iteration operation is to determine whether the number of iterations is reached, which may be preset. When the continuous iteration operation satisfies the number of iterations, for example, 10,000 times, the iteration operation at (n + 9999) ends or stops, and proceeds to step 316. If it is determined that the number of iterations has not been reached, the enhanced component image data of the normalized operation is used as an input to continue the iteration operation, and the iteration operation is performed. In addition, possible practices may include manual operations Make. For example, the operator may decide whether to perform several iterative calculations based on the observations of the intermediate reconstruction images 26 as shown in the second figure.

The iteration operation further includes reconstructing the color image data of the iteration times based on the enhanced component image data obtained based on the iteration times every time a preset iteration number or a multiple thereof is reached, as described in Reconstructed images. For example, when the number of iterations reaches 1,000 times or a multiple of 1,000, the color image data is reconstructed based on the enhanced component image data obtained by the last iteration operation to obtain one or more intermediary reconstructions. Image, and continue to iterate. As mentioned earlier, these intermediary reconstructed images can present continuous changes in the initial image during processing. This can be provided to the operator or clinician for observation.

Step 314, in other embodiments, the determining whether to end the iteration operation is to judge whether the difference between different operation results obtained by different iterations is less than a critical condition. In the continuous iterative operation, when the difference between the enhanced component image data (normalized) of a certain iterative operation and the enhanced component image data (normalized) of a previous iterative operation is less than the critical condition, the process ends Or stop continuing the iterative operation and proceed to step 316. For example, it is determined whether the overall pixel value difference between one intermediate reconstruction image obtained from an iterative operation and another intermediate reconstruction image obtained from another iterative operation is smaller than the critical condition.

The critical condition is that it can be a convergence condition. When the differences between these outputs of different iterative operations (that is, multiple intermediate reconstruction images or multiple enhanced component image data) show a convergence trend, that is, the convergence condition is met, the convergence Algebraic operations will end or stop.

In step 316, based on the judgment in step 314, that is, whether the number of iterations is satisfied or whether the iteration result is converged, the iteration operation is stopped and the termination is strengthened. In step 318, the color image data is reconstructed based on the enhanced component image data obtained by the last iteration operation. The reconstructing the color image data includes superimposing the enhanced component image data and the corresponding component image data of the color image data on each other after the last iteration operation is completed to generate a reconstructed image, as described in the following equation : X _final = r _final + g _final + b _final + X _initial ...... (10), where X _final is the reconstructed image data, and r _final , g _final , and b _final are the last time after the iteration operation stops. The enhanced component image data obtained by the iterative operation are superimposed on the corresponding component images r _n , g _n , b _n , that is, r _final + r _n , g _final + g _n , b _final + b _n . In step 320, the processor outputs the reconstructed image (such as the second reconstructed image 22) based on the reconstructed image data X _final , which can be displayed on the display device. This reconstructed image is different from the aforementioned intermediate reconstructed image. The reconstructed image here is generated after the iterative operation ends or stops. After the fundus image is reconstructed through the above-mentioned steps, the area (white) of the fundus hyperplasia film is more obvious.

In the above embodiment, the three primary colors model is taken as an example for illustration. In other embodiments, the three primary color models may be replaced with a CMYK model or an HSI model. For the CMYK model, the above operation may further include a complementary operation on color, for example, component image data including identifying and converting an input image data, as described by the following equation: X _initial = c + m + y + k . ..... (11), c _n = 255- c ...... (12), m _n = 255- m ...... (13), y _n = 255- y ... ... (14), k _n = 255- k ...... (15), where c , m , y , k are cyan, magenta, yellow, and black matrices, respectively, c _n , m _n, y _n, k _n is the converted as a basis for an iterative calculation of the weighted input and calculated. In addition, for the HSI model, it is based on a linear relationship between the three primary color models and the HSI model. The above operation includes a conversion operation. Those skilled in the art can appropriately adjust or modify the above-mentioned operation flow based on common knowledge, so that the present invention is applicable to other color models.

Although the foregoing invention has been described with certain details for the sake of clarity, we will understand that certain changes and modifications can be implemented within the scope of the patent application. Therefore, the above embodiments are used for illustration only, and are not limited, and the present invention is not limited to the details described herein, but may be modified in the scope of equivalent patent application fields and equivalents.

Claims (14)

A fundus image reconstruction method executed by at least one processor includes: receiving a color image data about a fundus image; and obtaining a plurality of component image data (n) according to the color image data as a stack An input of algebraic operation; based on the input to obtain a first proportion of the component image data related to a weight (W _n ), calculate respective weights of the component image data, and the weights of the respective component image data are respective components The mean value of the image data is divided by the sum of the mean values of all the component image data; performing the iterative operation includes: performing an enhanced operation to calculate a plurality of enhanced component image data (n + 1), where each enhanced component image data is Based on the sum of the other component image data multiplied by their respective weights, and based on a second ratio of the enhanced component image data, calculate the respective enhanced weights (n + 1) of the enhanced component image data; based on the corresponding enhanced weights And an average value of the enhanced component image data, perform normalization operations on the respective enhanced component image data; determine whether End the iterative operation, if not, use the enhanced component image data of the normalized operation as the input of the iterative operation, and after ending the iterative operation, reconstruct the color based on the enhanced component image data video material. For example, the method of claim 1 in the patent scope, wherein the component image data is a pixel value. For example, the method of claim 1 in the patent scope, wherein the second ratio is the ratio of the respective averages of the enhanced component image data. For example, the method of claim 3 in the patent scope, wherein the enhancement weighting of the respective enhanced component image data is the sum of the average of the respective enhanced component image data divided by the average of all the enhanced component image data. For example, the method of claim 1, wherein the plurality of component image data is executed based on a color model selected from one of a BGB model, a CMYK model, and an HSI model. For example, the method of the first scope of the patent application, wherein the normalization operation of the respective enhanced component image data is to calculate the product of the original respective enhanced component image data minus the original average of the respective enhanced component image data and 1 plus the respective enhanced weighting The calculation results are used as the image data of each enhancement component. For example, the method of applying for the first item of the patent scope, wherein the determining whether to end the iteration operation is to judge whether the number of iterations is reached. For example, if the method of the seventh item of the patent scope is applied, wherein it is determined that the number of iterations has not been reached, the enhanced component image data of the normalized operation is used as the input of the iteration operation, and the iteration operation is performed. For example, the method of claim 1 in the patent scope, wherein the iteration operation further includes, each time a preset number of iterations is reached, based on the enhanced component image data obtained by the number of iterations, reconstructing the Color image data. For example, the method of claim 9 in the patent application scope, wherein the determination of whether to end the iteration operation is to judge whether the difference between different operation results obtained by different iterations is less than a critical condition. The method according to item 1 of the patent application range, wherein the reconstruction of the color image data includes superimposing the enhanced component image data and the corresponding component image data of the color image data on each other to generate A reconstruction image. A fundus image capturing device includes: an image capturing device for capturing a fundus image; an image capturing device for capturing the fundus image to output a color image data; and a processor for obtaining a plurality of images based on the color image data The component image data is used as an input of an iterative operation, and based on the input, a first ratio of a weight of the component image data is obtained to calculate the respective weights of the component image data. The weight of the respective component image data is The mean value of the respective component image data is divided by the sum of the mean values of all the component image data, and the processor performs the iterative operation to strengthen the transparent area of the fundus image about the fundus image. The iterative operation includes an enhanced operation Calculate a plurality of enhanced component image data, wherein each enhanced component image data is based on the sum of other component image data multiplied by a respective weighting, and based on a second ratio of the enhanced component image data, calculate the enhanced component images The respective weighting of the data, after the processor finishes the iterative operation, To such strengthening component of the video data, the reconstruction of the color image data. For example, the image capturing device of the patent application No. 12 includes a memory that stores the color image data, the component image data and the enhanced component image data, and a display device that displays the reconstructed image after reconstruction. Transparent areas of the fundus image in the color image data. For example, the image capturing device of the scope of application for patent No. 13, wherein each time the processor reaches a preset number of iterations, the processor reconstructs the number of iterations based on the enhanced component image data obtained by the number of iterations. The reconstructed color image data, and the processor displays the reconstructed color image data on the display device according to an operation of an input interface to different iteration times.