A Large-Model-Fractal Joint-Assisted Intelligent Segmentation and Recognition Method for Mineral Images

Abstract

This invention relates to the field of computer vision and image processing technology. It provides a large-model-fractal joint-assisted intelligent segmentation and recognition method for mineral images, comprising the following steps: acquiring mineral image data and preprocessing it to obtain a mineral image dataset; constructing an initial joint model based on SAM and a fractal model, training the initial joint model to obtain a joint model; optimizing the joint model using the mineral image dataset to obtain an optimized joint model; segmenting the mineral image to be processed and labeling the boundaries of mineral particles in the image; performing fractal analysis on the labeled mineral particles using the optimized joint model, and outputting the segmentation mask and quantitative fractal information of the mineral particles. This method solves the problem that traditional mineral image analysis methods rely on manual observation and labeling, which is not only time-consuming and labor-intensive but also easily affected by subjective factors, leading to inconsistencies and uncertainties in the analysis results.

Description

Technical Field

[0001] This invention relates to the field of computer vision and image processing technology, and more specifically, to a method for intelligent segmentation and recognition of mineral images with large model-fractal joint assistance. Background Technology

[0002] In geology, mineralogy, and related engineering fields, mineral image segmentation and recognition is a crucial task. It not only provides fundamental data support for the exploration, mining, and utilization of mineral resources, but also plays a key role in research across multiple disciplines such as petrology and geochemistry.

[0003] Traditional mineral image analysis methods often rely on manual observation and annotation, which is not only time-consuming and labor-intensive, but also easily affected by subjective factors, leading to inconsistencies and uncertainties in the analysis results. Summary of the Invention

[0004] The purpose of this invention is to provide a large-model-fractal joint-assisted intelligent segmentation and recognition method for mineral images, which aims to solve the problem that traditional mineral image analysis methods rely on manual observation and annotation, which is not only time-consuming and labor-intensive, but also easily affected by subjective factors, leading to inconsistencies and uncertainties in the analysis results.

[0005] The present invention is achieved through the following technical solution:

[0006] A large-model-fractal joint-assisted intelligent segmentation and recognition method for mineral images includes the following steps:

[0007] Mineral image data is acquired, and the mineral image data is preprocessed to obtain a mineral image dataset;

[0008] An initial joint model is constructed based on SAM and fractal model, and the initial joint model is trained to obtain the joint model;

[0009] The joint model was optimized using a mineral image dataset to obtain an optimized joint model.

[0010] The joint model is optimized to segment the mineral image to be processed, and the boundaries of mineral particles in the mineral image are marked.

[0011] The labeled mineral particles are subjected to fractal analysis by optimizing the joint model, and the segmentation mask and fractal quantitative information of the mineral particles are output.

[0012] Optionally, the specific process of acquiring mineral image data and preprocessing the mineral image data to obtain an image dataset is as follows:

[0013] Obtain mineral image data containing several mineral samples; perform denoising processing on the obtained mineral image data; perform image enhancement on the denoised mineral image data to obtain a mineral image dataset.

[0014] Optionally, the mineral image data is obtained by performing microscopic imaging of the mineral sample using a high-resolution microscope or microscopic imaging system under both crossed polarization and single polarization systems.

[0015] Optionally, the specific process of constructing the initial joint model based on SAM and fractal model is as follows:

[0016] Using the SAM model as the basic segmentation model, an initial joint model is constructed by combining it with the analysis model. This initial joint model includes an image encoding module, a cue encoding module, a mask decoding module, and a fractal analysis module. The outputs of the image encoding and cue encoding modules are connected to the input of the mask decoding module, and the output of the mask decoding module is connected to the input of the fractal analysis module. Specifically, the image encoding module extracts feature information from the mineral image; the cue encoding module extracts cue information from the mineral image; the mask decoding module decodes the feature information and cue information of the mineral image into a segmentation mask image; and the fractal analysis module performs fractal analysis on the segmentation mask image.

[0017] Optionally, the specific process of training the initial joint model to obtain the joint model is as follows:

[0018] The mineral image dataset is divided into a training set, a validation set, and a test set;

[0019] The boundaries of mineral particles in the training set are labeled, and the point cue information and bounding box cue information of the mineral image are extracted by the cue encoding module to segment the labeled mineral particles;

[0020] The fractal analysis module calculates the fractal dimension of each segmented mineral particle, quantifying the morphological complexity and spatial distribution characteristics of the mineral particles.

[0021] Optionally, the specific process of calculating the fractal dimension of each segmented mineral particle through the fractal analysis module, and quantifying the morphological complexity and spatial distribution characteristics of the mineral particles, is as follows:

[0022] The fractal dimension is calculated using the box dimension method; the boundary region of the segmented mineral particles is divided into grids of different step lengths, and the step length of the grid is defined as ∈; for each step length ∈, the minimum number of boxes N(∈) required to cover the mineral particles is calculated on the boundary of the mineral particles; the expression for calculating the fractal dimension is as follows (1):

[0023]

[0024] Where D represents the fractal dimension; the fractal dimension D is obtained by fitting different step sizes ∈ and corresponding N(∈) values;

[0025] By calculating the fractal dimension distribution of all mineral particles, statistically analyzing the average fractal dimension and standard deviation of different types of mineral particles, the distribution characteristics of the fractal dimension of mineral particles are extracted.

[0026] Based on the distribution characteristics of the fractal dimension of mineral particles, a threshold range is set to classify mineral particles with matching morphological complexity into the same category.

[0027] Different types of mineral particles are classified and processed by setting threshold ranges.

[0028] Optionally, the specific process of optimizing the joint model using the image dataset to obtain the optimized joint model is as follows:

[0029] When fine-tuning the joint model, the parameters of the mask decoding module are set to fixed values; the parameters of the image encoding module and the prompt encoding module are updated by using the mask image of real mineral particles and the corresponding center point prompt information, and the optimized joint model is obtained.

[0030] Optionally, the optimization process further includes iterative optimization; obtaining the embedding vector of the mineral image dataset through the image encoding module; generating prompt information of the mineral image dataset through the prompt encoding module; inputting the embedding vector and prompt information into the mask decoding module to generate a predicted mask image; setting the total loss function and updating the parameters of the joint model through gradient descent until the loss function value is lower than a preset parameter threshold, at which point the iteration stops.

[0031] Optionally, the specific process of segmenting the mineral image to be processed by optimizing the joint model and marking the boundaries of mineral particles in the mineral image to be processed is as follows:

[0032] The image of the mineral to be processed is input into the image encoding module of the optimized joint model to extract the feature information of the image; the prompt encoding module extracts the prompt information of the image; the extracted feature information and prompt information are input into the mask decoding module to generate a segmentation mask image of the image; based on the segmentation mask image, the boundary of each mineral particle in the image is marked.

[0033] Optionally, the specific process of performing fractal analysis on the labeled mineral particles through optimizing the joint model and outputting the segmentation mask and quantitative fractal information of the mineral particles is as follows:

[0034] The labeled segmentation mask image of the mineral particles is input into the fractal analysis module of the optimized joint model; the fractal analysis module calculates the fractal dimension of each mineral particle; based on the calculated fractal dimension, the morphological complexity and spatial distribution characteristics of each mineral particle are quantified; and quantitative information containing the mineral particle segmentation mask and fractal dimension is output.

[0035] The technical solution of the present invention has at least the following advantages and beneficial effects:

[0036] Improve mineral image segmentation accuracy: By combining the powerful feature extraction capabilities of large models with the fine structure analysis capabilities of fractal models, mineral particles in mineral images can be segmented more accurately, especially for mineral particles with complex textures and boundaries, achieving a significant improvement in segmentation accuracy.

[0037] Enhancing mineral identification capabilities: Fractal analysis not only provides morphological characteristics of mineral particles but also reveals the complexity of their internal structures, which is crucial for accurate mineral identification. Combined with the classification capabilities of large models, it can more reliably identify different types of minerals, providing strong support for subsequent mineralogy research.

[0038] Optimize processing flow and improve efficiency: By constructing and optimizing the joint model, rapid processing of mineral image data is achieved, reducing manual intervention and improving processing efficiency, making the analysis of large-scale mineral image data possible.

[0039] It provides rich quantitative fractal information: It can output quantitative fractal information of mineral particles, such as fractal dimension, which is of great significance for understanding the formation process, physical properties and chemical composition of minerals, and provides a scientific basis for the development and utilization of mineral resources.

[0040] Highly adaptable and widely applicable: By combining large-scale models and fractal models, it has strong adaptability to different types of mineral images and can be widely used in fields such as geological exploration, mineral resource development, and materials science research, thus promoting the development of related disciplines. Attached Figure Description

[0041] Figure 1 This is a flowchart illustrating the intelligent segmentation and recognition method for mineral images using a large model-fractal joint assistance according to an embodiment of the present invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0043] Example 1

[0044] Reference Figure 1 A large-model-fractal joint-assisted intelligent segmentation and recognition method for mineral images, comprising the following steps:

[0045] Step 1: Obtain mineral image data and preprocess the mineral image data to obtain a mineral image dataset.

[0046] In this embodiment, mineral image data containing several mineral samples is acquired; the acquired mineral image data is denoised; and the denoised mineral image data is enhanced to obtain a mineral image dataset.

[0047] In this embodiment, various mineral samples, such as quartz, mica, and albite, are selected from geological laboratories or mineral specimen libraries to ensure representativeness and coverage of different mineral morphologies and structures. High-resolution microscopes or microscopic imaging systems are used to perform microscopic imaging of the mineral samples under orthogonal polarization and single polarization systems. The acquired images should clearly show the boundaries, textures, and internal structures of the mineral grains. Denoising can be performed using image denoising methods such as Gaussian filtering or median filtering. The acquired mineral image data is input into image processing software (such as MATLAB, Python's OpenCV library, etc.), and the selected denoising algorithm is applied to eliminate noise interference in the image, such as random noise and spots. By comparing the images before and after denoising, the clarity of mineral boundaries is ensured to be improved. Image enhancement can be performed by adjusting parameters such as brightness, contrast, and saturation of the image according to the characteristics of the mineral image. The image enhancement function in the image processing software is used to enhance the denoised image, making the boundaries of mineral grains more prominent and the texture clearer. Visual inspection or calculation of image quality indicators (such as signal-to-noise ratio, contrast, etc.) ensures that the image enhancement effect meets the requirements of subsequent segmentation and recognition. The preprocessed mineral image data are categorized and organized according to mineral type, optical system (cross-polarized light, single-polarized light), etc., to form a structured dataset.

[0048] Step 2: Construct an initial joint model based on SAM and the fractal model, and train the initial joint model to obtain the final joint model. SAM stands for Segment Aspecting Model, which is an image segmentation model based on deep learning.

[0049] In this embodiment, the specific process of constructing the initial joint model based on SAM and fractal model is as follows:

[0050] Using the SAM model as the basic segmentation model, an initial joint model is constructed by combining it with the analysis model. This initial joint model includes an image encoding module, a cue encoding module, a mask decoding module, and a fractal analysis module. The outputs of the image encoding and cue encoding modules are connected to the input of the mask decoding module, and the output of the mask decoding module is connected to the input of the fractal analysis module. Specifically, the image encoding module extracts feature information from the mineral image; the cue encoding module extracts cue information from the mineral image; the mask decoding module decodes the feature information and cue information of the mineral image into a segmentation mask image; and the fractal analysis module performs fractal analysis on the segmentation mask image.

[0051] In this embodiment, the specific process of training the initial joint model to obtain the joint model is as follows:

[0052] The mineral image dataset is divided into a training set, a validation set, and a test set; the training set is used for model training, the validation set is used for model tuning, and the test set is used to evaluate model performance.

[0053] The boundaries of mineral particles in the training set are labeled, and these labels will serve as the ground truth labels during model training. Point and bounding box cues are extracted from the mineral images using a cue encoding module to segment the labeled mineral particles. These cues help the model segment mineral particles more accurately. The training set is input into the joint model, and the loss is calculated through forward propagation, while the model parameters are updated through backpropagation. During training, a validation set is used to monitor model performance, and hyperparameters are adjusted as needed. Common segmentation loss functions such as cross-entropy loss or Di ce loss can be used to measure the difference between the segmented mask image and the ground truth labels.

[0054] In each iteration, fractal analysis is performed on the segmented mask image output by the model. The fractal analysis module calculates the fractal dimension of each segmented mineral particle, quantifying the morphological complexity and spatial distribution characteristics of the mineral particles. To enable the model to generate segmentation results with reasonable fractal dimensions, an additional fractal loss function can be designed. For example, the mean squared error between the predicted fractal dimension and the true fractal dimension can be calculated as the fractal loss.

[0055] Step 3: Optimize the joint model using the mineral image dataset to obtain the optimized joint model.

[0056] In this embodiment, the joint model is optimized using an image dataset. The specific process for obtaining the optimized joint model is as follows:

[0057] When fine-tuning the joint model, the parameters of the mask decoding module are set to fixed values ​​to reduce computational resource consumption and accelerate the training process. The parameters of the image encoding module and the cue encoding module are updated using backpropagation with masked images of real mineral particles and corresponding center point cues, resulting in an optimized joint model. The real masked images serve as supervisory information, while the center point cues provide additional guidance, both helping the model to segment mineral particles more accurately. To prevent overfitting, data augmentation techniques (such as random rotation, flipping, and scaling) are used to enhance the mineral images in the training set. This helps generate diverse training samples and improves the model's generalization ability.

[0058] In this embodiment, the optimization process further includes iterative optimization; obtaining the embedding vectors of the mineral image dataset through the image encoding module; generating cue information for the mineral image dataset through the cue encoding module; inputting the embedding vectors and cue information into the mask decoding module to generate a predicted mask image; setting a total loss function, defining a total loss function that includes segmentation loss and fractal analysis loss. The segmentation loss is used to evaluate the difference between the segmentation mask generated by the model and the true mask, and the fractal analysis loss is used to evaluate the difference between the fractal dimension calculated by the model and the true fractal dimension; updating the parameters of the joint model using gradient descent (or its variants, such as the Adam optimizer). In each iteration, the gradient of the total loss function is calculated, and the model parameters are adjusted according to the gradient to minimize the total loss; setting the number of iterations or a loss function threshold as a stopping condition, until the loss function value is lower than a preset parameter threshold, which is also the preset number of iterations or loss function threshold, and stopping the iteration. The performance of the optimized joint model is evaluated on a validation set. The model's performance is evaluated by calculating metrics such as segmentation accuracy, recall, and F1 score, as well as the accuracy and stability of fractal dimension calculation. Based on the evaluation results on the validation set, model parameters and data augmentation strategies are adjusted and optimized to improve model performance.

[0059] Step 4: Optimize the joint model to segment the mineral image to be processed and mark the boundaries of mineral particles in the image.

[0060] In this embodiment, the specific process of segmenting the mineral image to be processed by optimizing the joint model and marking the boundaries of mineral particles in the mineral image to be processed is as follows:

[0061] The mineral image to be processed is input into the image encoding module of the optimized joint model to extract its feature information. The mineral image can be a mineral sample image from a microscope, covering various mineral morphologies and structures. The image encoding module receives the input mineral image and begins to extract feature information from the image, including color, texture, and shape. The cue encoding module extracts cue information from the mineral image to guide the subsequent segmentation process. Point cue information is used to determine the approximate location of mineral particles, and bounding box cue information is used to further limit the segmentation region. The extracted feature information and cue information are input into the mask decoding module to generate a segmentation mask image of the mineral image to be processed. The mask image is a binary image where white areas represent mineral particles and black areas represent the background. Based on the segmentation mask image, the boundary of each mineral particle in the mineral image to be processed is marked. The boundary of the mineral particle can be obtained by calculating the contour of the white area in the mask image. The marked mineral particle boundaries are output as a visualization result. This result can be an original mineral image with the segmentation boundaries superimposed, or an image containing only the segmentation boundaries.

[0062] Step 5: Perform fractal analysis on the labeled mineral particles by optimizing the joint model, and output the segmentation mask and quantitative fractal information of the mineral particles.

[0063] In this embodiment, the specific process of performing fractal analysis on the labeled mineral particles by optimizing the joint model and outputting the segmentation mask and quantitative fractal information of the mineral particles is as follows:

[0064] The labeled segmentation mask image of the mineral particles is input into the fractal analysis module of the optimized joint model. The fractal analysis module calculates the fractal dimension of each mineral particle, using methods such as box counting and radius dimension. Based on the calculated fractal dimension, the morphological complexity and spatial distribution characteristics of each mineral particle are quantified. A higher fractal dimension indicates a more complex morphology and more irregular edges, while a lower fractal dimension indicates a simpler morphology and smoother edges. Furthermore, the spatial distribution characteristics, such as the degree of aggregation and dispersion, can be inferred by analyzing the differences in fractal dimensions among different mineral particles. The output contains quantitative information including the segmentation mask and fractal dimension of the mineral particles. This information can be presented in the form of images, tables, or text for subsequent analysis and processing. For example, an image containing the segmentation mask and fractal dimension of each mineral particle can be generated, where different colors or grayscale values ​​represent different fractal dimension values; or a table can be generated listing the segmentation mask coordinates and fractal dimension values ​​of each mineral particle.

[0065] Example 2

[0066] Based on Example 1, in this example, the specific process of calculating the fractal dimension of each segmented mineral particle through the fractal analysis module, and quantifying the morphological complexity and spatial distribution characteristics of the mineral particles, is as follows:

[0067] The fractal dimension is calculated using the box dimension method; the boundary region of the segmented mineral particles is divided into grids of different step lengths, such as 1 pixel, 2 pixels, 4 pixels, etc., and the step length of the grid is defined as ∈; for each step length ∈, the minimum number of boxes N(∈) required to cover the mineral particles is calculated on the boundary of the mineral particles; the expression for calculating the fractal dimension is as follows (1):

[0068]

[0069] Where D represents the fractal dimension; the fractal dimension D is obtained by fitting different step sizes ∈ and corresponding N(∈) values;

[0070] By calculating the fractal dimension distribution of all mineral particles, and statistically analyzing the average fractal dimension and standard deviation of different types of mineral particles, the distribution characteristics of the fractal dimension of mineral particles are extracted. Histograms of the fractal dimension distribution or kernel density estimation plots for each type of mineral particle can be plotted to visually display its distribution characteristics, analyze the degree of overlap in the fractal dimensions of different types of mineral particles, and determine whether there are significant distribution differences.

[0071] Setting a threshold range based on the distribution characteristics of the fractal dimension of mineral particles can be used to distinguish mineral particles with different morphological complexities. Classifying mineral particles with matching morphological complexities into the same category helps to further understand the morphological complexity and spatial distribution characteristics of mineral particles, and provides an important basis for subsequent mineral identification and analysis.

[0072] Different types of mineral particles are classified and processed by setting threshold ranges.

[0073] The above are merely preferred embodiments of the present invention and are not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A large-model-fractal joint-assisted intelligent segmentation and recognition method for mineral images, characterized in that, Including the following steps: Mineral image data is acquired, and the mineral image data is preprocessed to obtain a mineral image dataset; An initial joint model is constructed based on SAM and a fractal model, and then trained to obtain the final joint model. Specifically, the SAM model is used as the base segmentation model, combined with an analysis model to construct the initial joint model. This initial joint model includes an image encoding module, a cue encoding module, a mask decoding module, and a fractal analysis module. The outputs of the image encoding and cue encoding modules are connected to the input of the mask decoding module, and the output of the mask decoding module is connected to the input of the fractal analysis module. Specifically, the image encoding module extracts feature information from the mineral image; the cue encoding module extracts cue information from the mineral image; the mask decoding module decodes the feature information and cue information of the mineral image into a segmentation mask image; and the fractal analysis module performs fractal analysis on the segmentation mask image. The joint model was optimized using a mineral image dataset to obtain an optimized joint model. The joint model is optimized to segment the mineral image to be processed, and the boundaries of mineral particles in the mineral image are marked. The labeled mineral particles are subjected to fractal analysis by optimizing the joint model, and the segmentation mask and fractal quantitative information of the mineral particles are output.

2. The intelligent segmentation and recognition method for mineral images with large model-fractal joint assistance as described in claim 1, characterized in that, The specific process of acquiring mineral image data and preprocessing the mineral image data to obtain a mineral image dataset is as follows: Obtain mineral image data containing several mineral samples; perform denoising processing on the obtained mineral image data; perform image enhancement on the denoised mineral image data to obtain a mineral image dataset.

3. The intelligent segmentation and recognition method for mineral images with large model-fractal joint assistance as described in claim 2, characterized in that, The mineral image data is obtained by microscopic imaging of mineral samples using a high-resolution microscope or microscopic imaging system under both crossed polarization and single polarization systems.

4. The intelligent segmentation and recognition method for mineral images with large model-fractal joint assistance as described in claim 1, characterized in that, The specific process of training the initial joint model to obtain the joint model is as follows: The mineral image dataset is divided into a training set, a validation set, and a test set; The boundaries of mineral particles in the training set are labeled, and the point cue information and bounding box cue information of the mineral image are extracted by the cue encoding module to segment the labeled mineral particles; The fractal analysis module calculates the fractal dimension of each segmented mineral particle, quantifying the morphological complexity and spatial distribution characteristics of the mineral particles.

5. The intelligent segmentation and recognition method for mineral images with large model-fractal joint assistance as described in claim 4, characterized in that, The specific process of calculating the fractal dimension of each segmented mineral particle through the fractal analysis module, and quantifying the morphological complexity and spatial distribution characteristics of the mineral particles, is as follows: The fractal dimension is calculated using the box-count method; the boundary regions of the segmented mineral grains are divided into grids of varying step sizes, with the step size defined as... For each step size Calculate the minimum number of boxes required to cover the mineral particles at the boundaries of the mineral particles. The expression for calculating the fractal dimension is as follows (1): in, Representing the fractal dimension; through different step sizes and corresponding The value is fitted to obtain the fractal dimension. ; By calculating the fractal dimension distribution of all mineral particles, statistically analyzing the average fractal dimension and standard deviation of different types of mineral particles, the distribution characteristics of the fractal dimension of mineral particles are extracted. Based on the distribution characteristics of the fractal dimension of mineral particles, a threshold range is set to classify mineral particles with matching morphological complexity into the same category. Different types of mineral particles are classified and processed by setting threshold ranges.

6. The intelligent segmentation and recognition method for mineral images with large model-fractal joint assistance as described in claim 1, characterized in that, The specific process of optimizing the joint model using mineral image datasets to obtain the optimized joint model is as follows: When fine-tuning the joint model, the parameters of the mask decoding module are set to fixed values; the parameters of the image encoding module and the prompt encoding module are updated by using the mask image of real mineral particles and the corresponding center point prompt information, and the optimized joint model is obtained.

7. The intelligent segmentation and recognition method for mineral images with large model-fractal joint assistance as described in claim 6, characterized in that, The optimization process also includes iterative optimization; obtaining the embedding vector of the mineral image dataset through the image encoding module; generating prompt information of the mineral image dataset through the prompt encoding module; inputting the embedding vector and prompt information into the mask decoding module to generate a predicted mask image; setting the total loss function and updating the parameters of the joint model through gradient descent until the loss function value is lower than the preset parameter threshold, at which point the iteration stops.

8. The intelligent segmentation and recognition method for mineral images with large model-fractal joint assistance as described in claim 4, characterized in that, The specific process of segmenting the mineral image to be processed by optimizing the joint model and marking the boundaries of mineral particles in the mineral image to be processed is as follows: The image of the mineral to be processed is input into the image encoding module of the optimized joint model to extract the feature information of the image; the prompt encoding module extracts the prompt information of the image; the extracted feature information and prompt information are input into the mask decoding module to generate a segmentation mask image of the image; based on the segmentation mask image, the boundary of each mineral particle in the image is marked.

9. The intelligent segmentation and recognition method for mineral images with large model-fractal joint assistance as described in claim 8, characterized in that, The specific process of performing fractal analysis on labeled mineral particles by optimizing the joint model and outputting the segmentation mask and quantitative fractal information of the mineral particles is as follows: The segmented mask image of the labeled mineral particles is input into the fractal analysis module of the optimized joint model; the fractal analysis module calculates the fractal dimension of each mineral particle; based on the calculated fractal dimension, the morphological complexity and spatial distribution characteristics of each mineral particle are quantified. The output contains quantitative information on mineral grain segmentation mask and fractal dimension.

Images