Volume data rendering acceleration method for blank space jump

Abstract

The invention provides a volume data rendering acceleration method for blank space jump, which relates to the field of volume data rendering, and comprises the following steps: judging whether a volume data block to which a current light sampling point belongs has a non-empty voxel, if the volume data block is a blank volume data block, entering a Chebyshev jump stage, jumping the blank volume data block based on a Chebyshev distance map, and if not, entering a Chebyshev jump stage; and if not, entering a voxel-level jump stage. And in the voxel-level jumping stage, judging whether the voxel is non-empty, if the voxel is non-empty, executing a ray casting sampling process, and if the voxel is empty, entering the jumping stage based on the unsigned distance map to jump the blank voxel. And performing adaptive sampling step length adjustment based on the gradient amplitude, the volume data value and the viewpoint distance. And when the light is emitted out of the cube, iteration is stopped. According to the technical scheme, the problems that in the prior art, efficient skipping of empty voxels cannot be achieved, and the calculation efficiency of a ray casting algorithm is low are solved.

Description

Technical Field

[0001] The present invention relates to the field of volume data rendering, and in particular to a volume data rendering acceleration method oriented to blank space jumping. Background Art

[0002] Volume data, a numerical data structure used to describe objects or phenomena in three-dimensional space, is widely used in fields such as medical imaging, scientific computing, and engineering visualization. Volume rendering technology generates two-dimensional images directly from three-dimensional data, visually displaying its internal structure and properties. Current mainstream volume rendering methods include texture slicing, snowballing, and ray casting. Ray casting has become a research hotspot due to its high-quality imaging performance. Ray casting methods determine pixel values ​​and generate realistic images by calculating the color contribution at the intersection of a ray emitted from a viewpoint and a scene object. The core process of the algorithm involves emitting a ray from the viewpoint to intersect the volume data, sampling and color synthesis along the ray propagation path, and generating a volume rendering result on a two-dimensional screen. However, this algorithm suffers from a significant computational redundancy problem: it uses a uniform sampling strategy for both empty and non-empty voxels. This means that during the rendering process, the color and opacity values ​​of all voxels must be blended, while the actual imaging contribution comes only from the valid voxel regions. This means that a significant amount of computational resources is used to sample and interpolate empty voxels.

[0003] Therefore, there is a need for a volume data rendering acceleration method for blank space jumping that can achieve efficient skipping of empty voxels and improve the computational efficiency of the ray casting algorithm. Summary of the Invention

[0004] The main purpose of the present invention is to provide a volume data rendering acceleration method for blank space jumping, so as to solve the problems in the prior art of being unable to achieve efficient skipping of empty voxels and low computational efficiency of ray casting algorithms.

[0005] To achieve the above objectives, the present invention provides a volume data rendering acceleration method for blank space jumping, which specifically includes the following steps: S1: Divide the volume data into blocks of fixed size. As the ray traverses the volume data, determine whether the volume data block to which the current ray sampling point belongs contains a non-empty voxel. If it is a blank volume data block, the process enters the Chebyshev jump phase, in which the blank volume data block is jumped based on the Chebyshev distance map. If it is not a blank volume data block, the process enters the voxel-level jump phase.

[0006] In step S2, during the voxel-level jump phase, the voxel is determined to be non-empty. If so, the ray casting sampling process is executed. If not, the unsigned distance map-based jump phase is entered to jump to the empty voxels. In step S3, during the ray casting sampling process, the sampling step size is adaptively adjusted based on the gradient amplitude, volume data value, and viewpoint distance.

[0007] S4, calculate the distance m traveled from the incident light to the outgoing light. Each time the volume data is sampled, calculate the distance n that the light has traveled in the cube. If n>=m, it means that the light has exited the cube, and the iteration is stopped.

[0008] Furthermore, step S1 specifically includes the following steps: S1.1, assuming that the volume data dimension is , the occupancy map dimension is , the block size is , then for the normalized sampling point coordinates , the coordinates of the volume data in the occupancy map for: .

[0009] S1.2, assuming that the normalized step length between two adjacent sampling points is ,but Step size in the occupancy map for: .

[0010] S1.3, based on the occupancy map, verify whether the current block is an empty block; if it is empty, enter the Chebyshev jump phase, which includes: skipping the remaining part of the current block and skipping blank volume data blocks based on the Chebyshev distance.

[0011] Furthermore, step S1.3 specifically includes the following steps: S1.3.1, in a certain dimension The formula for jumping up the remaining part of the current block is: ; in, Indicates the current sampling point in the occupancy map The non-normalized coordinates in dimensions, Indicates the first The sampling distance in the dimension direction, To round down.

[0012] S1.3.2, based on the Chebyshev distance, the number of blank data blocks skipped is to calculate the Chebyshev distance from the current empty block to the nearest non-empty block through the occupancy map, and generate the corresponding Chebyshev distance map; Chebyshev distance Represents two vectors Each coordinate component in The maximum absolute value of the difference: ; in, To take the maximum value.

[0013] S1.3.3, combined with steps S1.3.2 and S1.3.3, the calculation formula for the jump distance in the Chebyshev jump phase is: ; in, Indicates the distance to jump to the next non-empty data block at the current position. represents a symbolic function; is the Heaviside step function, Indicates the offset of the current point in the original empty block, To get the minimum value, To round up.

[0014] S1.3.4, simplify the jump distance calculation formula in step S1.3.3 to obtain: .

[0015] Furthermore, step S2 specifically includes the following steps: S2.1, traverse all voxels in the volume data and determine whether each voxel is non-empty; set the UDF value of the non-empty voxel to 0, and initialize the distance of other empty voxels to positive infinity.

[0016] S2.2, for empty voxels, the unsigned distance graph jump phase is entered to perform distance propagation calculations along the x, y, and z axes in sequence. Each phase is based on the distance information of the previous phase, and a bidirectional traversal or alternating search strategy is adopted to calculate the distance value of each voxel to the nearest non-empty voxel using the Euclidean distance formula. The three-dimensional unsigned distance field is generated by iteratively taking the minimum value. The three-dimensional unsigned distance field stores the Euclidean distance from the current voxel to the nearest non-empty voxel. The calculation formula is: ; in, is the distance from the current voxel to the nearest non-empty voxel, To represent the distance value of the queried offset voxel in the x-axis direction, is the current voxel coordinate, To take the minimum.

[0017] S2.3, set the minimum jump threshold based on the volume data scale Each time the UDF distance value is queried, the UDF distance value is compared with the minimum jump threshold. When the distance value is greater than the minimum jump threshold, a jump is performed. The minimum jump threshold The calculation formula is: ; in, Represents the diagonal length of the volume data, is the maximum number of sampling points.

[0018] Furthermore, step S3 specifically includes the following steps: S3.1, normalize the current ray sampling coordinates and read the volume data value of the current ray sampling coordinates.

[0019] S3.2, copy the volume data value to the three RGB channels and expand it into grayscale color, obtain the color value size of the sampling point, and use the color value size and the transparency of the sampling point to perform color accumulation synthesis.

[0020] S3.3, obtain volume data, gradient amplitude, and viewing distance, determine the importance based on the attributes of the sampling points, and dynamically adjust the sampling density. The weight calculation formula for volume data value and gradient amplitude is: ; ; in, represents the gradient magnitude weight; Represents the weight of volume data value; The number of intervals into which the volume data values ​​are divided; The gradient amplitude falls within the interval The probability of The volume data value falls within the interval probability.

[0021] For viewpoint distance, the viewing range of the sampling point is divided into three segments: far, medium, and near, and a corresponding weight coefficient is assigned to each interval: ; in, is the viewpoint distance weight.

[0022] S3.4, combining the gradient amplitude weight, volume data value weight, and viewpoint distance weight, we obtain the adaptive sampling step adjustment formula: ; in, 、 、 are the normalized volume data gradient amplitude, volume data value and position distance, is the step length before adjustment, is the adjusted step size.

[0023] The present invention has the following beneficial effects: The present invention uses Chebyshev distance map (CDM) to quickly jump large-scale empty blocks to avoid invalid block traversal; The present invention implements empty voxel jumping based on unsigned distance field (UDF) inside non-empty blocks to avoid redundant sampling; The present invention combines volume data values, gradient amplitude and viewpoint distance to dynamically adjust the sampling step size and improve sampling efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work. In the drawings: Figure 1 A flow chart of a volume data rendering acceleration method for blank space jumping according to the present invention is shown. DETAILED DESCRIPTION

[0025] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0026] like Figure 1 The volume data rendering acceleration method for blank space jumping shown in the figure specifically includes the following steps: S1: Divide the volume data into blocks of fixed size. As the ray traverses the volume data, determine whether the volume data block to which the current ray sampling point belongs contains a non-empty voxel. If it is a blank volume data block, the process enters the Chebyshev jump phase, in which the blank volume data block is jumped based on the Chebyshev distance map. If it is not a blank volume data block, the process enters the voxel-level jump phase.

[0027] In step S2, during the voxel-level jump phase, the voxel is determined to be non-empty. If so, the ray casting sampling process is executed. If not, the unsigned distance map-based jump phase is entered to jump to the empty voxels. In step S3, during the ray casting sampling process, the sampling step size is adaptively adjusted based on the gradient amplitude, volume data value, and viewpoint distance.

[0028] S4, when calculating the intersection of the light and the volume data as the incident point and the exit point, calculate the travel distance m from the incident light to the exit light. When sampling the volume data each time, calculate the distance n that the light has traveled in the cube. If n>=m, it means that the light has exited the cube, and the iteration stops.

[0029] The method provided by the present invention quickly jumps invalid volume data contained in the volume data. Based on the volume rendering blank space jump acceleration method of the Chebyshev distance map and the volume rendering blank space jump acceleration method of the three-dimensional unsigned distance, the present invention proposes a volume rendering blank space jump acceleration method based on layered distance field and importance driving, thereby further reducing invalid voxel sampling and improving overall rendering efficiency. The method integrates the Chebyshev distance map and the unsigned distance field, respectively, to implement the block-level (Block-Level) and voxel-level (Voxel-Level) space jump mechanism, effectively eliminating redundant calculations of empty voxels. In addition, for the valid voxel area, the present invention integrates the dynamic step size adjustment mechanism of the volume data value, gradient amplitude and viewing distance to achieve adaptive adjustment of the sampling density.

[0030] Specifically, the volume rendering blank space jump acceleration method based on the Chebyshev Distance Map (CDM) first divides the volume data into blocks according to a fixed size, and traverses each data block in turn to determine whether there are non-empty voxels in it. If a block contains at least one non-empty voxel, it is considered a non-empty block, otherwise it is an empty block. The occupancy map constructed based on this records 0 or 1 for each block, indicating that the block is empty or occupied, respectively. Based on the occupancy map, a Chebyshev distance map is further constructed, and each voxel stores the minimum number of empty blocks required to cross from the current position along any direction to the nearest non-empty block. Step S1 specifically includes the following steps: S1.1, assuming that the volume data dimension is , the occupancy map dimension is , the block size is , then for the normalized sampling point coordinates , the coordinates of the volume data in the occupancy map for: .

[0031] S1.2, assuming that the normalized step length between two adjacent sampling points is ,but Step size in the occupancy map for: .

[0032] S1.3, based on the occupancy map, verifies whether the current block is empty. If so, the process enters the Chebyshev jump phase, which includes skipping the remainder of the current block, skipping empty blocks based on the Chebyshev distance, and then skipping subsequent empty blocks based on the Chebyshev distance to reach the nearest non-empty block.

[0033] Specifically, step S1.3 includes the following steps: S1.3.1, in a certain dimension The formula for jumping up the remaining part of the current block is: ; in, Indicates the current sampling point in the occupancy map The non-normalized coordinates in dimensions, Indicates the first The sampling distance in the dimension direction, To round down.

[0034] S1.3.2, based on the Chebyshev distance, the number of blank data blocks skipped is to calculate the Chebyshev distance from the current empty block to the nearest non-empty block through the occupancy map, and generate the corresponding Chebyshev distance map; Chebyshev distance Represents two vectors Each coordinate component in The maximum absolute value of the difference: ; in, To take the maximum value.

[0035] S1.3.3, combined with steps S1.3.2 and S1.3.3, the calculation formula for the jump distance in the Chebyshev jump phase is: ; in, Indicates the distance to jump to the next non-empty data block at the current position. represents the sign function, which is used to determine the sampling direction; is the Heaviside step function, which is used to handle special logic for negative direction jumps. Indicates the offset of the current point in the original empty block, To get the minimum value, This jump strategy comprehensively considers the minimum effective jump steps in the three dimensions, ensuring the maximum jump distance from the current position, effectively reducing unnecessary sampling and improving rendering efficiency.

[0036] S1.3.4, to simplify the implementation and further improve the jump efficiency, the present invention performs an approximate process on the spatial jump method based on the Chebyshev distance map. The original method requires the precise calculation of the jump step size based on the ray direction and the block boundary position to avoid out-of-bounds missampling; the present invention directly multiplies the shortest block-level distance stored in the distance map by the number of voxels contained in each block, estimates the maximum number of voxels that can be jumped, and completes the jump along the ray direction accordingly. The jump distance calculation formula of step S1.3.3 is simplified to obtain: .

[0037] The simplification is to directly multiply the shortest block-level distance stored in the distance map by the number of volume elements contained in each block to estimate the maximum jumpable distance.

[0038] This strategy omits the complex logic for determining block boundaries, significantly reducing computational overhead. Subsequent comparative experiments have shown that this simplified strategy can produce images with good visual quality while maintaining high rendering efficiency.

[0039] Specifically, a volume rendering white space jump acceleration method based on a 3D unsigned distance field (UDF) achieves precise voxel-level spatial jumps by querying the distance from the current point to the nearest non-empty voxel. Compared to jump methods based on Chebyshev distance maps, this method has higher spatial resolution and jump accuracy. Step S2 specifically includes the following steps:

[0040] S2.1, traverse all voxels in the volume data and determine whether each voxel is non-empty; set the UDF value of non-empty voxels to 0, and initialize the distance of other empty voxels to positive infinity;

[0041] S2.2, for empty voxels, the unsigned distance graph jump phase is entered to perform distance propagation calculations along the x, y, and z axes in sequence. Each phase is based on the distance information of the previous phase, and a bidirectional traversal or alternating search strategy is adopted to calculate the distance value of each voxel to the nearest non-empty voxel using the Euclidean distance formula. The three-dimensional unsigned distance field is generated by iteratively taking the minimum value. The three-dimensional unsigned distance field stores the Euclidean distance from the current voxel to the nearest non-empty voxel. The calculation formula is: ; in, is the distance from the current voxel to the nearest non-empty voxel, To represent the distance value of the queried offset voxel in the x-axis direction, is the current voxel coordinate, To take the minimum.

[0042] S2.3, since the spatial jump method based on the 3D unsigned distance field directly calculates the skippable sampling distance based on the distance value of the current sampling point, when the distance value is small, small step jumps may occur frequently, which will introduce additional computational overhead and reduce rendering efficiency. This invention introduces a minimum jump step limit strategy to set a minimum jump threshold based on the volume data scale. Each time the UDF distance value is queried, the UDF distance value is compared with the minimum jump threshold. When the distance value is greater than the minimum jump threshold, a jump is performed, thereby effectively avoiding inefficient small-step jumps and improving overall rendering performance. Minimum jump threshold The calculation formula is: ; in, Represents the diagonal length of the volume data, is the maximum number of sampling points.

[0043] Specifically, to further improve the rendering efficiency of non-empty voxel areas, the present invention introduces the concept of importance-driven and proposes an adaptive sampling step adjustment strategy based on gradient amplitude, volume data value, and viewpoint distance. This strategy determines the importance based on the attributes of the sampling points and dynamically adjusts the sampling density: in areas with a long viewing distance, small gradient amplitude, or low volume data value, the voxels in these areas contribute less to the final image and therefore have lower importance. The sampling step is dynamically increased for these areas to reduce redundant calculations and improve overall rendering performance. Step S3 specifically includes the following steps: S3.1, normalize the current ray sampling coordinates and read the volume data value of the current ray sampling coordinates.

[0044] S3.2, copy the volume data value to the three RGB channels and expand it into grayscale color, obtain the color value size of the sampling point, and use the color value size and the transparency of the sampling point to perform color accumulation synthesis.

[0045] S3.3, obtain volume data, gradient amplitude, and viewing distance, determine the importance based on the attributes of the sampling points, and dynamically adjust the sampling density. The weight calculation formula for volume data value and gradient amplitude is: ; ; in, represents the gradient magnitude weight; Represents the weight of volume data value; The number of intervals into which the volume data values ​​are divided; The gradient amplitude falls within the interval The probability of The volume data value falls within the interval probability.

[0046] For viewpoint distance, the viewing range of the sampling point is divided into three segments: far, medium, and near, and a corresponding weight coefficient is assigned to each interval: ; in, is the viewpoint distance weight.

[0047] S3.4, combining the gradient amplitude weight, volume data value weight, and viewpoint distance weight, derives an adaptive sampling step adjustment formula to dynamically adjust the sampling step, thereby improving rendering efficiency while ensuring image quality: ; in, 、 、 are the normalized volume data gradient amplitude, volume data value and position distance, is the step length before adjustment, is the adjusted step size.

[0048] This strategy can effectively control the sampling density without significantly affecting the image quality.

[0049] To verify the effectiveness of the proposed method, five volume datasets of different sizes and sparsity levels were selected from the Open Science Visualization Dataset Platform for simulation experiments, and the results of different methods were compared. The datasets are shown in Table 1:

[0050] Table 1 Dataset and its key indicators

[0051] While maintaining the same rendering window size, the rendering rates of five volume rendering methods are compared: the classic ray casting method, the space jumping method based on the occupancy map, the volume rendering blank space jumping acceleration method based on the Chebyshev distance map, the volume rendering blank space jumping acceleration method based on the unsigned distance field, and the volume rendering blank space jumping acceleration method based on the layered distance field and importance-driven proposed in this paper. The comparison results are shown in Table 2.

[0052] Table 2 Comparison of rendering acceleration effects of different methods on different datasets

[0053] The comparative experimental results are shown in Table 2. In terms of acceleration performance, the proposed method improves the average frame rate by 1.4-4.4 times on various datasets, significantly outperforming traditional ray casting algorithms (1.0x) as well as comparison methods such as OCCM (1.1x-3.1x), CDM (1.2-4.1x), and UDF (1.2-2.7x). This advantage is most prominent on the sparse aneurism dataset (1.01%) (4.38x speedup), fully verifying the effectiveness of the proposed method in accelerating empty voxel jumps.

[0054] The method of the present invention can significantly improve the rendering efficiency of volume data, thereby accelerating the rendering efficiency of volume data. While accelerating the rendering, it is necessary to ensure that the image quality is not significantly affected. The present invention further uses the structural similarity index (SSIM) to quantitatively analyze the rendering results. The experiment uses the rendering results of the traditional ray casting algorithm as a reference standard. By calculating the SSIM values ​​of the rendered images of each acceleration method and the reference image, the balance between performance improvement and quality maintenance of different methods is objectively evaluated. The results are shown in Table 3.

[0055] Table 3 Pulse signal screening and parameter estimation results

[0056] In terms of image quality preservation, quantitative evaluation of the structural similarity index SSIM shows that the method provided by the present invention can maintain a high fidelity of more than 0.95 (range 0.9587-0.9996) in all test scenarios, indicating that it can effectively improve the rendering speed while ensuring the rendering quality.

[0057] Therefore, the volume rendering blank space jump acceleration method based on layered distance field and importance-driven proposed in the present invention can effectively improve the rendering speed of volume data while ensuring high rendering quality.

[0058] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by technicians in this technical field within the essential scope of the present invention should also fall within the scope of protection of the present invention.

Claims

1. A volume data rendering acceleration method for blank space jumping, characterized in that: The specific steps include: S1: Divide the volume data into blocks of fixed size. When the ray traverses the volume data, determine whether there is a non-empty voxel in the volume data block to which the current ray sampling point belongs. If it is a blank volume data block, enter the Chebyshev jump phase and jump the blank volume data block based on the Chebyshev distance map. If it is not a blank volume data block, enter the voxel-level jump phase. S2, in the voxel-level jump phase, it is determined whether the voxel is non-empty. If it is non-empty, the ray casting sampling process is executed. If it is empty, the unsigned distance map-based jump phase is entered to jump blank voxels. S3, in the ray casting sampling process, adaptive sampling step size adjustment is performed based on gradient amplitude, volume data value and viewpoint distance; S4, calculate the distance m traveled from the incident light to the outgoing light. Each time the volume data is sampled, calculate the distance n that the light has traveled in the cube. If n>=m, it means that the light has exited the cube, and the iteration is stopped.

2. The method for accelerating volume data rendering for blank space jumping according to claim 1, characterized in that: Step S1 specifically includes the following steps: S1.1, assuming that the volume data dimension is , the occupancy map dimension is , the block size is , then for the normalized sampling point coordinates , the coordinates of the volume data in the occupancy map for: ； S1.2, assuming that the normalized step length between two adjacent sampling points is ,but Step size in the occupancy map for: ； S1.3, based on the occupancy map, verify whether the current block is an empty block; if it is empty, enter the Chebyshev jump phase, which includes: skipping the remaining part of the current block and skipping blank volume data blocks based on the Chebyshev distance.

3. The method for accelerating volume data rendering for blank space jumping according to claim 2, characterized in that: Step S1.3 specifically includes the following steps: S1.3.1, in a certain dimension The formula for jumping up the remaining part of the current block is: ； in, Indicates the current sampling point in the occupancy map The non-normalized coordinates in dimensions, Indicates the first The sampling distance in the dimension direction, To round down; S1.3.2, based on the Chebyshev distance, the number of blank data blocks skipped is to calculate the Chebyshev distance from the current empty block to the nearest non-empty block through the occupancy map, and generate the corresponding Chebyshev distance map; Chebyshev distance Represents two vectors Each coordinate component in The maximum absolute value of the difference: ； in, To obtain the maximum value; S1.3.3, combined with steps S1.3.2 and S1.3.3, the calculation formula for the jump distance in the Chebyshev jump phase is: ； in, Indicates the distance to jump to the next non-empty data block at the current position. represents a symbolic function; is the Heaviside step function, Indicates the offset of the current point in the original empty block, To get the minimum value, To round up; S1.3.4, simplify the jump distance calculation formula in step S1.3.3 to obtain: 。 4. The method for accelerating volume data rendering for blank space jumping according to claim 1, characterized in that: Step S2 specifically includes the following steps: S2.1, traverse all voxels in the volume data and determine whether each voxel is non-empty; set the UDF value of non-empty voxels to 0, and initialize the distance of other empty voxels to positive infinity; S2.2, for empty voxels, enter the jump stage based on the unsigned distance map; perform distance propagation calculations along the x, y, and z axes in sequence. Each stage is based on the distance information of the previous stage, using a bidirectional traversal or alternating search strategy, and calculates the distance value of each voxel to the nearest non-empty voxel using the Euclidean distance formula. The three-dimensional unsigned distance field is generated by iteratively taking the minimum value. The three-dimensional unsigned distance field stores the Euclidean distance from the current voxel to the nearest non-empty voxel. The calculation formula is: ； in, is the distance from the current voxel to the nearest non-empty voxel, To represent the distance value of the queried offset voxel in the x-axis direction, is the current voxel coordinate, To take the minimum; S2.3, set the minimum jump threshold based on the volume data scale Each time the UDF distance value is queried, the UDF distance value is compared with the minimum jump threshold. When the distance value is greater than the minimum jump threshold, a jump is performed. The minimum jump threshold The calculation formula is: ； in, Represents the diagonal length of the volume data, is the maximum number of sampling points.

5. The method for accelerating volume data rendering for blank space jumping according to claim 1, characterized in that: Step S3 specifically includes the following steps: S3.1, normalize the current ray sampling coordinates and read the volume data value of the current ray sampling coordinates; S3.2, copy the volume data value to the three RGB channels and expand it into grayscale color, obtain the color value of the sampling point, and use the color value and the transparency of the sampling point to perform color accumulation synthesis; S3.3, obtain volume data, gradient amplitude, and viewing distance, determine the importance based on the attributes of the sampling points, and dynamically adjust the sampling density. The weight calculation formula for volume data value and gradient amplitude is: ； ； in, represents the gradient magnitude weight; Represents the weight of volume data value; The number of intervals into which the volume data values ​​are divided; The gradient amplitude falls within the interval The probability of The volume data value falls within the interval probability; For viewpoint distance, the viewing range of the sampling point is divided into three segments: far, medium, and near, and a corresponding weight coefficient is assigned to each interval: ； in, is the viewpoint distance weight; S3.4, combining the gradient amplitude weight, volume data value weight, and viewpoint distance weight, we obtain the adaptive sampling step adjustment formula: ； in, 、 、 are the normalized volume data gradient amplitude, volume data value and position distance, is the step length before adjustment, is the adjusted step size.

Images