A method and system for calculating a stereoscopic image visual masking threshold

Abstract

The application relates to a method and system for calculating a stereoscopic image visual masking threshold, and belongs to the technical field of image visual distortion evaluation and application, and solves the problem of low accuracy of a stereoscopic image visual masking threshold in the prior art. The method comprises the following steps: acquiring binocular views of a stereoscopic image to be calculated, calculating a monocular masking threshold for each view based on brightness adaptability, contrast masking and direction modulation; judging a binocular interaction mode according to the noise direction of each view; and calculating a binocular masking threshold of the stereoscopic image based on the monocular masking threshold and neural gain modulation according to the binocular interaction mode. More accurate calculation of the binocular masking threshold is realized.

Description

Technical Field

[0001] This invention relates to the field of image visual distortion assessment and application technology, and in particular to a method and system for calculating the visual masking threshold of stereoscopic images. Background Technology

[0002] In recent years, significant progress has been made in the generation and processing of stereoscopic images / videos, with a growing focus on achieving "high perceptual quality" rather than just "high definition" to provide a more immersive experience. However, stereoscopic image processing often introduces noticeable distortion, negatively impacting the viewing experience. Therefore, ensuring high-quality perceptual performance while simultaneously improving the usability, controllability, and security of images and videos remains a challenge.

[0003] Visual masking refers to the phenomenon where the presence of a masking signal reduces the perceptibility of a given target signal. It is an inherent property of the visual system and plays an important role in stereo perception. Visual masking models are used to measure the maximum imperceptible distortion, i.e., Just Noticeable Distortion (JND) or masking threshold. Information hiding techniques, such as digital watermarking and encrypted communication, are used to hide information in images and videos without being detected. However, excessive embedding can lead to distortion and quality degradation, and focusing solely on ensuring high perceptual quality can reduce the strength and capacity of the embedding. Therefore, the study of visual masking effects and models is crucial in information hiding. Determining the JND threshold through visual masking models helps optimize the intensity of distortion while maintaining the perceived image / video quality. Similarly, research related to image and video compression coding calculates perceptual redundancy in images and videos using distortion perception models, and filters out details in images that are imperceptible to the visual system (HVS) through preprocessing or encoded embedding to achieve data compression.

[0004] Stereoscopic masking models play a crucial role in accurately simulating human binocular visual perception and obtaining corresponding masking thresholds. In the formation of binocular perception, visual masking occurs in two distinct loops. One loop involves masking effects within each monocular view, including brightness adaptation, contrast masking, and texture masking. These effects have been extensively studied in visual masking models. The other loop involves interocular visual masking, which is related to the mechanisms of binocular synthesis. Physiological studies have shown that the mechanisms of interocular interaction differ between binocular fusion and competition. When binocular views are similar, the brain merges them to generate a consistent and stable perception, where both views contribute to the stereoscopic experience while being partially masked by the contralateral eye. When there are significant differences between binocular views, binocular competition occurs. To eliminate binocular vision, the brain selectively suppresses some or all of the visual input in one view. The suppressed stimuli are effectively masked by the dominant view, leading to alternating perception between the two different views.

[0005] However, existing models have limited coverage and cannot fully capture the complexity and variability of the masking effect, leading to discrepancies between the determined threshold and psychophysical perception. Specifically, existing models have two main shortcomings:

[0006] (1) Existing models mainly consider the masking effect of background patterns on targets in 2D vision, and extend these effects in the case of binocular fusion. The masking effects involved are very limited, and they cannot describe the interocular masking effect caused by binocular interaction during the formation of stereo perception, nor can they describe the modulation of masking effect and threshold by changes in binocular target features. Current stereo vision-related models often regard visual masking as a means of determining the existence of binocular competition, rather than as an independent research topic. In addition, the human eye's sensitivity to direction affects the visibility and detection of stimuli, but research on how directional sensitivity affects the masking threshold in stereo vision is limited.

[0007] (2) Existing BJND models and their derivatives have measured the masking threshold of asymmetric noise in the case of binocular fusion. These models describe the process of combining binocular noise into monocular perception by using an equal-weighted summation method, completely ignoring the weight changes caused by binocular asymmetry, and failing to consider the role of interocular interaction in regulating stereo perception, resulting in the model being inconsistent with the principles and neural regulation mechanisms of human binocular perception.

[0008] The above two main defects prevent existing models from achieving optimal distortion intensity while ensuring subjective quality in applications such as stereo image information hiding and compression coding, and also prevent them from accurately obtaining the visual masking threshold of stereo images. Summary of the Invention

[0009] Based on the above analysis, the embodiments of the present invention aim to provide a method and system for calculating the visual masking threshold of stereoscopic images, so as to solve the problem of low accuracy of the visual masking threshold of stereoscopic images obtained by existing models.

[0010] On one hand, embodiments of the present invention provide a method for calculating the visual masking threshold of a stereoscopic image, including the following steps:

[0011] Obtain binocular views of the stereo image to be calculated, and calculate monocular masking thresholds for each view based on brightness adaptability, contrast masking, and orientation modulation.

[0012] The binocular interaction mode is determined based on the noise direction of each view;

[0013] Based on the binocular interaction mode, the binocular masking threshold of the stereo image is calculated using the monocular masking threshold and neural gain modulation.

[0014] Further improvements to the above method involve calculating a monocular visual masking threshold for each view based on brightness adaptability, contrast masking, and orientation modulation, including:

[0015] Convert the view to a grayscale image;

[0016] The brightness adaptability threshold is directionally modulated;

[0017] The monocular masking threshold is calculated based on the contrast masking threshold and the luminance adaptive threshold after direction modulation.

[0018] Based on the further improvement of the above method, the monocular masking threshold is calculated using the following formula:

[0019]

[0020]

[0021] p x (i,j)=α1×bg x (i,j) 2 +α2×bg x (i,j)+α3

[0022] Among them, lm L (i,j) represents the adaptive threshold at position (i,j) of the left eye view, bg x (i,j) represents the grayscale value at position (i,j) in the grayscale image of the monocular view, L represents the left view, R represents the right view, α1, α2, α3 represent the model parameters, and δ L (i,j) represents the orientation modulation term of the left view at position (i,j), δ R (i,j) represents the direction modulation term of the right view at position (i,j). This represents the monocular masking threshold of the left view at position (i,j). This represents the monocular masking threshold of the right view at position (i,j), in cm. L (i,j) represents the contrast masking threshold at position (i,j) in the left view, in cm. R (i,j) represents the contrast masking threshold at position (i,j) in the right view, p x (i,j) represents the modulation parameters at position (i,j) for monocular vision, α1, α2, and α3 represent the fitting parameters, and θ x Indicates the direction of noise in the left and right views.

[0023] Based on a further improvement of the above method, the brightness adaptability threshold is calculated using the following formula based on the grayscale image:

[0024] lm x (i,j)=p1·bg x (i,j) 5 +p2·bg x (i,j) 4 +p3·bg x (i,j) 3 +p4·bg x (i,j) 2 +p5·bg x (i,j)+p6,x∈{L,R}

[0025] Among them, lm x (i,j) represents the adaptive threshold at position (i,j) of the monocular view, bg x (i,j) represents the gray value at position (i,j) in the grayscale image of the monocular view, L represents the left view, R represents the right view, and p1, p2, p3, p4, p5 and p6 represent the fitting parameters.

[0026] Based on a further improvement of the above method, the contrast masking threshold is calculated using the following formula based on the grayscale image:

[0027] cm x (i,j)=K x (i,j)×eh x (i,j),x∈{L,R}

[0028] K x (i,j)=β1×(β2×bg x (i,j) 2 +β3×bg x (i,j))+β4

[0029] Among them, cm x(i,j) represents the contrast masking threshold at position (i,j) in the monocular view, bg x (i,j) represents the grayscale value at position (i,j) in the grayscale image of the monocular view, L represents the left view, R represents the right view, and eh x (i,j) represents the edge filtering result at position (i,j) in the monocular view, K x (i,j) represents the contrast masking threshold enhancement factor at position (i,j) of the monocular view, and β1, β2, β3 and β4 are fitting parameters.

[0030] Based on the further improvement of the above method, when the noise directions of the left view and the right view are the same, the binocular interaction mode is the binocular fusion mode; when the noise directions of the left view and the right view are different, the binocular interaction mode is the binocular competition mode.

[0031] Based on further improvements to the above method,

[0032] When the binocular interaction mode is binocular competition mode, the binocular masking threshold for the stereo image is:

[0033]

[0034] Among them, A L A represents the stereo vision masking threshold matrix for the left view. R This represents the stereo vision masking threshold matrix for the right view. This represents the monocular masking threshold of the left view. This represents the monocular masking threshold for the right view.

[0035] Based on the further improvement of the above method, when the binocular interaction mode is the binocular fusion mode, the binocular masking threshold of the stereo image is calculated using the following formula:

[0036]

[0037] Among them, A L A represents the stereo vision masking threshold matrix for the left view. R This represents the stereo vision masking threshold matrix for the right view. g represents the monocular masking threshold of the left view. c γ represents the gain modulation threshold, B represents the adjustment factor, and α represents the noise intensity perceived by the binoculars.

[0038] Compared with existing technologies, the method for calculating the stereoscopic image visual masking threshold provided in this embodiment first calculates monocular visual masking based on brightness adaptability, contrast masking, and orientation modulation, taking into account the perceptual distortion and visual acuity of the human eye when viewing stereoscopic images. Simultaneously, orientation modulation is used to consider the modulation relationship between orientation features and the brightness adaptability and contrast masking threshold, simulating the change of the masking threshold with the target orientation, thus calculating the masking threshold more accurately. By considering binocular interaction modes, different binocular masking threshold calculation methods are adopted according to different binocular interaction modes, overcoming the deficiency of existing stereoscopic visual masking models that lack consideration of the masking effect caused by inter-eye interaction, thus calculating the binocular masking threshold more accurately. Based on the above scheme, it is possible to improve the distortion intensity that stereoscopic images can tolerate while ensuring that the subjective quality of the stereoscopic image is not compromised. Therefore, in practical applications, it can improve technical indicators, such as the information embedding strength in information hiding technology and the compression rate in image compression technology, and can better balance the relationship between various technical indicators in image processing technology and image quality (viewing experience).

[0039] On the other hand, embodiments of the present invention provide a system for calculating a visual masking threshold for stereoscopic images, comprising the following modules:

[0040] The monocular masking threshold calculation module is used to acquire binocular views of the stereo image to be calculated, and calculate the monocular masking threshold for each view based on brightness adaptability, contrast masking and orientation modulation.

[0041] The interaction mode determination module is used to determine the binocular interaction mode based on the noise direction of each view;

[0042] The binocular masking threshold calculation module is used to calculate the binocular masking threshold of the stereo image based on the monocular masking threshold and neural gain modulation according to the binocular interaction mode.

[0043] Based on further improvements to the above system, a monocular visual masking threshold is calculated for each view based on brightness adaptability, contrast masking, and orientation modulation, including:

[0044] Convert the view to a grayscale image;

[0045] The brightness adaptability threshold is directionally modulated;

[0046] The monocular masking threshold is calculated based on the contrast masking threshold and the luminance adaptive threshold after direction modulation.

[0047] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from the description and drawings, which are particularly pointed out. Attached Figure Description

[0048] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0049] Figure 1 This is a flowchart illustrating the method for calculating the visual masking threshold of a stereoscopic image according to an embodiment of the present invention;

[0050] Figure 2 This is a block diagram of the system for calculating the visual masking threshold of stereoscopic images according to an embodiment of the present invention. Detailed Implementation

[0051] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0052] To address the shortcomings mentioned in the background technology, a large number of psychophysical experiments were conducted to study the stereoscopic visual masking effect, and the following findings were obtained:

[0053] Under binocular fusion conditions, increasing the masking threshold of one view will lead to decreasing the masking threshold of the other view; however, under binocular competition conditions, the masking threshold is largely independent of the threshold change of the other view.

[0054] Under binocular fusion and binocular competition conditions, the masking threshold of oblique noise is higher than that of horizontal and vertical noise; in addition, under medium brightness conditions, the threshold increase due to directional change is more significant than that at other brightness levels.

[0055] The neurocomputational theory explains binocular fusion as follows: during binocular fusion, neurons in the visual cortex provide excitatory signals to neurons on the same side while simultaneously imposing inhibitory signals to neurons on the contralateral side. Neurons in the visual cortex are regulated by both internal and external factors. Physiological studies show that the brain achieves adaptive responses to changing inputs through a series of modulated neural computations, namely neural gain modulation (also known as gain control). This modulation enables neurons to optimize the processing of sensory information by enhancing the representation of relevant stimuli and suppressing the influence of irrelevant information. Neural gain is a metric for measuring the adaptive response of neurons to different inputs. There is a strong correlation between the neuronal population response based on gain modulation and psychophysical performance. Therefore, a stereoscopic masking model needs to incorporate neural gain modulation to represent interocular masking effects, making it closer to human eye perception.

[0056] Based on psychophysical experimental results and neural gain modulation theory, a specific embodiment of the present invention discloses a method for calculating the visual masking threshold of stereoscopic images, such as... Figure 1 As shown, it includes the following steps:

[0057] S1. Obtain the binocular view of the stereo image to be calculated, and calculate the monocular masking threshold for each view based on brightness adaptability, contrast masking and orientation modulation.

[0058] S2. Determine the binocular interaction mode based on the noise direction of each view;

[0059] S3. The stereo image's stereo masking threshold is calculated based on the monocular masking threshold and neural gain modulation according to the stereo interaction mode.

[0060] In practice, the binocular views of the stereo image to be calculated are the left and right views of the stereo image. It should be noted that this invention is implemented under binocular symmetry. Under binocular symmetry, the left and right eyes are equivalent, and L and R correspond to the two views of the binocular stereo image, but do not necessarily correspond specifically to the left and right eyes.

[0061] The stereoscopic image visual masking model using the method for calculating the stereoscopic image visual masking threshold of this invention is called the IM-JND model.

[0062] Compared with existing technologies, the method for calculating the stereoscopic image visual masking threshold provided in this embodiment first calculates monocular visual masking based on brightness adaptability, contrast masking, and orientation modulation, taking into account the perceptual distortion and visual acuity of the human eye when viewing stereoscopic images. Simultaneously, orientation modulation is used to consider the modulation relationship between orientation features and the brightness adaptability and contrast masking threshold, simulating the change of the masking threshold with the target orientation, thus calculating the masking threshold more accurately. By considering binocular interaction modes, different binocular masking threshold calculation methods are adopted according to different binocular interaction modes, overcoming the deficiency of existing stereoscopic visual masking models that lack consideration of the masking effect caused by inter-eye interaction, thus calculating the binocular masking threshold more accurately. Based on the above scheme, it is possible to improve the distortion intensity that stereoscopic images can tolerate while ensuring that the subjective quality of the stereoscopic image is not compromised. Therefore, in practical applications, it can improve technical indicators, such as the information embedding strength in information hiding technology and the compression rate in image compression technology, and can better balance the relationship between various technical indicators in image processing technology and image quality (viewing experience).

[0063] Specifically, in step S1, a monocular visual masking threshold is calculated for each view based on brightness adaptability, contrast masking, and orientation modulation, including:

[0064] Convert the view to a grayscale image;

[0065] The brightness adaptability threshold and contrast masking threshold are calculated based on the grayscale image, respectively.

[0066] The brightness adaptability threshold is directionally modulated;

[0067] The monocular masking threshold is calculated based on the contrast masking threshold and the luminance adaptive threshold after direction modulation.

[0068] During implementation, the left and right views are converted into grayscale images respectively. For example, when converted to an 8-bit quantized grayscale image, the pixel values ​​in the grayscale image are numbers in the range of 0 to 255.

[0069] Then, based on the grayscale images of the left and right views, the brightness adaptation threshold and contrast masking threshold of the left and right views are calculated respectively.

[0070] Brightness adaptability refers to the eye's ability to adapt to changes in environmental brightness, and this threshold is calculated based on grayscale values. In practice, psychophysical experiments are conducted to collect data on the brightness adaptability threshold exhibited by the human eye under different image brightness levels. A fifth-order polynomial fitting is then used to derive the relationship between the brightness adaptability threshold and grayscale values. Finally, the brightness adaptability threshold of the view is calculated based on this experimentally obtained relationship.

[0071] Specifically, the brightness adaptability threshold is calculated based on the grayscale image using the following formula:

[0072] lm x (i,j)=p1·bg x (i,j) 5 +p2·bg x (i,j) 4 +p3·bg x (i,j) 3 +p4·bg x (i,j) 2 +p5·bg x (i,j)+p6,x∈{L,R} (1)

[0073] Among them, lm x (i,j) represents the adaptive threshold at position (i,j) of the monocular view, bg x (i,j) represents the gray value at position (i,j) in the grayscale image of the monocular view, L represents the left view, R represents the right view, and p1, p2, p3, p4, p5 and p6 represent the fitting parameters, that is, the parameters in the fitting relationship between the brightness adaptability threshold and the grayscale value.

[0074] During implementation, the fitting parameters were p1 = -7.622 × 10⁻⁶. -11 p2 = 5.564 × 10 -8 p3 = -1.465 × 10 -5 , p4=0.001697, p5=-0.0766, p6=2.46.

[0075] Contrast masking refers to the masking effect of a high-contrast target on the surrounding low-contrast area.

[0076] In practice, the contrast masking threshold is calculated based on the grayscale image using the following formula:

[0077] cm x (i,j)=K x (i,j)×eh x (i,j),x∈{L,R} (2)

[0078] K x (i,j)=β1×(β2×bg x (i,j) 2 +β3×bg x (i,j))+β4 (3)

[0079] Among them, cm x (i,j) represents the contrast masking threshold at position (i,j) in the monocular view, bg x(i,j) represents the grayscale value at position (i,j) in the grayscale image of the monocular view, L represents the left view, R represents the right view, and β1, β2, β3, and β4 are fitting parameters. x (i,j) represents the edge filtering result at position (i,j) in the monocular view, K x (i,j) represents the contrast masking threshold enhancement factor at position (i,j) of the monocular view.

[0080] During implementation, the following formula can be used to calculate eh. x (i,j):

[0081]

[0082]

[0083] Among them, G k (h,v) denotes a 5×5 Sobel operator, where H represents the horizontal direction, V represents the vertical direction, and G represents the vertical direction. H G represents the convolution kernel used to detect horizontal changes in an image. H E represents the convolution kernel used to detect vertical variations in an image. H E represents the horizontal change of the view. V This represents the vertical change in the view. In practice, similarly, psychophysical experiments were conducted to collect the contrast masking threshold exhibited by the human eye under different image brightness levels, and the relationship between the contrast masking threshold and the grayscale value was fitted. In practice, the fitting parameter was β1 = -10. -6 , β2=0.7, β3=32, β4=0.07.

[0084] In practice, eh(i,j) is defined by the 5×5 Sobel operator G at position (i,j). H and G V Calculated.

[0085]

[0086] Because the human eye has varying sensitivities to different directions, the noise masking threshold also differs in different directions. Psychophysical experiments show that, at the same brightness, the threshold first increases and then decreases as the direction θ changes from vertical (θ=0) to horizontal (θ=π / 2). Therefore, after obtaining the brightness adaptive threshold, directional modulation is applied to it, based on the contrast masking threshold and the directionally modulated brightness adaptive threshold.

[0087] Specifically, the monocular masking threshold is calculated using the following formula:

[0088]

[0089]

[0090] p x (i,j)=α1×bg x (i,j) 2 +α2×bg x (i,j)+α3 (8)

[0091] Among them, lm L (i,j) represents the adaptive threshold at position (i,j) of the left eye view, bg x (i,j) represents the grayscale value at position (i,j) in the grayscale image of the monocular view, L represents the left view, R represents the right view, α1, α2, α3 represent the model parameters, and δ L (i,j) represents the orientation modulation term of the left view at position (i,j), δ R (i,j) represents the direction modulation term of the right view at position (i,j). This represents the monocular masking threshold of the left view at position (i,j). This represents the monocular masking threshold of the right view at position (i,j), in cm. L (i,j) represents the contrast masking threshold at position (i,j) in the left view, in cm. R (i,j) represents the contrast masking threshold at position (i,j) in the right view, p x (i,j) represents the modulation parameters at position (i,j) for monocular vision, α1, α2, and α3 represent the fitting parameters, and θ x Indicates the direction of noise in the left and right views.

[0092] In practice, for the stereo image to be calculated, the left and right views can first be directionally filtered based on the texture features of the image using a filter, such as a Gabor filter, to obtain the texture directions of the left and right views. For each view, multiple texture directions may be obtained. The direction modulation term is a convex function with respect to the noise direction. As the direction θ changes from vertical (θ=0) to horizontal (θ=π / 2), the threshold first increases and then decreases. According to equation (7), when θ x for δ x Since (i,j) is the largest, we take the closest values ​​for the left and right views respectively. The texture direction is the noise direction.

[0093] In practice, a psychophysical experiment was conducted by adding binocular noise in different directions to pairs of stereo images with varying background brightness, resulting in a series of stereo test images with different brightness and directions of binocular noise. For each pair of stereo images (background brightness, noise direction), initially, the noise intensity of both the left and right views was 0. The noise intensity of the left view was fixed at 0, and the noise intensity of the right view was continuously increased from 0 in increments. Subjects viewed these images until they could just perceive the noise, at which point the noise intensity of the left and right views was recorded. Then, the noise intensity of the left view was increased by one increment, and the noise intensity of the right view was continuously increased from 0 in increments. Subjects viewed these images until they could just perceive the noise, and the noise intensity of the left and right views was recorded. This process was repeated until the noise intensity of the left view increased to the point where the noise intensity of the right view was 0, at which point the subject could just perceive the noise, and the noise intensity of the left and right views was recorded, ending the experiment.

[0094] For each set of stereo images (background brightness, noise direction), when the noise intensity in the left view is 0, the noise in the right view is just perceptible. The monocular masking threshold is such that when the noise intensity in the right view is 0, the noise in the left view is just perceptible.

[0095] During implementation, the noise direction θ x Using radians, the range of noise direction variation is:

[0096] Then, the parameters α1, α2, and α3 can be obtained by fitting according to the above formula. In practice, the fitted parameters are α1 = 1.297 × 10 -5 , α2=-0.002769, α3=-0.106.

[0097] Psychophysical experiments have shown that under binocular fusion conditions, an increase in the masking threshold of one view leads to a decrease in the masking threshold of the other view; while under binocular competition conditions, the masking threshold is largely independent of the threshold change of the other view.

[0098] To more closely approximate human stereoscopic perception, stereoscopic masking models need to encompass a more comprehensive and complete range of stereoscopic masking effects, especially considering that interocular interaction is the most crucial element in the formation of stereoscopic perception. When constructing a stereoscopic masking model, the role of interocular interaction in modulating stereoscopic perception needs to be considered to more accurately simulate human visual perception. Therefore, this invention overcomes the deficiency of existing stereoscopic masking models that lack consideration of the masking effects caused by interocular interaction. When calculating the binocular masking threshold, it also considers the influence of binocular target orientation configuration on interocular interaction, and calculates the binocular masking threshold of the stereoscopic image based on the binocular interaction pattern.

[0099] Specifically, the binocular interaction mode is determined based on the noise directions of the left and right views. When the noise directions of the left and right views are the same, the binocular interaction mode is a binocular fusion mode; when the noise directions of the left and right views are different, the binocular interaction mode is a binocular competition mode. That is, when the noise directions θ of the left and right views are the same... L ≠θ R This can trigger binocular competition; when the noise direction θ of the left and right views... L =θ R At this time, it will trigger the fusion of the two eyes.

[0100] In the case of binocular competition, the masking thresholds of the left and right views are independent of each other. Therefore, under the condition of binocular competition, the binocular masking threshold is obtained by the following formula:

[0101]

[0102] Among them, A L A represents the stereo vision masking threshold matrix for the left view. R This represents the stereo vision masking threshold matrix for the right view. This represents the monocular masking threshold of the left view. This represents the monocular masking threshold for the right view.

[0103] For binocular fusion, considering the mutual influence of binocular masking thresholds, this invention proposes a two-layer iterative gain-controlled nonlinear binocular summation function based on neural gain control theory to effectively describe the relationship between masking thresholds and stereo perception in binocular views, as shown in the following equation:

[0104]

[0105] Among them, A L A represents the stereo vision masking threshold matrix for the left view. R ε represents the stereoscopic masking threshold matrix of the right view. L (A L ) and ε R (A R) represent the noise energy matrices for the left and right views, respectively. B represents the noise intensity perceived by both eyes.

[0106] In binocular fusion (assuming zero disparity between the left and right views), the inputs for the left and right views are the same, therefore... exist The relationship between the noise energy ε and the noise energy is shown in the following formula. Where g c γ is the threshold for gain modulation, and γ is the adjustment factor.

[0107]

[0108] Among them, g c and and Positive correlation, as shown in the following formula, where p7 is g c Fitting parameters:

[0109]

[0110] Therefore, according to equations (10)-(12), in the case of binocular fusion, the binocular masking threshold of the stereo image is calculated using the following formula:

[0111]

[0112] Among them, A L A represents the stereo vision masking threshold matrix for the left view. R This represents the stereo vision masking threshold matrix for the right view. g represents the monocular masking threshold of the left view. c Let represent the gain modulation threshold, γ represent the adjustment factor, B represent the noise intensity perceived by the binoculars, and α represent the parameter. In practice, the fitted result was p7 = 0.6395, and the adjustment factor γ = 1.49.

[0113] Equation (13) The following formula is A obtained from equations (10) and (11). R A quintic equation in one variable. In the case of binocular symmetry... According to equation (14), given B and A... L When solving, we get A R .

[0114] This invention employs a nonlinear summation function and performs two-layer iterative gain control based on neural gain control theory, effectively describing the relationship between masking threshold and stereo perception in binocular vision.

[0115] The effectiveness of the IM-JND model was verified through subjective validation experiments using psychophysical methods. Noise from different directions (covering both binocular fusion and binocular competition scenarios) was added to real texture images in the experiments. The noise intensity was calculated using both the IM-JND model presented in this invention and the existing BJND model. The experiments showed that the IM-JND model has good agreement with the psychophysical experimental data and effectively captures the masking phenomenon in stereoscopic vision perception. By comparison, the IM-JND model can reduce the peak signal-to-noise ratio (PSNR) by 5.06%–15.05% without compromising image perception quality (a lower PSNR indicates the addition of higher intensity noise).

[0116] A specific embodiment of the present invention discloses a system for calculating the visual masking threshold of a stereoscopic image, such as... Figure 2 It includes the following modules:

[0117] The monocular masking threshold calculation module is used to acquire the binocular view of the stereo image to be calculated, and calculate the monocular masking threshold for each view based on brightness adaptability, contrast masking and orientation modulation.

[0118] The interaction mode determination module is used to determine the binocular interaction mode based on the noise direction of each view;

[0119] The binocular masking threshold calculation module is used to calculate the binocular masking threshold of the stereo image based on the monocular masking threshold and neural gain modulation according to the binocular interaction mode.

[0120] Preferably, a monocular visual masking threshold is calculated for each view based on brightness adaptability, contrast masking, and orientation modulation, including:

[0121] Convert the view to a grayscale image;

[0122] The brightness adaptability threshold is directionally modulated;

[0123] The monocular masking threshold is calculated based on the contrast masking threshold and the luminance adaptive threshold after direction modulation.

[0124] The above-described method and system embodiments are based on the same principles, and their related aspects can be referenced from each other to achieve the same technical effects. For specific implementation processes, please refer to the foregoing embodiments, which will not be repeated here.

[0125] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0126] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for calculating the visual masking threshold of a stereoscopic image, characterized in that, Includes the following steps: Obtain binocular views of the stereo image to be calculated, and calculate monocular masking thresholds for each view based on brightness adaptability, contrast masking, and orientation modulation. The binocular interaction mode is determined based on the noise direction of each view; The binocular masking threshold of the stereo image is calculated based on the monocular masking threshold and neural gain modulation according to the binocular interaction mode. A monocular visual masking threshold is calculated for each view based on brightness adaptability, contrast masking, and orientation modulation, including: Convert the view to a grayscale image; The brightness adaptability threshold is directionally modulated; The monocular masking threshold is calculated based on the contrast masking threshold and the brightness adaptive threshold after direction modulation; The monocular masking threshold is calculated using the following formula: in, Indicates the left eye view The fitness threshold of location, Grayscale image representing a monocular view The grayscale values ​​for the location, where L represents the left view and R represents the right view. Indicates the left view in Position direction modulation term, Indicates the right view is in Position direction modulation term, Indicates the left view in The monocular masking threshold for the location, Indicates the right view is in The monocular masking threshold for the location, Represents the left view Contrast masking threshold at location, Represents the right view Contrast masking threshold at location, Indicates monocular vision Position modulation parameters, , , Represents the fitted parameters, Indicates the direction of noise in the left and right views; The brightness adaptability threshold is calculated based on the grayscale image using the following formula: , in, Represents a monocular view The fitness threshold of location, Grayscale image representing a monocular view The grayscale values ​​for the location, where L represents the left view and R represents the right view. and Indicates the fitted parameters; The contrast masking threshold is calculated based on the grayscale image using the following formula: in, Represents a monocular view Contrast masking threshold at location, Grayscale image representing a monocular view The grayscale values ​​for the location, where L represents the left view and R represents the right view. Represents a monocular view Location edge filtering results Represents a monocular view Contrast masking threshold enhancement factor at location , and These are the fitting parameters.

2. The method for calculating the visual masking threshold of a stereoscopic image according to claim 1, characterized in that, When the noise directions of the left and right views are the same, the binocular interaction mode is a binocular fusion mode; when the noise directions of the left and right views are different, the binocular interaction mode is a binocular competition mode.

3. The method for calculating the visual masking threshold of a stereoscopic image according to claim 2, characterized in that, When the binocular interaction mode is binocular competition mode, the binocular masking threshold for the stereo image is: in, This represents the stereo vision masking threshold matrix for the left view. This represents the stereo vision masking threshold matrix for the right view. This represents the monocular masking threshold of the left view. This represents the monocular masking threshold for the right view.

4. The method for calculating the visual masking threshold of a stereoscopic image according to claim 2, characterized in that, When the binocular interaction mode is binocular fusion mode, the binocular masking threshold of the stereo image is calculated using the following formula: in, This represents the stereo vision masking threshold matrix for the left view. This represents the stereo vision masking threshold matrix for the right view. This represents the monocular masking threshold of the left view. The threshold representing gain modulation. B represents the adjustment factor, and B represents the noise intensity perceived by the binoculars. Indicates a parameter.

5. A system for calculating the visual masking threshold of a stereoscopic image, characterized in that, Includes the following modules: The monocular masking threshold calculation module is used to acquire binocular views of the stereo image to be calculated, and calculate the monocular masking threshold for each view based on brightness adaptability, contrast masking and orientation modulation. The interaction mode determination module is used to determine the binocular interaction mode based on the noise direction of each view; The binocular masking threshold calculation module is used to calculate the binocular masking threshold of the stereo image based on the monocular masking threshold and neural gain modulation according to the binocular interaction mode. A monocular visual masking threshold is calculated for each view based on brightness adaptability, contrast masking, and orientation modulation, including: Convert the view to a grayscale image; The brightness adaptability threshold is directionally modulated; The monocular masking threshold is calculated based on the contrast masking threshold and the brightness adaptive threshold after direction modulation; The monocular masking threshold is calculated using the following formula: in, Indicates the left eye view The fitness threshold of location, Grayscale image representing a monocular view The grayscale values ​​for the location, where L represents the left view and R represents the right view. Indicates the left view in Position direction modulation term, Indicates the right view is in Position direction modulation term, Indicates the left view in The monocular masking threshold for the location, Indicates the right view is in The monocular masking threshold for the location, Represents the left view Contrast masking threshold at location, Represents the right view Contrast masking threshold at location, Indicates monocular vision Position modulation parameters, , , Represents the fitted parameters, Indicates the direction of noise in the left and right views; The brightness adaptability threshold is calculated based on the grayscale image using the following formula: , in, Represents a monocular view The fitness threshold of location, Grayscale image representing a monocular view The grayscale values ​​for the location, where L represents the left view and R represents the right view. and Indicates the fitted parameters; The contrast masking threshold is calculated based on the grayscale image using the following formula: in, Represents a monocular view Contrast masking threshold at location, Grayscale image representing a monocular view The grayscale values ​​for the location, where L represents the left view and R represents the right view. Represents a monocular view Location edge filtering results Represents a monocular view Contrast masking threshold enhancement factor at location , and These are the fitting parameters.

Images