STAIR code stripe level identification method for dynamic three-dimensional measurement

Abstract

The invention relates to the technical field of optical three-dimensional measurement, in particular to an STAIR code stripe level identification method for dynamic three-dimensional measurement. The method comprises the following specific steps: generating projection fringes through a variable phase shift coding strategy, projecting the projection fringes on a measured object through projection equipment, collecting a deformed fringe image on the measured object, calculating a truncation phase and a step phase modulated by the height of the measured object according to the deformed fringe image, and calculating the phase of the measured object according to the truncation phase. Converting the step phase into a discrete integer coding level diagram to obtain an initial integer coding level diagram; dynamically performing perceptual directional mode filtering correction on the initial integer coding level diagram according to the truncation phase value to obtain a corrected coding level diagram; and the absolute stripe level is distinguished based on the displacement principle. By adopting the STAIR code stripe level identification method based on dynamic three-dimensional measurement, full-field stripe level identification can be realized under a high-frequency condition, and the pattern decoding reliability and robustness under the high-frequency condition are improved.

Description

Technical Field

[0001] This invention relates to the field of optical three-dimensional measurement technology, and in particular to a method for identifying the STAIR code stripe levels in dynamic three-dimensional measurement. Background Technology

[0002] Fringe projection profilometry (FPP) is an important optical 3D measurement technique. Its basic principle is to project a series of structured light fringe patterns onto the surface of an object using a projector. A camera captures the fringe image, which is distorted by the object's height modulation. Then, phase calculation and unwrapping are used to recover the object's absolute phase, and finally, the object's 3D shape is reconstructed based on system calibration parameters. In FPP, due to the periodicity of the arctangent function, the directly calculated phase value is confined to the interval (-π, π], known as the truncated phase. A continuous absolute phase must be obtained through phase unwrapping techniques. Temporal phase unwrapping (TPU) technology identifies the order of each fringe by projecting additional coded patterns, thus achieving point-by-point phase unwrapping and exhibiting better robustness to complex surfaces and isolated objects.

[0003] Traditional TPU methods, such as multi-frequency heterodyne or Gray code methods, typically require projecting multiple patterns, usually more than 10, which severely limits the measurement speed and makes them difficult to apply to dynamic scenes. Although some Variant Phase Shift Coding (VPSC) methods attempt to acquire phase and order information simultaneously with fewer patterns, their order recognition reliability drops sharply under high-frequency fringe projection, and they are prone to order misidentification due to noise, motion blur, and uneven surface reflection, leading to reconstruction errors.

[0004] Therefore, there is an urgent need for a method that offers high measurement speed, accuracy, and robustness, and remains applicable even under conditions of noise, motion ambiguity, and uneven surface reflection. Summary of the Invention

[0005] The purpose of this invention is to provide a method for identifying the stripe levels of STAIR codes using dynamic three-dimensional measurement, thereby solving the aforementioned technical problems.

[0006] To achieve the above objectives, this invention provides a method for identifying the stripe order of STAIR codes using dynamic three-dimensional measurement, the specific steps of which are as follows: Step S1: Generate projection stripes using a variable phase shift coding strategy, project the projection stripes onto the object under test using a projection device, and acquire images of the deformed stripes on the object under test. Step S2: Calculate the truncated phase and step phase modulated by the height of the measured object based on the deformed fringe image; Step S3: Convert the step phase in step S2 into a discrete integer coding level map to obtain an initial integer coding level map; dynamically perform perceptual directional modal filtering correction on the initial integer coding level map according to the truncated phase value of each pixel to obtain the corrected coding level map; Step S4: Identify the absolute fringe order based on the principle of order displacement; Step S4 is as follows: Step S41: Measure the coding level map and absolute stripe level at the reference point, and determine the maximum stripe level offset; Step S42: Construct a mapping relationship between encoding levels and stripe levels based on the data from step S41; Step S43: Obtain the initial level shift based on the corrected coding level diagram and the mapping relationship in step S42; Step S44: Determine the absolute fringe order value of the object point by resolving order ambiguity through spatial monotonicity constraint parameters.

[0007] Preferably, in step S1, the variable phase-shift coding strategy generates the following projected fringe intensity: in, For the first The intensity of the frame stripe image, , , as well as All are constants. The pixel coordinates of the projection device. For a stepped phase that includes fringe order, , For coding levels, The half-length of the ladder phase code, For phase spacing, , This represents the maximum value of the encoding level. The variable phase-shift coding strategy uses a periodic coding sequence, with each period being... The sequence length is related to the stripe frequency, and each number in the sequence corresponds to a stripe period.

[0008] Preferably, in step S2, the acquired deformed stripe image is as follows: in, It is the first Frame-distorted stripe image, , and These are the background light intensity and modulation amplitude, respectively. To truncate the phase, The phase is the step phase; the calculation formula is as follows: .

[0009] Preferably, in step S3, the step phase is converted into a discrete integer coding level diagram, and the conversion formula is as follows: in, This is a hierarchy diagram for discrete integer encoding.

[0010] Preferably, in step S3, the corresponding directional filter is dynamically selected based on the truncated phase value of each pixel. The sensing directional mode filtering correction process is as follows: When pixel When filtering is needed, a right-window filter is used. When pixel At that time, a center window filter is used for filtering; When pixel At that time, a left window filter is used for filtering; Window size is , The width of the filter window is half its pixel count. The neighborhood set is as follows: The corrected coding level diagram is as follows , This is used to calculate the mode operator.

[0011] Preferably, in step S41, the object being measured is within the calibration volume, and the maximum fringe order offset is related to the maximum depth change of the reference plane. In step S42, at a given maximum order displacement Under these conditions, a lookup table is constructed based on the mapping relationship between the coding level and the stripe level. The formula for the lookup table is as follows: in, and Points on the camera image plane The corresponding stripe level lookup table and code level lookup table, To represent the absolute order distribution of fringes on the reference plane, This represents the corresponding coding level distribution on the reference plane.

[0012] Preferably, in step S43, the initial level displacement calculation formula is as follows: in, For the initial order displacement, This function returns the index of a specified element in the lookup table, with the encoding level and the shift level being the same as the stripe level.

[0013] Preferably, the process for resolving hierarchical ambiguity in spatial monotonicity constraint parameters is as follows: The encoding level graph is labeled with multiple gray-level connected components, and a unique label is assigned to each phase period. The feature function... for: in, For a given pixel A collection of tags; The forward difference graph and backward difference graph of this domain are defined as follows: in, and They are respectively Forward difference plot and backward difference plot, The characteristic functions of the left and right boundaries are as follows: in, and They are respectively The left and right boundary neighborhoods; The boundary order is obtained from the above as follows: in, and They are left-level and right-level, respectively. monotonicity parameter The calculation formula is as follows: Through monotonic parameters Determine the monotonicity of the corresponding pixel for resolution. Ambiguity in the text.

[0014] Therefore, the present invention employs the above-mentioned dynamic three-dimensional measurement method for STAIR code stripe level recognition, which has the following beneficial effects: (1) Based on the principle of order displacement, the encoding and decoding of stripe order is realized. The robustness of phase expansion is ensured by using local monotonicity and spatial pattern inference. Full-field stripe order recognition is realized under high frequency conditions, thereby improving the reliability of pattern decoding under high frequency conditions.

[0015] (2) The truncated phase-sensing directional modal filtering is adopted. Through the adaptive window selection and majority voting pattern filtering mechanism, misalignment artifacts and spike artifacts are corrected, so that the hierarchical recognition of complex surface patterns can be effectively completed in complex scenes, and the three-dimensional reconstruction of the measured object can be realized.

[0016] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0017] Figure 1 This is a flowchart of a dynamic three-dimensional measurement method for identifying the stripe levels of a STAR code according to the present invention.

[0018] Figure 2 The diagram shows a comparison between the present invention and conventional encoding, wherein (a) from top to bottom are stripe intensity curve, truncated phase curve, conventional continuous phase image and image of this embodiment, and (b) from top to bottom are encoding curve of this embodiment, conventional connected phase encoding cross-section and encoding cross-section of this embodiment.

[0019] Figure 3 The diagram shows a comparison between the present invention and a traditional stripe hierarchy, where (a) is a traditional stripe hierarchy diagram and (b) is a stripe hierarchy diagram of this embodiment.

[0020] Figure 4 The diagram shows the process of identifying the absolute fringe order based on the order displacement principle. (a) is the fringe order lookup table and the coded order lookup table, (b) is the unambiguous retrieval under the right projection, (c) is the ambiguous retrieval under the right projection, and (d) is the ambiguous retrieval under the left projection.

[0021] Figure 5 The images shown are from Experiment 1. (a) is an image of the standard ceramic ball; (b) is a deformed fringe pattern after the standard ceramic ball is projected with stripes; (c) is the encoding order pattern before correction; (d) is the encoding order pattern after correction; and (f) is the absolute fringe order pattern.

[0022] Figure 6 The following are the relevant error plots for Experiment 1, where (a) is the spatial depth error plot for sphere I, (b) is the spatial depth error plot for sphere II, (c) is the two-dimensional error histogram for sphere I, and (d) is the two-dimensional error histogram for sphere II. Figure 7 The images are related to Experiment 2, where (a) is an actual image of the complex plaster sculpture; (b) is a deformed fringe image of the complex plaster sculpture after the fringe is projected; (c) is the coding level image before correction; (d) is the coding level image after correction; and (f) is the absolute fringe level image.

[0023] Figure 8The following are comparison results for Experiment 2: (a) is the result of traditional 3D reconstruction, (b) is the result of 3D reconstruction using three-step phase shift + complementary Gray code, (c) is the result of 3D reconstruction using this embodiment, and (d) is a comparison curve between three-step phase shift + complementary Gray code and this embodiment.

[0024] Figure 9 The following are magnified views of the relevant three-dimensional reconstructions in Experiment 3, where (a) is a magnified view of the traditional three-dimensional reconstruction; (b) is a magnified view of the three-step phase-shift + complementary Gray code three-dimensional reconstruction; and (c) is a magnified view of the three-dimensional reconstruction in this embodiment.

[0025] Figure 10 This is a schematic diagram of the equivalent single-frame reconstruction process.

[0026] Figure 11 The deformation correlation diagrams for the pressure cushion in Experiment 4 are shown, where (a) is a picture of the actual pressure cushion, (b)-(e) are three-dimensional reconstructed point clouds at four different times, and (f)-(i) are three-dimensional reconstructed depth maps at four different times.

[0027] Figure 12 The images are related to Experiment 5, where (a) is the acquired image with marked rotation feature points and background intensity map, (b) are the intensity map, stripe map and stripe level map at different times from top to bottom, (c) are the three-dimensional point cloud at different times, and (d) are the rotation trajectory fitting curves of the left and right quarter spheres. Detailed Implementation

[0028] In the description of this invention, it should be noted that the terms "upper," "lower," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product is in use. They are used only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," and "connect" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0029] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0030] like Figure 1 As shown in the figure, this embodiment presents a method for identifying the stripe levels of a STAR code using dynamic three-dimensional measurement. The specific steps are as follows: Step S1: Generate projected fringes using a variable phase-shift coding strategy, such as... Figure 2 As shown, the method in this embodiment uses fewer independent stages to achieve stage recognition of high-frequency phases. The variable phase-shift coding strategy uses a periodic coding sequence, with each period being [missing information]. The sequence length is related to the stripe frequency, and each number in the sequence corresponds to a stripe period. Figure 2 Figure (b) shows a comparison between a traditional monotonic sequence and the sequence of this embodiment. This embodiment uses 8 independent digits to encode 32 stripes, and the quantization interval between adjacent levels is significantly larger than that of a traditional monotonic continuous sequence. If the half-width of the quantization interval... The phase interval of the monotonically continuous order is This embodiment The stripe recognition conditions are: This embodiment improves the noise tolerance of the level identification process by 400%. From Figure 3 As can be seen, compared with conventional methods that generally produce random level errors in the entire image, this example only produces a small number of error lines in local areas, demonstrating higher level recognition accuracy and stability.

[0031] Projection stripes are projected onto the object being measured using a projection device, and images of the deformed stripes on the object are captured.

[0032] Step S2: Calculate the truncated phase and step phase modulated by the height of the object under test based on the deformed fringe image.

[0033] Step S3: Convert the stepped phase from Step S2 into a discrete integer coding level map to obtain an initial integer coding level map. To ensure the robustness and accuracy of the final level recovery process, the initial integer coding level map needs to be corrected. Error bars and misalignment artifacts exist in the corrected coding level. Based on the truncated phase value of each pixel, a suitable directional filter is dynamically / adaptively selected, and local neighborhood information in a specific direction is used to achieve accurate error detection and correction. The initial integer coding level map is dynamically corrected using perceptual directional modal filtering based on the truncated phase value, resulting in the corrected coding level map. Because truncated phase perceptual directional modal filtering dynamically adapts to the truncated phase of each pixel, it not only avoids interference from adjacent periods but also achieves strict alignment between the coding level and the truncated phase transition at the pixel level. Therefore, truncated phase perceptual directional modal filtering can simultaneously correct error bars and misalignment artifacts.

[0034] The data on the effectiveness of the perception-oriented modal filtering correction are shown in Table 1.

[0035] Table 1. Effectiveness data of sensing-oriented modal filtering correction in this embodiment.

[0036] Given the widespread and robust application of median filtering in traditional time-phase expansion methods, this paper selects it as the baseline method. Compared with the proposed filtering algorithm, median filtering exhibits a higher mean absolute error (MAE) in phase error, a larger root mean square error (RMSE) in depth reconstruction, generates more erroneous pixels, and leads to a decrease in overall accuracy. In contrast, truncated phase-aware directional modal filtering can effectively correct boundary errors, overcome the limitations of traditional filtering, and significantly improve the decoding robustness and overall reconstruction quality under the encoding strategy in this embodiment.

[0037] Step S4: Identify the absolute fringe order based on the principle of order displacement. The identification process is as follows: Figure 4 As shown, if the absolute stripe level and encoding level of a pixel on the reference plane are 26 and 4 respectively, and its encoding level on the object is 8, and the index of the first element is set to 1, the level shift is calculated to be 4, then the absolute stripe level of the object point is 22.

[0038] To ensure the uniqueness of the order displacements, additional criteria (or constraints) are introduced to resolve ambiguity, such as the monotonicity indicator. , Indicates an increasing trend. Indicating a decreasing trend, in Figure 4 In (c), the encoded value 6 is in It appears twice, but the corresponding monotonicity is different, according to Distinguish them and resolve ambiguities in their levels.

[0039] To verify the performance of this embodiment, the following test experiments were conducted: Experiment 1: This experiment demonstrates the setup of the experimental system. The projector uses a high-speed DLP LightCrafter 4500 with a native resolution of 912×1140 pixels; the imaging end uses a Hikvision MV-CA013-21UC industrial camera, capturing images in grayscale mode with a resolution of 1024×1280 pixels. The center distance between the camera and the projector is approximately 202.82 mm, and their optical axis angle is approximately 24.44°. The captured image contains 76 fringe periods, each period occupying 12 pixels in width horizontally.

[0040] The test objects in Experiment 1 were a pair of standard ceramic spheres with radii of, respectively, as shown in Figure 1. Figures 5-6As shown, the spheres are 25.391 mm and 25.388 mm, and the center-to-center distance between them is 100.048 mm. The truncated phase and the initial step phase are reconstructed separately. Obvious fringe order boundary error lines are visible in the coarse step phase. Correction is performed using the truncated phase and a perceptual directional modal filtering algorithm to obtain a clean and refined step phase. Combining the truncated phase with the corrected absolute fringe order yields the 3D reconstruction result. The fitting data for the standard sphere is shown in Table 2.

[0041] Table 2. Fitting data for standard spheres (unit: mm)

[0042] Table 1 presents the fitting results for sphere I and sphere II, including the coordinates of the sphere centers, radius, radius error, root mean square error relative to the true value, and the distance between the sphere centers and its deviation. This demonstrates that the method of this embodiment has high measurement accuracy.

[0043] Experiment 2: The objects measured in Experiment 1 are different, such as... Figures 7-8 As shown, the static scene measured in this experiment includes a complex plaster statue to verify the effectiveness of the algorithm in reconstructing complex, isolated scenes. Compared with the traditional phase unrolling method, which shows significant quantization error, this embodiment is comparable to the ground truth obtained using three-step phase shifting + complementary Gray code (CGC), verifying the superior performance of the technical solution in high-frequency 3D reconstruction.

[0044] Experiment 3: The objects measured are different from those in Experiment 2. Figure 9 As shown, this embodiment measures a static umbrella, providing a magnified view of the TPU reconstruction results. The selected area corresponds to the printed text on the umbrella surface, which is an extremely thin material layer with uneven reflectivity, making it highly susceptible to noise. Despite the harsh conditions, this embodiment accurately captures the details of the thin printed layer and maintains the overall texture in low-reflectivity areas. Compared with the ground truth (GT), no key details are lost, thus achieving high-quality reconstruction.

[0045] Experiment 4: Unlike the object measured in Experiment 1, the dynamic scenario in this experiment is the deformation of a pressure cushion.

[0046] The camera captured 4000 frames at 200Hz. Compared to the static experiment, the dynamic experiment requires equivalent single-frame reconstruction, such as... Figure 10 As shown, for the first frame, using four initial fringe patterns, the truncated phase and phase shift change are calculated, and the absolute fringe order is calculated. Based on this, phase unwrapping and 3D reconstruction are completed. For each subsequent frame, only one additional fringe pattern is needed to form a four-pattern group with equivalent phase shift to complete the reconstruction. This process is recursively applied frame by frame, and dynamic reconstruction can be achieved by adding only one image per frame.

[0047] like Figure 11As shown, this experiment recorded the deformation of the pillow during the compression and rebound process, showing the transition from an irregular surface pattern to a smooth surface pattern. This indicates that the system can continuously track the surface geometry that evolves over time, capture significant temporal pattern changes, and reliably identify complex surface patterns and accurately recover subtle geometric features even under conditions of uneven reflectivity and dynamics.

[0048] Experiment 5: This experiment differs from Experiment 4 in that it uses a rotating platform to track the circular motion trajectory of feature points of an object, such as... Figure 12 As shown in Figure 12(a), the acquired image and background intensity map with annotated rotation feature points are presented. Figure 12(b) shows the intensity map, fringe map, and fringe order map at different times during the rotation. The corresponding 3D point cloud is shown in Figure 12(c). To reveal the motion pattern, Figure 12(d) shows the rotation trajectories of the left and right quarter-spheres, respectively. The fitted xz-plane center and radius are: left (–67.50, –663.37) mm and right (–67.90, –663.51) mm, respectively. The results are highly consistent with the recorded motion video, verifying the robustness of the method in dynamic reconstruction in this embodiment.

[0049] Experiments 1-5 demonstrate that the method of this embodiment can achieve full-field stripe level recognition under high-frequency conditions, improve the reliability of pattern decoding under high-frequency conditions, and effectively complete the level recognition and three-dimensional reconstruction of complex surface patterns in complex scenarios such as single-pixel-level motion, non-uniform reflectivity, and dynamic deformation, showing its application potential in industrial inspection, in-situ metrology, and dynamic three-dimensional pattern recognition.

[0050] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for identifying the order of a STAIR code stripe in a dynamic three-dimensional measurement, characterized by The specific steps are as follows: Step S1: generating a projection fringe by a variable phase shift coding strategy, projecting the projection fringe on the measured object by a projection device, and collecting a deformed fringe image on the measured object; Step S2: calculating a truncated phase and a step phase modulated by the measured object according to the deformed fringe image; Step S3: converting the step phase in step S2 into a discrete integer coding order map to obtain an initial integer coding order map; the initial integer coding order map is dynamically corrected by a perceptual orientation modal filter according to the truncated phase value of each pixel point to obtain a corrected coding order map; Step S4: identifying the absolute fringe order based on the order displacement principle; Step S4 is as follows: Step S41: measuring the coding order map and the absolute fringe order at the reference point, and determining the maximum fringe order displacement; Step S42: constructing a mapping relationship between the coding order and the fringe order according to the data in step S41; Step S43: obtaining an initial order displacement according to the corrected coding order map and the mapping relationship in step S42; Step S44: determining the absolute fringe order value of the object point by eliminating the order ambiguity through a spatial monotonicity constraint parameter.

2. The method of claim 1, wherein the method is a method of STAIR fringe order identification for dynamic 3D measurement. In step S1, the variable phase shift coding strategy generates a projection fringe intensity as follows: wherein is the first frame fringe image, , , and are constants, is a pixel coordinate of the projection device, is a step phase containing fringe orders, , is an encoding order, is a half length of the step phase encoding, is a phase distance, , is a maximum value of the encoding order; The variable phase-shift coding strategy employs a periodic coding sequence, each period of which is where the sequence length is related to the stripe frequency, and each number in the sequence corresponds to a stripe period.

3. The method of claim 2, wherein the method further comprises: determining the phase of the STAIR code stripe pattern based on the phase of the first and second phase signals. In step S2, the collected deformed fringe image is as follows: wherein is the first frame deformation fringe image, , and are the background light intensity and the modulation amplitude, respectively, is the cut-off phase, is the step phase; the calculation formula is as follows: 。 4. The method of claim 3, wherein the method further comprises: determining the phase of the STAIR code stripe pattern based on the phase of the first and second phase signals. In step S3, the step phase is converted into a discrete integer coding order map, and the conversion formula is as follows: wherein is a discrete integer encoding level map.

5. The method of claim 4, wherein the method further comprises: determining the order of the STAIR code stripe based on the first and second phase values. In step S3, the corresponding directional filter is dynamically selected according to the truncated phase value of each pixel, and the perceptual directional modal filter correction process is as follows: When the pixel point is a right window filter is used for filtering; When the pixel point is a center window filter is used for filtering; When the pixel point is the left window filter is used for filtering; The window size is , is the half width of the filter window, and the neighborhood set of the pixel point is as follows: The corrected encoding level graph is , is the statistical majority operator.

6. The method of claim 5, wherein the method further comprises: determining the order of the STAIR code stripe based on the first and second phase values. In step S41, the measured object is in a calibration volume, and the maximum fringe order displacement is related to the maximum depth variation of the reference plane; In step S42, a lookup table is constructed according to the mapping relationship between the encoding order and the stripe order under the condition that the maximum order displacement is given, and the lookup table formula is as follows: wherein and are the coordinates of the point corresponding fringe order lookup table and encoded order lookup table, is a representation of the distribution of absolute fringe orders on the reference plane, is a representation of the corresponding encoded order distribution on the reference plane.

7. The method of claim 6, wherein the method further comprises: determining the order of the STAIR code stripe based on the first and second phase values. In step S43, the initial order displacement calculation formula is as follows: wherein, is the initial order displacement, is a function that returns the specified element index in the lookup table, the coding order is the same as the stripe order displacement order.

8. The method of claim 7, wherein the method further comprises: determining the order of the STAIR code stripe based on the first and second phase values. The process of eliminating order ambiguity through a spatial monotonicity constraint parameter is as follows: The coding level graph is used to label the multi-gray connected domain, and a unique label is given to each phase period. The characteristic function is: f(x, y) = i, wherein, a label set for a given pixel point ; The forward difference map and the backward difference map of the domain are defined as follows: wherein and are respectively the forward difference map and the backward difference map of the characteristic functions of the left and right boundaries of wherein and respectively are left and right boundary neighborhoods of The boundary order is obtained as follows: wherein and L and R are left and right order respectively, monotonicity parameter The calculation formula is as follows: By monotonicity parameter Determining the monotonicity of the pixel point, used to resolve ambiguities in

Images