A three dimensional surface profile imaging method and apparatus described in U.S. Pat. No. 5,675,407 (“the '407 patent”), the disclosure of which is incorporated herein by reference in its entirety, conducts imaging by projecting light through an optical filter, such as linear variable wavelength filter (LVWF), thereby projecting light having a known, spatially distributed wavelength spectrum on the objects being imaged. The LVWF is a rectangular optical glass plate coated with a color-filtering film that gradually varies in color, (i.e., wavelength). If the color spectrum of a LVWF is within the visible light region, one edge of the filter rectangle may correspond to the shortest visible wavelength (i.e. blue or violet) while the opposite edge may correspond to the longest visible wavelength, (i.e. red). The wavelength of light passing through the coated color-filtering layer is linearly proportional to the distance between the position on the filter glass where the light passes and the blue or red edge. Consequently, the color of the light is directly related to the angle θ, shown in
Referring to
The light reflected from the surface of the object (104) is then detected by the camera (102). If a visible spectrum range LVWF (400-700 nm) is used, the color detected by the camera pixels is determined by the proportion of its primary color components, i.e., the Red, Green, and Blue components (RGB). The color spectrum of each pixel has a one-to-one correspondence with the projection angle (θ) of the plane of light due to the fixed geometry of the camera (102) lens and the LVWF characteristics. Therefore, the color of light received by the camera (102) can be used to determine the angle θ at which that light left the rainbow light projector (100). Other spectrum ranges can also be used in similar fashion.
As described above, the angle α is determined by the physical relationship between the camera (102) and the coordinates of each pixel on the camera's imaging plane. The baseline B between the camera's 102 focal point and the center of the cylindrical lens of the light projector (100) is fixed and known. Given the value for angles α and θ, together with the known baseline length B, all necessary information is provided to easily determine the full frame of three-dimensional range values (x,y,z) for any and every visible spot on the surface of the objects seen by the camera (102).
As shown in
As a result, the three-dimensional imaging system described above can capture full-frame, high spatial resolution three-dimensional images using a standard camera, such as a charge coupled device camera, in real time without relying on any moving parts. Further, because the imaging system does not rely on a laser, it does not pose any hazard to the eyes when used in clinical applications. Also, because the wavelength of the light projected onto the object surface continuously varies, there is no theoretical limitation on the measurement accuracy that can be achieved by the system. The actual accuracy of a specific system will depend on system implementation and will be affected primarily by limiting factors such as the optical system design, the quality and resolution of the camera, light spectral emission of the light source projector; noise level and resolution of the frame grabber, calibration algorithms, and the three-dimensional imaging processing algorithms.
To avoid allowing the ambient light on the object being imaged from affecting the imaging results, the system may obtain an image of the object under normal light conditions before projecting the filtered light onto the object. The image obtained under normal light conditions is then subtracted from the image obtained under LVWF light conditions to eliminate the effects of the ambient light on the image.
With reference to
where (xp, yp) is the location of the rainbow light projector, (xc, yc) is the location of imaging sensor, B is the baseline between the rainbow projector and the imaging sensor (CCD), α=π−θ−B, O is a surface point on the object in the scene, and R is the three-dimensional range, that is, the distance between (xc, yc) and O.
Note that all of the variables, θ, α, and B, in the equation (2) may introduce error in the three-dimensional range calculation. In the following error sensitivity analysis, considered with reference to
The dα, dθ, and db are all functions of the position error of the light projector (dxp, dzp) and the color mismatch dx. Notice that:
Therefore
To simplify the analysis, the effect of a mismatch error in the world coordinate, dx, caused by imaging sensor's spectral noise can be indicated as follows:
The linear relationship between the mismatch error dx in the world coordinate and the mismatch error in CCD camera's pixel coordinate, di, allows approximation of dx and di as if they are the same order of magnitude, resulting in the expression:
dx=k*di (7)
where k is a constant. From this expression, the range error in terms of the color mismatch error in CCD pixel coordinate can be expressed as follows:
In the imaging system shown in
The accuracy of the color match operation, of course, affects the accuracy of the three-dimensional surface measurement. Referring to
As shown in
di2<di1. (9)
Using Equ. (8) in Equ. (9), the expression becomes:
dR2<dR1. (10)
The above expressions suggest that mismatch errors can be reduced by using a spatially varying projection pattern that has a large variation rate and therefore a large derivative.
As noted above, the imaging system projects a light pattern that could be described as a “single rainbow projection” pattern because the pattern ranges over the visible light spectrum a single time. This type of pattern, however, has a variation rate that depends on numerous factors, including the field of view of the camera/image sensor, the distance between the light projector and the object being imaged, as well as other factors. As a result, it is difficult to adjust the spectral variation rate of the single rainbow projection pattern to fit the accuracy requirements of many three-dimensional imaging tasks.
Accordingly, the present invention is directed to a system and method that generates a light pattern that contains multiple cycles of light spectra or sub-patterns. In one embodiment, the light pattern includes two or more cycles of spectra of visible or infrared light, arranged contiguously with each other. The light pattern may also be generated by sequentially emitting two of more sub-patterns.
In another embodiment of the invention, the multiple spectra light pattern is generated by projecting the light from the light source through a planar member, such as a slotted panel, that is used in conjunction with color filtering to produce a pattern containing multiple iterations of the visible color spectrum. The multiple spectra used to illuminate the object being imaged improves resolution and measurement accuracy by reducing color mismatch error.
a and 5b are graphs comparing the slopes of the illumination pattern on measurement accuracy.
a and 6b are light pattern spatial displacement graphs corresponding to embodiments of the present invention.
a, 7b and 7c illustrate a multiple spectra light projection pattern according to embodiments of the present invention.
a is an illustration of another embodiment of a system according to the present invention in which a multiple spectra light pattern is used to generate a three-dimensional image of an object or scene.
b is an illustration of the operation of a dichroic filter.
a and 23b illustrate the structure and nomenclature of the human ear.
a and 6b illustrate one manner in which the light projection pattern can be modified to improve resolution and measurement accuracy.
Because this embodiment uses a single rainbow projection pattern as the light pattern, the one-to-one correspondence between the wavelength of a light sheet in the light pattern and its projection angle still applies in this embodiment.
Referring now to
First, adaptive control can be used to determine the initial condition in the search. More particularly, an adaptive initial point calculation mechanism may be used to ensure that the initial point at which the iterative search method begins is at least somewhat close to the correct result, particularly because the final result of iterative processes tends to depend on the selected initial condition. By limiting the search range, as shown in
One method of increasing the chance for having a good initial point in a search is to use an adaptive initial point calculation mechanism. The initial point of the search is determined by a correct match point in the search of a neighboring pixel. The underlying assumption here is that most portions of the surface of a physical object being imaged are continuous. Thus the projected, i.e., reflected, color is similar for neighboring portions of the object, and therefore the initial projection angle value should also be very similar.
As can be seen in
Note that although the look-up table for a multiple rainbow projection system has multiple projection angles associated with any given wavelength when the light pattern is viewed as a whole, the system restricts the search space in the color matching operation within a smaller range (e.g., within one sub-pattern) so that, as a practical matter, there is a one-to-one correspondence between the wavelength and the projection angle when the search is restricted to within the sub-pattern.
Although all of the above examples assume that there is only one camera or image sensor used to capture images of the illuminated object, multiple cameras can also be used to capture multiple rainbow projection pattern data. Referring to
The color spectrum of pixels on the captured images is determined by the proportion of Red, Green and Blue components (RGB) associated with the individual pixel. As a result, pixel-to-pixel registration can be easily performed based on matching the color components of counterparts. Because the length of the baseline B between the two cameras is known, and the geometrical and optical parameters of two cameras can be obtained from a prior calibration procedure, the locations of a pair of corresponding pixels in a stereo image pair provide sufficient information to determine viewing angles α1 and α2. The range values, R, associated with each pixel between a camera and surface points can be obtained using a straightforward triangulation:
Thus, a full frame of a three-dimensional image can be obtained from a single snap shot, and a stream of three-dimensional images can be generated at the camera's frame rate (e.g., 30 frames per second or higher).
By using multiple cameras in conjunction with the multiple rainbow projection system, three-dimensional measurement accuracy is greatly improved, particularly when measuring objects and scenes that do not have many distinguishing surface features. More particularly, the stereo matching and triangulation methods used in this embodiment reduces the dependence of distinguishing features, such as edges and corners, on an object's surface in conducting the measurement.
Referring now to
Note that the above examples assumed that an optical filter, i.e., an LWVF, is used to produce the illumination pattern, such as a rainbow pattern, on the object or scene being imaged. In this type of system the light from the light source is projected through the LWVF to generate the light pattern. The quality of the light pattern, however, depends greatly on the quality of the filter, which can be expensive and difficult to produce.
Consequently, it would be preferable to have a system in which an LWVF is not required to generate the multi-rainbow pattern. It would also preferable to have a system in which the multi-rainbow pattern can be created in a single shot without having to exposing the object or scene of interest to a sequence of three separate light patterns in three different color components.
As indicated above, the accuracy of the color to projection angle matching operation has a major effect on the accuracy of resulting 3D images. The accuracy of the color-to-angle match, in turn, is significantly determined by the color variation rate of the projected light pattern.
This analysis leads to the “Multi-Rainbow Projection (MRP)” concept, as shown in FIG. 13. Instead of changing the wavelength in single cycle, a projection pattern is used in which the wavelength varies across the spectrum with the pattern being repeated several times crossing the field of view.
To implement the MRP scheme, a projector is needed that can generate the multi-rainbow illumination pattern, preferably without requiring costly color filters such as an LWVF. As indicated above, the LVWF in the Rainbow 3D camera is the main contributor to the high cost of the disclosed 3D camera system.
Thus, instead of using an LWVF, a flat plate (114) with slots (115) on it, as shown in
The slot plate (114) discussed above is only capable of generating an intensity varying projection pattern, not a pattern that varies in color or wavelength. To implement the multi-rainbow projection (MRP) concept, we need to produce the zigzag intensity patterns for Red, Green, and Blue (Primary colors) in the projected illumination.
A multiple rainbow projection pattern can be represented as a combination of three individual bands of color, namely Red, Green, and Blue (or other colors in different color space definition). Each band of color components (R, G, B) can be projected independently. These images are then combined together to form a composite “color” image.
As shown in
To vary the intensity of the light pattern from each source, the light may be projected through a slotted plate (P1, P2 and P3). The light from each source (S1, S2 and S3) is then directed to an object of interest (3D object). For example, one light source (S1) may be directed at the object of interest, while the light from the other two sources (S2 and S3) is redirected by a beam-splitter, mirror or half-mirror (F1 and F2) to the object of interested. The result is the desired MRP on the object of interest which can be recorded and processed as described above by a camera (102).
In some embodiment, the light of each source (S1, S2 and S3) is of a different color component. In other embodiments, each of the light sources (S1, S2 and S3) may project white light and a color filter may be provided over each of the slotted plates (P1, P2 and P3) so that light of a particular color is passed through the slots of that plate.
For example, the filtering on one plate may pass red and green light, while blocking blue light. The filtering on another plate may pass green and blue light, while blocking red light. And, finally, the filtering on another plate may pass blue and red light, while blocking green light. This arrangement has been discovered to provide color bands of equal width in the MRP projected onto the object being imaged.
This same effect can also be accomplished with the use of dichroic filters in place of, or with, the mirrors (F1 and F2) or other elements that direct the light of the various light sources to the object of interest. The dichroic spectrum filters (F1 and F2) are used to combine light projections from different spectrum into a combined multi-color pattern. Typical spectrum properties of the dichroic filters are illustrated in
Alignment among multiple projection patterns is important for certain applications. Typically, for three pattern configuration, three pattern filters should be aligned such that their pattern period phases are ⅓ off with respect to each other.
Another possible configuration for the source of the MRP is illustrated in FIG. 17. As shown in
The light from the lens (CL1-CL2) then passes through a pattern filter (P1-P3). The pattern filter (P1-P3) may be a slotted plate as shown in FIG. 14. The pattern filter (P1-P3) give the light a pattern in which strips of light vary continuously in intensity across the width or height of the pattern. (See, FIG. 15).
The three light patterns produced by the three pattern filters (P1-P3) are then combined. This is accomplished with mirrors (M2 and M3) and dichroic filters (DC2-DC3). The dichroic filters (DC2-DC3) operate as described above. Finally, the combined patterns, each representing a different color component, or the absence of a color components, are projected onto the object or scene of interest through a projection lens (CL).
It must be remembered that using an MRP creates a situation in which there is not a one-to-one relationship between projection angle and color. Rather, a given color may be projected at several different angles depending on the number of spectra in the MRP. Therefore, additional procedures are needed for the new 3D Camera to be able to distinguish the correct projection angle from multiple candidates resulting from the one-to-many table lookup.
These procedures are discussed above with regard to
In restricting the search range to a single spectrum, it is useful to use an adaptive control scheme to determine the initial search point. When using a local search method in a one-to-many MRP, the outcome of the search relies heavily upon the initial condition. If a search starts within the neighborhood (e.g., the rainbow cycle) of the solution, the search can achieve a correct result. If the search starts in a point far away from the solution, the search may end up with a wrong answer.
An adaptive control scheme increases the chance of having a good initial point in a search. The initial point of a search is determined by a previously obtained correct match for a point on the object that is adjacent to the point for which range is now being determined. The underlying assumption here is that a large portion of surface of a physical object is continuous, thus the projected color is similar, and projection angle value should be very similar for adjacent points.
The algorithm starts with a known 3D point (provided by the 3D camera hardware design), and uses this point to search for the correct color matching points in its neighborhood (8 points around the starting point). It is then selects a best point among the 8 neighborhood points according to a predefined criteria to be the starting point for the next round of the search process. This recursive process is a simple and effective way to process the adaptive multi-rainbow projection calculation problem.
The colored pattern filters (CP1, CP2, CP3) are used to form desirable illumination properties in each spectrum band. The purpose of the pattern filters is to produce spatially varying intensity of the illumination for each color band. One candidate of such patterns is saw-tooth wave as shown in FIG. 15. Other similar patterns can be easily implemented. The lengths of optical paths for each channel should be approximately the same. Alignment among multiple projection patterns is critical for certain applications. Typically, for three pattern configuration, three pattern filters should be aligned such that their pattern period phases are ⅓ off with respect to each other.
The polarizing filters (PF1 and PF2) are used to combine light projections from different bands. For example, the PS1 and PS2 are combined by using the PF1 based on their polarization directions. The resulting lights are of both polarization directions. A 45-degree polarizer (W1) is used to pick up light from both polarization directions. The first stage combined light now has single polarization, which can be combined with the polarization light from PS3 using the same method with that of the first one. The projected multi-spectrum pattern onto the 3D object will be imaged by an image sensor to acquire three dimensional information as described above using the camera (102).
Each light source is able to produce a single spectrum intensity varying pattern on the object using previously described approaches. The offset between multiple light sources allow for the multiple projection pattern in different spectrum bands to be offset by desirable phases (say ⅓ period).
To this point, a number of embodiments of an improved 3D imaging device have been described. We will now discuss an important application for these devices, the manufacture of custom fit hearing aids or other items to be worn in a patient's ear. Any of the embodiment described herein or in related applications for 3D imaging can be used in the following application, but the improved 3D imaging devices described herein are preferred.
More than 28 million Americans suffer from some type of hearing impairment according to statistics from National Institute on Deafness and Communication Disorders (NIDCD). It is estimated that over 260 million people are hearing impaired worldwide. Based upon the huge population involved, hearing impairment is arguably the number one disability in today's world. Fortunately, many of these people can benefit from the use of a hearing aid. However, hearing aids cannot work for everyone. Those who can be helped need to be carefully fitted in order to gain the enhanced hearing functionality.
The two basic types of hearing loss are conductive and sensorineural. Conductive hearing loss involves the outer and middle ear. It can result from a blockage of wax, a punctured eardrum, birth defects, ear infections, or heredity. Usually, conductive hearing loss can be corrected medically or surgically. Sensorineural, or “nerve” hearing loss involves damage to the inner ear. It can be caused by aging, prenatal and birth-related problems, viral and bacterial infections, heredity, trauma [such as a severe blow to the head], exposure to loud noises, the use of certain drugs, fluid buildup in the inner ear, or a benign tumor in the inner ear. Only in rare cases can sensorineural hearing loss be medically or surgically corrected. It is this type of hearing loss that is most commonly managed with a hearing aid.
A hearing aid is an electronic device that picks up sound waves with a tiny microphone. The microphone makes weaker sounds louder and sends them to the ear through a tiny speaker. Because a hearing aid is an amplification device, a person must have some hearing to benefit from its use. Hearing aids usually need to be custom-made to fit each individual patient. Because hearing loss has a variety of patterns and degrees of severity and affects people in different ways, no single hearing aid is right for everyone.
Each person has a different shape to his or her ear. The two part of the external ear are the auricle (the projecting part of the ear, consisting of skin overlaying an irregularly shaped elastic fibrocartilage), and the external acoustic meatus (the ear canal coursing through the auricle to the tympanic membrane).
The geometric shape of the ear impression suggests that a 3D ear camera should have the capability to acquire images that cover both the auricle and the canal areas that can be used in place of a physical impression. Image registration techniques are also needed to put multiple un-calibrated images together seamlessly to generate complete digital ear impression that can replace the traditional physical impression.
There are five common models for hearing aids in the market: behind-the-ear (BTE); in-the-ear (ITE); in-the-canal (ITC); mini-canal (MC); and completely-in-the-canal (CIC). Each requires a different shell design and fabrication technique. As indicated, however, the shell of each hearing aid needs to be custom-made to fit the shape of the individual's ear. The custom-fit is crucial to patients for ensuring both the comfort and operability of the hearing aid.
The geometric fit of the hear aid is defined as the close matching between the shape of an external ear and/or ear canal and the shell of the hearing aid. Poorly fitted shell shape creates serious problems in the performance and comfort of the hearing aid. For example, the user cannot attach the device on his or her ear. If the shape of the hearing aid shell does not match closely with its user's ear shape, there would be not enough frictional force to retain the device in its proper position. This may also create discomfort or pain. A close fit creates a comfortable wearing experience.
Additionally, the close fit of the hearing aid shell is not only for comfortable wearing, but also for ensuring the hearing aid's functional performance. If the shell cannot seal the internal ear chamber from the outside, interference between inner and outer air vibration will occur and the performance of the hearing aid will be severely deteriorated.
The current manufacturing process for custom-fit shells of hearing aids is a highly labor-intensive and manual process. Moreover, the quality control of the fitting/performance of hearing aids is difficult. The custom-fitting process starts with taking an ear impression of a patient at the office of an audiologist or hearing aid dispenser. The impression is taken with a soft material that can conform to and then represent the shape of the patient's ear and/or canal.
The impression is then shipped to manufacturer's laboratory. Each shell has to be custom-made by skilled technicians using manual techniques. The quality and consistency of the fitting vary significantly with the technician's skill level. A typical process of producing a shell takes about 40 minutes from start to finish. Once the shell is finished, electrics will be installed, calibrated, and a quality check will typically be performed. The hearing aid device will then be sent back to the audiologist or dispenser who will try the device on the patient. If for any reason the hearing aid device does not fit well (only 75% fit well in the first installation), this lengthy, costly, and uncomfortable process will have to be repeated.
Conventional hearing aid fabrication procedures suffer major problems from audiologists' viewpoint (1) the process of taking an ear impression is very uncomfortable to most patients; (2) the impression procedure exerts force to on ear structures that may cause deformation of those structures which affects the measurement accuracy, thus affects the quality of fitting; (3) it is very difficult for the audiologist to obtain an immediate feedback on the quality of the impression, until the second appointment; (4) physical handling of impressions requires a sterilization process (to eliminate contamination); and (5) an ear impression only records 3D ear shapes in a solid mold.
The manual fabrication process also suffers major drawbacks from hearing aid manufacturers' viewpoint: (1) the manual and time-consuming fabrication process is not scalable for mass production; (2) the mailing of physical impressions from dispensers to manufacturers takes several days to deliver; (3) there is a lack of consistency of quality, resulting in a high level of products that have to be remade or are returned; and (4) there is a shortage of skilled worker and a long training time to train a new worker.
To address these issues, any of the 3D cameras described herein can be used as a 3D ear camera that enables audiologists to acquire single or multiple 3D images of the external ear (auricle) and ear canal and to produce a complete 3D digital ear model that serves as a “digital ear impression.” The digital ear impression data can then be sent instantly to a manufacture's lab via the Internet, for example, thereby reducing dramatically the delivery time.
The digital impression enables the hearing aid manufacturers to take advantages of the latest breakthroughs of computer-aided-design (CAD) and computer aided manufacturing (CAM) technologies, including rapid prototyping, to produce a customized hearing aid device within, for example, a one-day time frame. After the 3D model of a custom-fit hearing aid shell is obtained, the data will be converted into the format that can be easily input to a CAD software package. The post processing and shell design will be performed. There are commercial CAD/CAM software packages that can perform such a design job, such as SolidWorks and AutoCAD. Special software also is offered by a company such as Raindrop Geomagic, NC.
The basic design steps are as follows: (1) obtain 3D point data from the 3D camera; (2) form the geometric surface; (3) cut the base; (4) trim the canal tip; (5) round the tip; (6) thicken the shell; (7) create the vent; and (8) create the hole. The completed design of the custom-fit shell is then sent to the rapid prototype (RP) machine for fabrication. The technology of RP production is quite mature and easily accessible. For example, using the RP machine from the leading provider 3D Systems, the shell can be built directly using the biocompatible materials.
Even including the quality assurance, electronics calibration, and shipping of the hearing-aid device, the entire process of making a custom-fit hearing aid device would be shortened from weeks to a few days. More importantly, the digital impression technology improves the quality of fit, thus enhance the hearing functionality for impaired people. This process is illustrated in FIG. 24.
Within the three major components, CAD software has been an active R&D area in the past decades, and commercial software packages and visualization tools (SolidWorks, AutoCAD, eShell, etc) are currently available. A number of CAM fabrication systems (such as 3D Systems SLA, Stratisys) are currently in the commercial market, thanks to the rapid technology advances and commercialization in the Rapid Prototyping industry. This specification provides the handheld 3D ear camera for taking the “Digital Ear Impression”.
On exemplary embodiment for the proposed 3D Ear Camera is illustrated in FIG. 25. Inside an ergonomically styled case (254) is a specially designed 3D Camera. To keep a compact dimension, a miniature color CCD image sensor (251) is used to collect both 2D and 3D images of the ear. A set of optical lenses and a reflecting mirror (252) are used to direct the light ray from a window (253) at the tip of the casing (254) to the CCD camera (251). The narrow neck of the casing (254) allows an easy access inside a patient's ear and the convenience of taking 3D measurement of a complex surface profile. The multi-rainbow light projection is generated through an optical fiber bundle (255). Detailed optics design will be performed to ensure a reasonable length of baseline.
There is a second independent light source (256) in the 3D ear camera. By turning off the rainbow source (255) and turning on the ordinary diverging white light source (256), the camera can take 2D images with accurate color representation of ear. This enables the ear camera to output a “textured 3D digital model” of the ear. This 3D modeling capability can be extended to building a comprehensive 3D database of ear structures for interactive virtual reality (VR) surgical simulation applications. As indicated above, the 3D ear model can be transferred over the Internet. Therefore the powerful 3D ear digitization capability provides a needed tool for ear surgery planning, tele-consultation, and education/training of audiologists/surgeons. Adding the second light source feature also allows the camera to also be used as a standard 2D camera for audiologists, enhancing its marketability.
Since the 3D ear camera is a handheld device, it is therefore unrealistic to expect that the multiple images acquired by a user contain any accurate geometric calibration information with respect to each other. Thus, the 3D ear camera system includes a set of robust algorithms and software able to automatically register multiple 3D images of the same ear acquired by the 3D ear camera to produce a complete 3D ear model. There are a number of commercial software packages available that will provide this functionality, such as the Videobrush software developed by Sarnoff.
Seamless registration of multiple un-calibrated 3D images is among the most difficult problems in the field of advanced 3D surface imaging technology, yet it is a critical and indispensable step leading to practical applications of the 3D camera. Conventional methods of 3D image integration rely on the pre-calibrated camera positions to align multiple 3D images, and require extensive ad hoc manual routines to merge the aligned 3D images into a complete 3D model.
Ideally, intelligent 3D image registration software could register a number of 3D images without the need of any prior knowledge of the cameras used to generate the images or their placement. Such a system works under the premise that the problem of integrating multiple 3D images can be considered as a sequence of 3D registration problems dealing with only a pair of 3D images. For example, one can start registering a pair of 3D images (say I1 and I2) within a given set of N frames of 3D images. After completing the integration of I1 and I2, the obtained 3D image becomes a new I1 image that is ready for registering with the I3 image. This process continues until all N images are integrated into a complete 3D ear model.
After a pair of 3D images is selected, the system will perform certain pre-processing for the 3D images to ensure their quality. Pre-processing operations include 3D image filtration, “bad 3D data” elimination, and “cutting off” the unreliable or undesired areas. The pre-processed 3D image pair is then sent to the image alignment module. Alignment of a pair 3D images is based solely upon the 3D surface features (such as surface curvature) that are independent from any coordinate system definition or illumination conditions. Once the corresponding points are found in the overlapping areas between two images, a 4×4 homogenous coordinate transformation is performed to align one image with the other in a global coordinate system.
The next step is to merge (i.e., blend) two aligned 3D images to form a uniformed 3D data set. Two images are usually not in the same spatial resolution, and the sampling of 3D surface data is not in the same grid orientation. The problem is exacerbated by noise in the data and errors in the alignment. A “resembling” procedure based on the Fuzzy Logic principle generates a uniformed 3D surface and blends the boundary areas. A 3D image merged from two “raw” images then becomes a new 3D image ready for the next integration process.
The process of registering pairs of 3D images continues by selecting the next “best” 3D image to integrate. The new image should overlap an area in the existing 3D image. This process continues until all the “raw” 3D images are processed and merged.
The 3D registration does not stop here. After getting a complete 3D model from multiple 3D images, the system enters into the “post-processing” stage to further enhance quality of 3D modeling. The 3D model will be compressed to the detail level specified by the user. Miscellaneous editing functions of 3D model will be necessary to fill holes and perform final touch to the 3D model. The 2D texture and color overlay will also be performed as an option to give the 3D model a more realistic look. Finally, a 3D visualization tool will be available that allows users to assess the 3D registration results and extract useful parameters from the completed 3D ear model.
For a more detailed explanation of the integration of a plurality of 3D images, please refer to Applicant's U.S. patent application Ser. No. 09/617,867, filed Jul. 17, 2000 and which is incorporated herein by reference in its entirety.
In summary, the proposed 3D ear camera for custom-fit hearing aid will provide a revolutionary advance in the hearing aid industry. The implementation of the 3D ear camera would lead to quality enhancement of care to the hearing impairment population, customer satisfaction and lower return rate for hearing aid manufacturers. The novel 3D ear camera technology disclosed herein fills in the gap of the digital audiology paradigm.
The preceding description has been presented only to illustrate and describe embodiments of invention. It is not intended to be exhaustive or to limit the invention to any precise form disclosed. Many modifications and variations are possible in light of the above teaching. It is intended that the scope of the invention be defined by the following claims.
This application claims the benefit under 35 U.S.C. §119(e) of prior U.S. provisional patent application No. 60/375,934, filed Apr. 26, 2002, the disclosure of which is incorporated herein by reference in its entirety. This application is also a continuation-in-part of previous U.S. patent application Ser. No. 09/771,531, filed Jan. 29, 2001, the disclosure of which is incorporated herein by reference in its entirety. U.S. patent application Ser. No. 09/771,531 claims the benefit of U.S. Provisional Appln. No. 60/178,695, filed Jan. 28, 2000, the disclosure of which is incorporated herein by reference in its entirety.
Number | Name | Date | Kind |
---|---|---|---|
4088400 | Assouline et al. | May 1978 | A |
6147760 | Geng | Nov 2000 | A |
6341016 | Malione | Jan 2002 | B1 |
6356298 | Abe et al. | Mar 2002 | B1 |
6552754 | Song et al. | Apr 2003 | B1 |
6636310 | Ben-Dov et al. | Oct 2003 | B1 |
6700669 | Geng | Mar 2004 | B1 |
20030002052 | Hoffmann | Jan 2003 | A1 |
20030016366 | Takeda et al. | Jan 2003 | A1 |
20030164952 | Deichmann et al. | Sep 2003 | A1 |
Number | Date | Country | |
---|---|---|---|
20030223083 A1 | Dec 2003 | US |
Number | Date | Country | |
---|---|---|---|
60375934 | Apr 2002 | US | |
60178695 | Jan 2000 | US |
Number | Date | Country | |
---|---|---|---|
Parent | 09771531 | Jan 2001 | US |
Child | 10426083 | US |