Minori Noguchi

Real-time focus range sensor

Structures of dynamic scenes can only be recovered using a real-time range sensor. Depth-from-defocus offers a direct solution to fast and dense range estimation. It is computationally efficient as it circumvents the correspondence problem faced by stereo and feature tracking in structure-from-motion. However, accurate depth estimation requires theoretical and practical solutions to a variety of problems including the recovery of textureless surfaces, precise blur estimation, and magnification variations caused by defocusing. Both textured and textureless surfaces are recovered using an illumination pattern that is projected via the same optical path used to acquire images. The illumination pattern is optimized to ensure maximum accuracy and spatial resolution in the computed depth. The relative blurring in two images is computed using a narrow-band linear operator that is designed by considering all the optical, sensing and computational elements of the depth-from-defocus system. Defocus-invariant magnification is achieved by the use of an additional aperture in the imaging optics. A prototype focus range sensor has been developed that produces up to 512/spl times/480 depth estimates at 30 Hz with an accuracy better than 0.3%. Several experimental results are included to demonstrate the performance of the sensor.<<ETX>>

Microscopic shape from focus using active illumination

Shape from focus relies on surface texture for the computation of depth. In many real-world applications, surfaces can be smoothly shaded and lacking in detectable texture. In such cases, shape from focus generates inaccurate and sparse depth maps. This paper presents a novel extension to the original shape from focus method. A strong texture is forced on imaged surfaces by the use of active illumination. The exact pattern of the projected illumination is determined through a careful Fourier analysis of all the optical effects involved in focus analysis. When the focus operator used is a 2D Laplacian, the optimal illumination pattern is found to be a checkerboard whose pitch is the same size as the distance between adjacent elements in the discrete Laplacian kernel. This analysis also reveals the exact number of images required for accurate shape recovery. These results are experimentally verified using an optical microscope. Surfaces lacking in texture, such as, three-dimensional structures on silicon substrates and solder joints on circuit boards were used in the experiments. The results show that the derived illumination pattern is in fact optimal, facilitating accurate shape recovery of complex and pertinent industrial samples.

Real-time computation of depth from defocus

A new range sensing method based on depth from defocus is described. It uses illumination pattern projection to give texture to the object surface. Then the image of the scene is split into two images with different focus settings and sensed simultaneously. The contrast map of the two images are computed and compared pixel by pixel to produce a dense depth map. The illumination pattern and the focus operator to extract the contrast map are designed to achieve finest spatial resolution of the computed depth map and to maximize response of the focus operator. As the algorithm uses only local operations such as convolution and lookup table, the depth map can be computed rapidly on a data-flow image processing hardware. As this projects an illumination pattern and detects the two images with different focus setting from exactly the same direction, it does not share the problem of shadowing and occlusion with triangulation based method and stereo. Its speed and accuracy are demonstrated using a prototype system. The prototype generates 512 by 480 range maps at 30 frame/sec with a depth resolution of 0.3% relative to the object distance. The proposed sensor is composed of off-the-shelf components and outperforms commercial range sensors through its ability to produce complete three-dimensional shape information at video rate.

Fine Particle Inspection Down to 38 nm on Bare Wafer with Micro Roughness by Side-Scattering Light Detection

A system for detecting fine particles on LSI wafers is proposed. This system illuminates the wafer with a laser at a large angle of incidence, and uses a detector with a small pixel size. In this way, optical noise from the bare wafer is reduced, enabling the detection of standard 38-nm particles. Optical noise from a wafer surface is analyzed and shown to be caused by diffracted light. Simulation with a diffraction model indicates that the optical noise level is related to the wafer roughness. To detect fine particles with a high probability of 95%, a high illumination energy of 180 J is needed. With small pixels, a low illumination energy of 1.2 J will also detect small 50-nm particles, but with a low probability of 10%. This technique will be useful in monitoring the cleanliness of submicron LSI production.

Focus Range Sensors

Structures of dynamic scenes can only be recovered using a real-time range sensor. Focus analysis offers a direct solution to fast and dense range estimation. It is computational efficient as it circumvents the correspondence problem faced by stereo and feature tracking in structure from motion. However, accurate depth estimation requires theoretical and practical solutions to a variety of problems including recovery of textureless surfaces, precise blur estimation, and magnification variations caused by defocusing. Both textured and textureless surfaces are recovered using an illumination pattern that is projected via the same optical path used to acquire images. The illumination pattern is optimized to ensure maximum accuracy and spatial resolution in computed depth. A prototype focus range sensor has been developed that produces up to 512×480 depth estimates at 30 Hz with an accuracy better than 0.3%. In addition, a microscopic shape from focus sensor is described that uses the derived illumination pattern and a sequence of images to recover depth with an accuracy of 1 micron. Several experimental results are included to demonstrate the performances of both sensors. We conclude with a brief summary of our recent results on passive focus analysis.

Resolution enhancement of stepper by complementary conjugate spatial filter

An ultra-high resolution stepper is proposed, which combines annular illumination and a complementary conjugate spatial filter. This technique reproduces deep submicron complex non-repeated patterns by relatively emphasizing the diffracted light from high spatial-frequency patterns. Images of several patterns are simulated for a conventional stepper, a stepper with phase shifting, and for the proposed technique. The results indicate that this technique has the potential to pattern 0.3-|im images using an i-line stepper with 0.5 N.A. with conventional masks. It should enable stepper ranges to be extended to the next generation of DRAMs.

Defect inspection on CMP process and its application

A high-throughput high-sensitivity defect-detection technique has been developed for manufacturing 0.15-0.25- micrometers LSI devices. It incorporates a high-resolution detection systems using multi-channel detectors and a high- resolution imaging system using spatial filtering and collimated focused-beam illumination. A new algorithm called correlated local area statistical threshold enables this technique to achieve a sensitivity of 0.15 micrometers on front- end processes and 0.3 micrometers on back-end processes and a high throughput.

40-nm-particle high-probability detection for bare wafer using side-scattered light

A high probability particle detection system for LSI wafers is proposed. In order to detect fine particles on bare wafers, optical noise from them are studied. Simulation with a diffraction model indicates that the optical noise is caused by diffracted light on the wafer roughness, and can be reduced by a large incident angle illumination and a detector with small pixel size. The side-scattering light detection system which has a illumination of incident angle of 80 degrees and a detector with a pixel size of 0.3 micrometers was confirmed experimentally to detect standard 38-nm particles in high signal-to-noise ratio, and the detection results were verified by SEM. Probability study of detection indicates that the detection probability reduces rapidly as detection light level is low. Two types of systems are proposed, a high detection probability system of 95% with energy of 600 mJ/cm2 a low illumination energy system of 4 mJ/cm2 with probability of 10%.