Kunio Nobori

Spatio-temporal clustering of probabilistic region trajectories

We propose a novel model for the spatio-temporal clustering of trajectories based on motion, which applies to challenging street-view video sequences of pedestrians captured by a mobile camera. A key contribution of our work is the introduction of novel probabilistic region trajectories, motivated by the non-repeatability of segmentation of frames in a video sequence. Hierarchical image segments are obtained by using a state-of-the-art hierarchical segmentation algorithm, and connected from adjacent frames in a directed acyclic graph. The region trajectories and measures of confidence are extracted from this graph using a dynamic programming-based optimisation. Our second main contribution is a Bayesian framework with a twofold goal: to learn the optimal, in a maximum likelihood sense, Random Forests classifier of motion patterns based on video features, and construct a unique graph from region trajectories of different frames, lengths and hierarchical levels. Finally, we demonstrate the use of Isomap for effective spatio-temporal clustering of the region trajectories of pedestrians. We support our claims with experimental results on new and existing challenging video sequences.

Image Synthesis System Using 3D Model-based Coding - Simulates Facial Expressions And Aging

The Algorithm and the System The model-based coding technique uses some knowledge of facial structure to recognize and to synthesize images. First, fitting algorithm uses a polygonal wireframe model and input source image to obtain a 3D face model. This 3D face model uses rules to obtain synthesized 3D face model. Then, a 2D source image is mapped onto the new synthe- sized 3D face model by texture-mapping of the composite of the image. In this way, a 3D computerized face is created We discuss a new image synthesis system which can gen- erate facial expressions and aging effects on facial images for a personal computer (PC). This system has wide applica- tions in many areas such as simulation in beauty parlors and cosmetic surgery clinics and montage systems for criminal investigation.

Video generation for high spatio-temporal resolution imaging

We propose a color video generation method for spatio-temporal high-resolution video imaging in dark conditions. The proposed method consists of two steps. First, RGB-separated video sequences with different spatio-temporal resolution sets are captured to increase the amount of captured light. Second, a high spatio-temporal resolution color video is reconstructed from those input video sequences in the regularization framework. We show the advantages of our method using a prototype camera system and simulation results.

Photometric Linearization under Near Point Light Sources

We present a method for classifying image pixels of real images into multiple photometric factors: specular reflection, diffuse reflection, attached shadows and cast shadows. Conventional photometric linearization methods cannot correctly classify pixels under near point light sources, since they assume parallel light. To satisfy this assumption, our method utilizes a photometric linearization method that divides images into small regions. It also propagates linearization coefficients from neighboring regions. Our experimental results show that the proposed method can correctly classify image pixels into photometric factors, even if images are obtained under near point light sources.

Compressive sensing reconstruction using collaborative sparsity among color channels

This study describes a reconstruction method for compressive sensing using collaborative sparsity among multi-frame images and color channels. The proposed method reduces the artifact for compressive sensing and obtains better image quality. Experimental results reveal that the proposed method reconstructs 6.1 dB higher quality images than the conventional one for complex texture at occlusion boundaries.

Compressive color sensing using random complementary color filter array

We propose a new color imaging system based on a compressive sensing technique. Our system consists of a random complementary color filter array (CFA) for random projection and a color reconstruction method for demosaicing. Our CFA overlaps two complementary color filters and consists of six color filters: cyan (C), yellow (Y), magenta (M), C+Y, C+M, and Y+M. By arranging these six color filters randomly, our imaging system achieves pseudo random projection among red (R) /green (G) / blue (B) colors, which is the key technology of compressive sensing. Because this CFA can retain more color information than RGB CFA, the proposed color reconstruction method reduces artifacts at monochromatic edges and in high-frequency regions, and obtains better image quality. As an additional contribution, we introduce saturation consistency to suppress color artifacts in saturated areas, then achieve to 3.3 dB higher quality images than the conventional method.

A surround view image generation method with low distortion for vehicle camera systems using a composite projection

This paper proposes a surround view image generation method for vehicle camera systems. To assist the driver during parking, a view with easy comprehension of distance and direction between the vehicle and objects is desirable. However, the conventional method of using an equidistant projection for generating a surround image of wide field of view causes image distortion, with straight lines appearing curved. This prevents the driver from correctly understanding the distance and direction of objects. Our proposed method uses a composite projection that combines two projection models: perspective projection and equidistant projection. This strategy can generate an image without distortion by using perspective projection near the vehicle and provides a wide field of view using equidistant projection. The experiments demonstrate the generation from parking scene images, using our proposed method, of a surround image with a wide field of view and no distortion near the vehicle.

A model-based facial image recognition method for image synthesizing system

This paper reports a new facial feature point extraction method for an image synthesizing system. The method works successfully with images of more than 1000 Japanese and 500 Americans. A sample of application is reported.