The University of Kitakyushu Masahiro Okuda Lab.｜北九州市立大学 奥田正浩研究室

Medical Image Restoration by Deep Neural Network深層学習を用いた医歯用画像の復元～北九州医療ICT基盤の構築～

We conduct research on medical and dental image analysis using signal processing and artificial intelligence jointly with the University of Occupational and Environmental Health and Kyushu Dental University. We form Kitakyushu medical ICT infrastructure through this research project.
信号処理，人工知能を用いた医科・歯科画像解析の研究を産業医科大学，九州歯科大学と共同で行う．本研究プロジェクトを通して北九州医療ICT基盤を形成する．

Submillimeter Endoscope Imaging with Sparse Modelingスパースモデリングを用いたサブミリメータ内視鏡イメージング

Ministry of Internal Affairs and Communications, Strategic Information and Communications R&D Promotion Programme (Principal Researcher: Masahiro Okuda)

High Dynamic Range Imaging高ダイナミックレンジ（HDR）画像処理

We propose an image restoration technique that uses multiple image integration. The detail of the dark area when acquiring a dark scene is often deteriorated by sensor noise. Simple image integration inherently has the capability of reducing random noises, but it is especially insufficient in scenes that have a dark area. We introduce a novel image integration technique that optimizes the weights for the integration.

高ダイナミックレンジ画像とは

Sparse Modeling for Multi-channel Spectral Restorationマルチチャネルスペクトル復元のためのスパースモデリング

We propose a method for local spectral component decomposition based on the line feature of local distribution. Our aim is to reduce noise on multi-channel images by exploiting the linear correlation in the spectral domain of a local region.

分光スペクトル情報からハイパースペクトルセンサにより観測したソーンの物質を特定する技術、高画質化する技術に取り組んでいます。

Compressive sensing and Image restoration圧縮センシングと画像復元

We propose a new regularization function, named as Arranged Structure tensor Total Variation (ASTV), for multichannel image restoration. ASTV is based on an arranged structure tensor that becomes an approximately low-rank matrix when a multichannel image of interest has strong correlation among their channels. This observation suggests that penalizing the nuclear norm of the arranged structure tensor is a reasonable regularization for multichannel images, leading to the definition of ASTV.

少ない観測サンプルから原画像を推測する圧縮センシングと、スパースモデリングを用いた画像復元に取り組んでいます。

Misaligned Image Integration位置ズレ画像の統合

We present a new image integration technique for flash and long-exposure image pairs to capture a dark scene without incurring blurring or noisy artifacts. We formulate image integration as a convex optimization problem with the local linear model. The proposed method makes it possible to integrate the color of the long-exposure image with the detail of the flash image without causing any harmful effects to its contrast, where the images do not need perfect alignment by virtue of our new integration principle.

連写した複数の画像を統合することでより高精細な画像を得る技術は重要である。ここでは位置ズレのある複数画像を一枚に統合する手法を提案している。

Reflectance Estimation using convex optimization凸最適化を用いた反射率推定

Using a flash/no-flash image pair, we propose a novel white-balancing technique that can effectively correct the color balance of a complex scene under multiple light sources. In the proposed method, by using multiple images of the same scene taken under different light- ing conditions, we estimate the reflectance component of the scene and the multiple shading components of each image. The reflectance compo- nent is a specific object color which does not depend on scene illumination and the shading component is a shading effect caused by the illumination lights. Then, we achieve white balancing by appropriately correcting the estimated shading components.

Reference-based Image Filtering参照画像を用いたフィルタリング

We introduce a novel edge-aware filter that manipulates the local covariances of a color image. A covariance matrix obtained at each pixel is decomposed by the singular value decomposition (SVD), then diagonal eigenvalues are filtered by characteristic control functions. Our filter form generalizes a wide class of edge-aware filters. Once the SVDs are calculated, users can control the filter characteristic graphically by modifying the curve of the characteristic control functions, just like tone curve manipulation while seeing a result in real-time. We also introduce an efficient iterative calculation of the pixel-wise SVD which is able to significantly reduce its execution time.

3D imaging3次元画像処理

3D mesh parameterization is a method which converts the complicated 3D mesh into the flat and non-overlapped 2D mesh, and is used for "texture-mapping" to make the correspondence between a texture-image and a 3D mesh in 2D space, and "remeshing" to convert irregular meshes into more manageable meshes. In this paper, we propose a 3D mesh parameterization method which is able to express more detailed shape of the 3D model. However, this one-sided emphasis on "remeshing" incurs texture-distortions in practice. So, we also propose a texture-mapping method which uses a transform-map of texture-coordinates to keep "texture-mapping" qualities.

ギターサウンドモデリング

入力音声をもとに真空管の非線形特性、トーンスタック、各種エフェクトなどのモデルを自動的に構築するアルゴリズムを提案します。この問題は非凸最適化問題に帰着されるが、ここでは進化論的計算手法により各ステップのパラメータを同時に求め、所望のサウンドを表現することを目的としている。

Sparse Filters

In this paper, we present a numerical algorithm for the design of FIR filters. Our method minimizes the number of nonzero entries in the impulse response together with the least squares error of its frequency response. We show that the FIR filters with sparse coefficients can outperform a conventional least suares approach and the Parks-McCllelan method under the condition of the same number of multipliers.