Computer vision focuses on the automated extraction, analysis, and interpretation of visual information from images and videos. This discipline integrates image processing, pattern recognition, and machine learning to address tasks such as object detection, segmentation, recognition, and 3D reconstruction. Research explores advancements in convolutional neural networks, optical flow, depth estimation, and scene understanding, enabling applications across robotics, surveillance, medical imaging, and autonomous systems. The following collection presents recent scholarly publications and patented 기술 that advance methodologies and applications within computer vision:
본 자료는 컴퓨터 비전 분야의 최신 영문 논문 및 특허를 엄선하여 정리한 것으로, 다양한 온라인 과학 저널에 게재된 자료들을 이미지 분할, 객체 탐지, 특징 추출, 합성곱 신경망, 이미지 분류, 광류, 스테레오 비전, 깊이 추정, 이미지 인식, 장면 이해, 이미지 재구성, 이미지 향상, 패턴 인식, 얼굴 인식, 모션 트래킹, 3D 재구성, 이미지 정합, 의미론적 분할, 인스턴스 분할, 비주얼 SLAM, 이미지 스티칭, 텍스처 분석, 에지 검출, 객체 추적, 초해상도, 이미지 노이즈 제거, 비디오 분석, 이상 탐지 및 이미지 캡션 등의 주제별로 분류하고 있습니다.
Product assessment system for e-commerce
Patent published on the 2026-06-11 in WO under Ref WO2026118348 by SHANDONG HUAYU UNIV OF TECHNOLOGY [CN] (Fan Huawei [cn])
Abstract: A product assessment system for e-commerce. The system comprises: a rating data collection module, which is used for receiving multi-dimensional ratings from a plurality of users on a plurality of products, and providing raw rating data for correction processing; a user bias correction module, which is used for modeling and correcting individual biases in user ratings, wherein the rating data collection module is connected to the user bias correction module, and outputs corrected rating data; a [...]
Our summary: The system collects multi-dimensional ratings from users for products. It corrects individual biases in ratings and performs dimensionality reduction using principal component analysis. Finally, it enhances feature expression through nonlinear mapping and aligns recommendations with user group characteristics.
product assessment, e-commerce, user bias correction, feature extraction
Patent
Image analysis apparatus and image analysis method
Patent published on the 2026-06-11 in WO under Ref WO2026122394 by SONY INTERACTIVE ENTERTAINMENT INC [JP] (Sugiyama Shotaro [jp])
Abstract: Image analysis apparatus and method are disclosed. An example image analysis apparatus includes one or more processors which acquire a first type optical flow that represents the distribution of bidirectional motion vectors between existing frames of a video image. The search range corresponding to the target pixel set in the interpolated frame plane is set in the existing preceding and following frame planes. Among the pairs of motion vectors composed of a motion vector in the forward direction[...]
Our summary: The disclosed image analysis apparatus includes processors that acquire optical flow representing bidirectional motion vectors. It sets a search range for target pixels in interpolated frame planes. The apparatus detects the most likely motion vector pair from existing frames to determine the target pixel s motion vector.
image analysis, optical flow, motion vectors, video processing
Patent
Monitoring system, monitoring method, and monitoring program
Patent published on the 2026-06-11 in WO under Ref WO2026120763 by NEC CORP [JP] (Nishida Masataka [jp])
Abstract: In the present invention, an object position calculation unit calculates the position of an object detected by a fixed-point sensor which is installed at a fixed position. A bird s-eye view creation unit creates, from an image captured with a moving camera mounted on a moving object, a bird s-eye view in which the capturing direction of the moving camera is viewed from above. An integration unit generates an integrated image which is an image obtained by integrating the detected object in a corr[...]
Our summary: The system calculates the position of an object using a fixed-point sensor. It creates a bird s-eye view from images captured by a moving camera. An integration unit generates and outputs an integrated image based on the detected object s position.
Object position calculation, bird s-eye view, fixed-point sensor, integrated image
Patent
Time-to-digital converter, analog-to-digital converter, and electronic device
Patent published on the 2026-06-11 in WO under Ref WO2026120914 by SONY SEMICONDUCTOR SOLUTIONS CORP [JP] (Nogamida Takeru [jp], Suzuki Atsuya [jp])
Abstract: The present invention improves performance in a TDC including a circuit for detecting edges. This time-to-digital converter is provided with a predetermined number of phase comparators and a predetermined number of edge detection units. Each of the phase comparators compares the respective phases of a pair of input signals having different edge phases, and outputs a comparison result. Each of the edge detection units comprises a first leading edge detection unit that detects an edge of one hav[...]
Our summary: The invention enhances the performance of a time-to-digital converter by integrating a circuit for edge detection. It includes multiple phase comparators that compare input signal phases and output results. Additionally, it features edge detection units for leading and lagging edge detection, ensuring precise measurements.
time-to-digital converter, edge detection, phase comparator, electronic device
Patent
Image recognition method and recognition system
Patent published on the 2026-06-11 in WO under Ref WO2026120852 by MURATA MACHINERY LTD [JP] (Hachiki Junya [jp])
Abstract: This image recognition method includes a preparation step, an extraction step, an additional training step, and a recognition step. In the preparation step, a base model for recognizing each item from an image of items is generated or acquired on the basis of images of items. In the extraction step, an item of a type for which the recognition rate did not satisfy a criterion as the result of recognizing each item from the image of items using the base model is extracted. In the additional traini[...]
Our summary: This method involves generating a base model for item recognition from images. It includes extracting underperforming items and performing additional training to create a LoRA layer. Finally, the base model and LoRA layer are utilized to recognize multiple item types in images.
Image recognition, model training, feature extraction, recognition system
Patent
Secure check processing system and related method
Patent published on the 2026-06-11 in US under Ref US20260162094 by IWALLET INC [US] (Kolchin Dmitriy [us])
Abstract: [0000] An AI-based method for integrating payment processing across disparate systems enables seamless transaction reconciliation and payment status updates without relying on APIs. The AI agent autonomously retrieves transaction data from third-party processors, scheduling systems, or CRMs using techniques such as web scraping, optical character recognition (OCR), and pattern recognition. The agent dynamically adapts to system workflows through reinforcement learning, performs risk assessments [...]
Our summary: The method enables seamless payment processing across different systems without APIs. It uses AI to autonomously retrieve transaction data and adapt to workflows. The system enhances operational efficiency and compliance through real-time updates and audit logs.
payment processing, AI integration, transaction reconciliation, risk assessment
Patent
Learning-based segmentation of diffusion-weighted MR images with arbitrary q -space samplings
Published on 2026-06-02 by @MIT
Abstract: AbstractSegmenting anatomical regions is a crucial step in many diffusion-weighted MRI (dMRI) workflows, such as region-of-interest analysis or anatomically-constrained tractography, which enable in vivo studies of brain microstructure and connectivity. However, convolutional neural networks (CNNs)—the foundation of most state-of-the-art segmentation models—require structured inputs with a fixed number of channels. This makes them ill-suited for dMRI, where acquisition protocols vary widely [...]
Our summary: This work presents a novel method for segmenting diffusion-weighted MRI data using geometric deep learning. It directly maps unstructured dMRI data to anatomical segmentations without requiring diffusion model fits. The proposed approach achieves robust generalization and superior performance compared to existing methods.
segmentation, diffusion-weighted MRI, geometric deep learning, convolutional neural networks
Publication
Cross-reference attention for region-aware super-resolution of blood cell images
Published on 2026-05-01 by @OXFORD
Abstract: AbstractBlood cell image analysis plays a critical role in clinical diagnostics, as white blood cells (WBCs) provide diagnostic cues for infections, cancers, and immune disorders. While low-magnification microscopy offers a wide field of view (FOV) but insufficient resolution for meaningful assessment, high-magnification microscopy offers more detailed but narrower FOV images. To obtain a high-magnification, large FOV image, current practice relies on stitching multiple high-magnification images[...]
Our summary: This study introduces a multi-reference-based super-resolution framework for blood cell microscopy. It employs a Cross-Reference Attention Module to enhance image quality and efficiency. The method outperforms traditional stitching techniques, providing large field-of-view images with improved white blood cell detail.
super-resolution, blood cell imaging, cross-reference attention, region-aware architecture
Publication
