High-Resolution Capture in Dashcam Systems

Video enhancement method to improve the quality of low-resolution, low-quality videos. The method involves enhancing specific regions of interest (ROI) using a higher complexity neural network while using a lower complexity network for the rest of the frame. This allows focusing computing resources on key areas. The enhanced ROI branches are fused with the full frame branch after temporal smoothing to obtain the final enhanced video. It enables better visual effects compared to uniformly enhancing the whole frame.

Source

Video processing system that selectively enhances image quality of areas containing important objects while compressing other areas to reduce video size. It uses temporal changes in recognition results to identify objects that need closer focus. The system sharpens image quality of gaze areas containing extracted gaze targets while compressing other areas. This allows selective enhancement of critical objects without over-compression.

Source

Road monitoring video defogging method that improves clarity of foggy video to enhance visibility and enable tasks like vehicle monitoring and license plate recognition. The method involves dividing each frame into sky and non-sky areas, calculating transmittance separately, and combining them to avoid distortion. It also uses guided filtering to keep edges smooth and accurate after defogging. This provides more realistic fog removal compared to prior methods.

Source

Vehicle vision system with cameras that use a thermoelectric cooler to prevent image degradation at high temperatures. The cameras have a sealed housing with a thermoelectric cooler that draws heat away from the camera circuitry which reduces image sensor temperature. A heat transfer element connects the cooler to the circuit board containing the image sensor. The cooler is powered and responds to a temperature sensor to maintain optimal operating temperature.

Source

Image processing system and method for automotive electronic rearview mirrors that improves image quality and perception for drivers by enhancing images captured by vehicle-mounted cameras. The processing steps include synchronizing camera and inertial data, stabilizing and correcting the input image, classifying high-brightness images, enhancing image quality, and assessing image quality. The system uses techniques like electronic image stabilization, distortion correction, fast image classification, image enhancement, and quality assessment to process the input images and output high-quality images for the electronic rearview mirror.

Source

Enhancing image quality from vehicle cameras to improve reliability and safety of autonomous driving. It uses a GAN model trained with images from other vehicles to enhance images from the target vehicle's camera. When the target vehicle captures a low-quality image, it uses the trained GAN to enhance it.

Source

Vehicle body structure with a camera that captures images around the vehicle, wherein a moth-eye structure is formed on the window to suppress image degradation due to internal reflections. The moth-eye structure is formed on the inner surface of the window within the camera's field of view. This reduces reflections from multiple surfaces that can create false images and degrade captured images. Attaching a selectively positioned film with the moth-eye structure to the window focuses the anti-reflection benefits within the camera's view while leaving other areas unobstructed. A support member fixed to the window surrounds the camera's field of view.

Source

Using a vehicle camera for improved parking assistance. The camera has both a wide-angle lens for parking space recognition and an integrated depth camera for obstacle detection. This allows detecting obstacles more accurately in 3D compared to just using a regular camera. By combining wide-angle and depth imaging in one camera, it provides enhanced parking assistance capability over just using traditional cameras.

Source

Vehicle camera system that provides high quality images while shielding direct sunlight and glare from the lens. The camera has a hood that limits the incident angle of light to the lens. It also captures two images, a high-resolution one for the rear view monitor and a smaller one for other displays. The hood shields light outside the smaller image area. This allows shielding without protruding beyond vehicle limits. The camera corrects distortion based on detected image height changes due to temperature.

Source

Method for displaying a realistic 3D model of a vehicle's surrounding area by harmonizing color information from multiple wide-angle camera views. The method uses a lattice structure to assign pixels from overlapping camera images. Color differences at lattice points are evaluated and weighted to determine a global color transformation. This transformation is then applied to adapt one camera view's colors to match the other view, providing a consistent appearance. The adapted images are used to create the 3D model with realistic color and brightness across views.

Source

Real-time image enhancement for low-quality video captured by cameras in conditions like low light and backlight, using an FPGA implementation of a biological vision model. The method involves processing the video using mechanisms inspired by the human retina, such as receptive field modeling and median filtering. It aims to improve video quality in challenging lighting conditions in real-time on embedded devices.

Source

Video processing system with automatic analysis and processing of video images. The system has a video collector, video analysis module, DSP processor, FPGA chip, video memory, encoding/decoding module, and video transmission unit. The video collector gathers footage, the analysis module enhances quality, DSP processes further, FPGA compresses, memory stores, encoding converts, and transmission sends the improved video. The system automatically analyzes and improves video sharpness, noise, color, etc.

Source

Smart roadside unit with high and low sensitivity cameras for improved vehicle detection accuracy. The roadside unit has two cameras with different brightness sensitivities but the same field of view. One camera captures a clear image under bright conditions and the other captures a clear image under dark conditions. The roadside unit extracts vehicle information from both images to improve accuracy in scenes with large lighting differences like tunnels.

Source

In-vehicle camera design that reduces electromagnetic noise generation without sacrificing signal quality. The camera has a conductive housing to prevent noise emitting into the vehicle. A spacer separates the flexible cable from the housing to prevent electrical contact and current flow that causes noise. The spacer lacks electromagnetic properties to avoid absorbing or reflecting signals.

Source

Device, system, and method to capture non-deformed images from a moving object like a vehicle using a line scan camera. It measures the object's speed and sets the line frequency of the camera to match, preventing image stretch or compression. A control module triggers camera capture at the matching line frequency.

Source

System and method to automatically mitigate image distortion in autonomous vehicles caused by vibrations. The system uses vehicle state data, inertial measurements, and strain data from the image capture assembly to predict and correct image distortion. This allows stabilizing and enhancing the image quality from cameras on autonomous vehicles to improve perception algorithms for environment sensing and decision making.

Source

Compact camera module for vehicle windshield mounting that captures a wide angle view. The camera module includes a wide angle lens and a hood to block light outside the imaging target range. The hood design allows imaging the external environment while preventing bright light interference. The hood has an open top and side walls that extend from around the lens towards the windshield. The height of the side walls is set to avoid blocking the full angle of view.

Source

Camera assembly structure for a vehicle camera module with improved optical alignment and reliability compared to adhesive bonding. The structure uses solder-jet bonding to fix the lens and image sensor in accurate alignment after pre-assembly. The housing posts and board holes have precise clearance for solder to fill. This replaces adhesives prone to thermal deformation. The solder-jet provides fast, localized, and temperature-controlled bonding without curing issues.

Source

Reducing the resolution and distance limitations of camera modules in vehicles to allow accurate long-range object recognition. The camera module has two cameras - one wide-angle and one narrow-angle. The narrow-angle camera with the same pixel count as the wide-angle camera can recognize distant objects more clearly. The wide-angle camera captures the overall scene. The dual camera setup provides both context and detail for accurate long-range object recognition.

Source

Windshield-mounted camera for vehicles that reduces glare by using a coated glare shield. The camera is mounted behind the windshield and captures images of the road. A light-absorbing coating is applied to a smooth surface below the camera lens. This reduces reflection of glare onto the lens. The camera captures clearer images without stray light interference.

Source