35 results about "Depth enhancement" patented technology

Motion picture scenes to be colorized/depth enhanced (2D->3D) are broken into separate elements, backgrounds/sets or motion/onscreen-action. Background and motion elements are combined into composite frame which becomes a visual reference database that includes data for all frame offsets used later for the computer controlled application of masks within a sequence of frames. Masks are applied to subsequent frames of motion objects based on various differentiating image processing methods, including automated mask fitting/reshaping. Colors and/or depths are automatically applied to masks throughout a scene from the composite background and to motion objects. Areas never exposed by motion or foreground objects in a series of images may be partially or fully realistically drawn or rendered and applied to the occluded areas of the background and then automatically applied throughout the images to generate of minimal artifact or artifact-free secondary viewpoints when translating foreground objects horizontally during 2D->3D conversion.

The invention provides a depth map enhancement method blending a high-resolution color image. The method comprises the following steps: firstly, calculating the maximum depth difference of pixel points in a neighborhood according to a neighborhood difference method; dividing a low-resolution depth image into an edge region and an non-edge region; conducting secondary dividing on the non-edge region by utilizing local neighborhood variance; obtaining depth values, corresponding to a mapping point, of a point to be interpolated by utilizing a mean interpolation method and a three-dimensional convolution interpolation method respectively; sampling a high-resolution image on the low-resolution depth image based on the principle of bilateral filtering by taking the high-resolution color image as a reference image; obtain the depth values, corresponding to the mapping point, of each pixel; and finally outputting a depth enhancement image subjected to interpolation processing. The method can not only effectively remove noise interference, reduce the complexity of a depth enhancement algorithm, but also maintain the integrity of image edge information.

The invention discloses a vehicle obstacle avoidance method and device. The method comprises steps that driving data acquired by a vehicle sensor of a vehicle is acquired, and the driving data comprises the obstacle information of the driving path and sensor data of the vehicle; based on the driving data, an obstacle avoidance strategy model is utilized to determine an obstacle avoidance control instruction, and the obstacle avoidance strategy model is acquired through training by employing a depth enhancement learning algorithm in an end to end mode on the basis of historical obstacle avoidance records; the obstacle avoidance control instruction is sent to a corresponding control system to make the control system perform corresponding obstacle avoidance operation. The method is advantaged in that the obstacle avoidance success rate of the vehicle can be enhanced.

The invention discloses a depth image enhancement method based on anisotropic diffusion. The similarity among pixel points of a color image corresponding to a depth image is used as a basis of depth diffusion, and based on the known depth of the depth image, filling of a depth missing region of the depth image is finished. Through the depth image enhancement method based on the anisotropic diffusion, the defect that in a traditional depth enhancement method based on interpolation, edges of an object are blurry is overcome, meanwhile, the limit of the size of a wave filtering window in a depth enhancement method based on wave filtering is broken through, and the depth image enhancement method based on the anisotropic diffusion has the advantages of being high in universality, high in robustness and the like. The depth image enhancement method based on the anisotropic diffusion can be widely used for a variety of depth images which have missing depth regions, and in a practical application process, based on the color image and the depth image which are in correspondence to each other, a good depth image missing region filling effect can be achieved.

The invention provides a three dimensional SLAM (simultaneous localization and mapping) method based on events with a depth enhanced vision sensor. The method comprises the following components: the embedded dynamic vision sensor and an individually movable depth induction sensor are combined in order to obtain an enhanced pixel event with depth information; the pixel event is used as an only input of the three dimensional SLAM method based on events, and a panoramic map is generated by selecting a particle increment model, generating a local map from discrete probability sparse voxel grid modeling, and updating the local map by iteration. Specific hardware is not needed, the processing speed is 20 times faster than the real time speed, and position update is carried out at hundred Hertz frequency; the method has the advantages of good effects, low memory demand, low power consumption, high efficiency calculating rate, etc.

The invention provides a content recommendation method and a device based on depth reinforcement learning. The method comprises the following steps: training a depth reinforcement function to obtain atraining result for a parameter set in the depth reinforcement function; obtaining an ordered candidate set of recommended contents and a number of pieces of selected recommended contents; based on the training results of the parameter set, the comprehensive reward value of each recommended content in the candidate set being calculated by using the depth enhancement function; a comprehensive award value for each recommendation content being associated with the recommendation content and other recommendations ranked after the recommendation content; according to the result of calculation, selecting the recommended contents as the selected contents and outputting them in order. The invention synthetically considers the recommended content and the ranking of the recommended content by usingthe method of depth reinforcement learning, thereby obtaining a better recommended result.

According to one embodiment, a method for compensating for inadequate bit resolution in a light processing system includes receiving a plurality of values each indicative of an intensity level for a pixel to be displayed. Each of the values is represented by a plurality of bits of data. The method also includes determining a quantization step size for the plurality of bits of data. For at least one particular pixel of the pixels, a set of consecutive pixels including the particular pixel is selected. The method also includes determining a difference between the value associated with the particular pixel in the set and each value associated with the other pixels in the set, and also determining that all of the determined differences are less than or equal to the quantization step size. In response, a filtered value for the particular pixel in the set is generated based at least on some of the pixels in the set in addition to the particular pixel.

The invention discloses an image bit depth enhancing method. The image bit depth enhancing method includes the steps: detecting an original image to obtain an overexposure area, and then determining whether the overexposure area is an SR overexposure area or an LS overexposure area; utilizing corresponding overexposure processing methods to perform overexposure processing of the original image; and utilizing a bit depth enhancing algorithm of a normal exposure image to perform bit depth enhancement of the image after overexposure processing. Compared with the overall transformation, the image bit depth enhancing method takes the characteristics of different areas in the image into account. Compared with the local transformation, the image bit depth enhancing method processes the overexposure area in advance, and can utilize the same transformation function for the whole image during the subsequent bit depth enhancing process so that the complexity of bit depth enhancement is reduced and the efficiency of the whole algorithm is improved.

A video system includes an analyzer and a bit depth predictor. The analyzer receives a first coded video signal, which is indicative of first values for pixels. The first values are associated with a first bit depth. The analyzer, for each pixel, analyzes the first values for the pixels located in a neighborhood that contains said each pixel. The bit depth predictor, based at least in part on the analysis, generates a second coded video signal that is indicative of second values for the pixels. The second values are associated with a second bit depth that is different than the first bit depth.

Processing an image having a first bit depth includes performing two or more iterations of a bit depth enhancement operation that increases the bit depth of the image to a second bit depth that is higher than the first bit depth. The bit depth enhancement operation includes dividing the image into a plurality of areas, performing an edge detection operation to identify one or more areas from the plurality of areas that do not contain edge features, and applying a blur to the one or more areas from the plurality of areas that do not contain edge features. In a first iteration of the of the bit depth enhancement operation, the plurality of areas includes a first number of areas, and the number of areas included in the plurality of areas decreases with each subsequent iteration of the bit depth enhancement operation.

The present disclosure relates to a depth-enhancing screen for producing a simulated 3D image. The screen comprises a multi-curved Fresnel lens which when viewed in cross-section along the or each longest line linking two points on the edge of the lens, has a curved cross-section with an apex in the central region of the lens, and wherein each end of the curve flattens before it reaches the edge of the lens.

The present invention relates to a texture-based depth boundary correction method. The method comprises the steps of A1, inputting a texture image and a corresponding depth image, wherein the images are acquired by a depth sensor (such as a Kinect); A2, extracting the boundary of the texture image and the boundary of the depth image, and acquiring a depth boundary dislocation figure with the boundary of the texture image as the reference; A3, calculating the pixel value difference between an accurate depth point (sweet spot) and an error depth point (dead point) and determining a boundary dislocation area; A4, according to the distribution characteristics of the boundary dislocation area, adaptively determining the side length of a square window for depth enhancement processing, correcting the depth of the dead point in the window, eliminating the boundary dislocation area. According to the invention, the accuracy and the time-domain stability of the boundary of the depth image, acquired by the Kinect and other low-end depth sensors, are significantly improved. The method is applied to the fields of three-dimensional reconstruction, free viewpoint video coding and the like, and can effectively improve the scene three-dimensional reconstruction quality and the coding efficiency.