Abstract: A light field image processing method is disclosed for removing occluding foreground and blurring uninterested objects, by differentiating objects located at different depths of field and objects belonging to distinct categories, to create see-through effects. In various embodiments, the image processing method may blur a background object behind a specified object of interest. The image processing method may also at least partially remove from the rendered image any occluding object that may prevent a viewer from viewing the object of interest. The image processing method may further blur areas of the rendered image that represent an object in the light field other than the object of interest. The method includes steps of constructing a light field weight function comprising a depth component and a semantic component, where the weight function assigns a ray in the light field with a weight; and conducting light field rendering using the weight function.

Abstract: A light-adjusting structure, a method for manufacturing the light-adjusting structure, and a light-adjusting module are provided, which belong to the field of display technology, the light-adjusting structure includes: a first flexible substrate and a second flexible substrate oppositely arranged; a first transparent electrode and a second transparent electrode which are located between the first flexible substrate and the second flexible substrate; a first alignment layer located on a side of the first flexible substrate facing towards the second flexible substrate; a second alignment layer located on a side of the second flexible substrate facing towards the first flexible substrate; and a spacer and a dye liquid crystal layer which are located between the first alignment layer and the second alignment layer. The solutions of the present disclosure can meet light-adjusting requirements of vehicle windows.

Abstract: The present invention discloses an image interpolation method and device based on RGB-D images and a multi-camera system, wherein the method comprises performing camera calibration on each camera in the multi-camera system; clarifying a position of a new camera for interpolation according to position information of the each camera in the multi-camera system, and calculating a camera pose of the new camera according to camera calibration data; calculating a plurality of initial interpolated images that have a one-to-one correspondence with designated images captured by the each camera of the multi-camera system according to a projection relationship of the camera and the pose information of the each camera; performing image fusion on each initial interpolated image to obtain a fused interpolated image; and performing pixel completion on the fused interpolated image so as to obtain an interpolated image related to the new camera.

Abstract: A flexible circuit board comprises: a flexible substrate, a lead layer, a first insulating layer and a pin layer. The lead layer comprises a plurality of leads provided at one side of the flexible substrate; the first insulating layer is provided at a side of the lead layer facing away from the flexible substrate, the first insulating layer is provided with a plurality of via holes, and the orthographic projection of each via hole on the flexible substrate at least partially overlaps with the orthographic projection of the corresponding lead on the flexible substrate; the pin layer comprises a plurality of pins provided at a side of the first insulating layer facing away from the flexible substrate, the plurality of pins are provided in one-to-one correspondence with the plurality of leads, and the pins are respectively connected to the corresponding leads through the via holes.

Abstract: The present invention discloses a method for real-time rendering of giga-pixel images. Image data are subject to offline pre-processing, and then are subject to data decoding and redirection through a decoding module. A corresponding scheduling strategy is determined according to different image inputs, and rendering is executed by a renderer. The present invention may realize the real-time rendering of a giga-pixel panoramic view in a conventional display device, to greatly reduce the resource allocated for rendering of giga-pixel images. The present invention may render an image originally requiring a 40G or more video memory capacity on a common video card with a 1G-4G video memory capacity.

Abstract: The present invention discloses a method for real-time rendering of giga-pixel images. Image data are subject to offline pre-processing, and then are subject to data decoding and redirection through a decoding module. A corresponding scheduling strategy is determined according to different image inputs, and rendering is executed by a renderer. The present invention may realize the real-time rendering of a giga-pixel panoramic view in a conventional display device, to greatly reduce the resource allocated for rendering of giga-pixel images. The present invention may render an image originally requiring a 40G or more video memory capacity on a common video card with a 1G-4G video memory capacity.

Abstract: A human hairstyle generation method based on multi-feature retrieval and deformation includes: acquiring a hair style mask; identifying feature points of a human face and match the feature points with a hair style database; aligning an image with a standard human face to acquire a corresponding hair region; calculating Minkowski distances between the hair region and hair masks of all frontal faces in the hair style database; assigning corresponding weights after sorting the Minkowski distances from small to large; training a deep learning network to detect hair styles of hair basic blocks at different scales; and taking out a most similar hair style picture. The present invention utilizes a single frontal photo of the human face, and retrieves a three-dimensional hair model most similar to the photo by retrieving a database in a mass three-dimensional hair style database, to avoid manual modeling, thereby improving efficiency and ensures high fidelity.

Abstract: A human hairstyle generation method based on multi-feature retrieval and deformation includes: acquiring a hair style mask; identifying feature points of a human face and match the feature points with a hair style database; aligning an image with a standard human face to acquire a corresponding hair region; calculating Minkowski distances between the hair region and hair masks of all frontal faces in the hair style database; assigning corresponding weights after sorting the Minkowski distances from small to large; training a deep learning network to detect hair styles of hair basic blocks at different scales; and taking out a most similar hair style picture. The present invention utilizes a single frontal photo of the human face, and retrieves a three-dimensional hair model most similar to the photo by retrieving a database in a mass three-dimensional hair style database, to avoid manual modeling, thereby improving efficiency and ensures high fidelity.

Abstract: A light field image processing method is disclosed for removing occluding foreground and blurring uninterested objects, by differentiating objects located at different depths of field and objects belonging to distinct categories, to create see-through effects. In various embodiments, the image processing method may blur a background object behind a specified object of interest. The image processing method may also at least partially remove from the rendered image any occluding object that may prevent a viewer from viewing the object of interest. The image processing method may further blur areas of the rendered image that represent an object in the light field other than the object of interest. The method includes steps of constructing a light field weight function comprising a depth component and a semantic component, where the weight function assigns a ray in the light field with a weight; and conducting light field rendering using the weight function.