206 results about "Texture synthesis" patented technology

Systems and methods for providing convolutional neural network based image synthesis using localized loss functions is disclosed. A first image including desired content and a second image including a desired style are received. The images are analyzed to determine a local loss function. The first and second images are merged using the local loss function to generate an image that includes the desired content presented in the desired style. Similar processes can also be utilized to generate image hybrids and to perform on-model texture synthesis. In a number of embodiments, Condensed Feature Extraction Networks are also generated using a convolutional neural network previously trained to perform image classification, where the Condensed Feature Extraction Networks approximates intermediate neural activations of the convolutional neural network utilized during training.

The invention discloses a method for carrying out the three-dimensional human face modeling by a single human face photo to obtain a realistic human face, and the method belongs to the field of graphics and images. The method utilizes a priori knowledge of a human face structure to estimate the depths of two-dimensional feature points on the photo, thereby obtaining similar three-dimensional coordinates; the similar three-dimensional feature points are taken as control points, and the dirichlet free deformation algorithm is adopted to realize the fitting of the neutral human face to the specific human face. Then, a method which is based on the regional expansion is adopted to realize the texture synthesis and the mapping, and the region which is not covered is repaired by adopting the interpolation or the block filling method. The method has simple calculation, high modeling accuracy and can rapidly utilize a small number of feature points to realize the three-dimensional modeling of the human face on the single photo, and the generated three-dimensional human face model has very strong realistic feeling under various postures due to the accurate estimation of the depths of the feature points. The method can also be used for the three-dimensional reconstruction of other objects.

A novel multi-resolution block sampling based texture analysis/synthesis algorithm. A reference texture is assumed to be sample from a probability function. The synthesis of a similar, but distinctive, synthetic texture is handled in a process and by an apparatus that first estimates and then resamples the probability function. In order to achieve good and fast estimation of the probability function for a reference texture and in order to retain the texel structural information during the synthesis, a novel concept of block sampling and a corresponding novel texture synthesis scheme based on multi-resolution block sampling is employed. As a result of this novel approach, the computational complexity of the present invention is much lower than that of other approaches to the problem. In addition, for textures that exhibit a high degree of directionality, a process, which integrates estimation of dominant texture direction and the synthesis algorithm is employed to handle directional textures. The dominant direction is used to orient and then control the synthesis process so as to preserve the dominant reference image direction.

In accordance with a first aspect, filling disocclusions in a virtual view video has identifying a background portion of a first virtual view image of the virtual view video based on a depth/disparity map associated with a first virtual view image, updating a background sprite image based on the background portion of a first virtual view image, and filling a disoccluded area of a second virtual view image of the virtual view video based on a corresponding portion of the background sprite image. According to a second aspect, filling a disoccluded area in a virtual view image is performed by patch-based texture synthesis being responsive to a depth/disparity map associated with a virtual view image.

The invention provides an image-based method for generating novel visual appearance in a new image. Synthetic texture is stitching together from small patches in existing images. First, we use a least cost path determination to determine the local boundaries between the patches. Second, we perform texture transfer by rendering an arbitrary object with a synthetic texture taken from a different object. More generally, we provide methods for rendering entire images in styles of different images. The method works directly on pixel images, and does not require 3D information.

The invention discloses a self-adaptation camouflage design method based on a background. According to the method, on the basis of a texture synthesis algorithm based on pixels, a whole camouflage image is synthesized pixel by pixel; in the synthesis process, a self coding algorithm is adopted to extract a texture primitive containing background sample complete texture features, and a neighborhood of the pixels to be synthesized is determined; by means of the neighboring correlation principle and a helix supplement search mechanism, a neighborhood matched with a target pixel neighborhood is searched for in a background sample; similarity matching is carried out through a pixel and texture similarity matching rule, pixels in the neighborhoods which are most similar in the sample are selected to serve as synthesis pixels, pixels of a target image are synthesized according to the sequence of scanning lines, and after synthesis, color quantization is carried out on the image by means of the clustering algorithm to generate a final camouflage image. By means of the method, similar camouflage images can be synthesized quickly according to the background, and the application aim of quickly camouflaging a target in a self-adaptation mode can be achieved.

The invention relates to an image restoring method which dose automatically restoring course after user enters into the restored area. The automatically restoring course uses the edge side as the unit of primary iteration packing and then iterates the following four steps until packs all the restored area: first step, computing all the precedent degree of the edge; second step, ascertaining the packing edge side of the current iteration by the precedent degree and the similarity degree criterion; third step, adopting line synthesis method and expends the known line and structure information; fourth step, updating the synthetic picture element ascertain value of the current iteration side.

The present invention relates to an apparatus for time-coherent texture synthesis including a texture preprocessor for receiving as input information a 2D texture image and a 3D triangular mesh, and preprocessing the 2D texture image in a form suitable to rapid searching, a vector field generator for defining a vector field on a 3D surface of the 3D triangular mesh, a color search unit for finding a color of each edge of a triangle having the defined vector field in consideration of a previous frame, and a texture synthesizer for determining texture coordinates of the triangle using the found colors. The texture preprocessor further receives information regarding a size of the texture to be synthesized and an initial vector field orientation.

The invention provides equipment and a method for modeling a three-dimensional (3D) object. By the equipment and the method, a depth image and a multi-view-angle color image can be fused, so that the 3D object is automatically modeled on the basis of a predefined 3D object template. The equipment comprises an image receiving unit, a 3D template providing unit and an object modeling unit, wherein the image receiving unit is used for receiving the shot color image and depth image; the color image comprises a positive color image and at least two lateral color images at other view angles; the depth image is a positive depth image; the 3D template providing unit is used for providing the 3D object template predefined for a specific type of object; the 3D object template has a shape feature and a texture feature; and the object modeling unit is used for deforming the 3D object template provided by the 3D template providing unit based on the color image and the depth image which are received by the image receiving unit, and performing texture synthesis on the color image and the deformed 3D object template to generate a 3D object model.

The present invention provides a generation method and generation device of a pencil drawing fusing skeleton strokes and textural features. The generation device of a pencil drawing comprises an original image acquisition unit, a gray image generation unit, a gradient image generation unit, a skeleton stroke layer generation unit, a tone layer generation unit, a texture feature rendering layer generation unit and a pencil drawing image generation unit. The generation method provided by the invention comprises: a gradient image is obtained through graying, deburring processing and edge detection of and inputted image; the gradient image is processed to be a skeleton stroke layer; the skeleton stroke layer is processed to be a tone layer; a texture feature rendering layer is obtained through texture synthesis and establishment of a tone mapping mathematical model; and multiply mixing of the skeleton stroke layer and the texture feature rendering layer is performed to obtain a final pencil drawing with a pencil hand-painting drawing effect. According to the invention, the step of generation of a pencil drawing is simple, the efficiency is high and the hand-painting effect is outstanding.

The invention discloses a texture synthesis and image repair method based on wavelet transformation. The method comprises the following steps: performing the wavelet transformation on an input damaged image, and decomposing the image into a low-frequency sub-image and three high-frequency sub-images; marking an repair region of the original image to form a binary mask image and sampling the mask image to match resolution of the mask image with a wavelet sub-image; analyzing structural information of the original image by a high-frequency coefficient of wavelet decomposition to determine an repair sequence of defect pixels, and filling and repair various sub-images by the texture synthesis method; and performing wavelet inversion on various wavelet sub-images to obtain the repaired image. The method helps repair the image in combination with structure and texture information of the image on the basis of the wavelet transformation, has excellent effects of repairing large-area damage images and greatly improves repair efficiency.

The invention belongs to the field of computer vision and computer graphics, and provides a method which can ensure a newly-generated video sequence to be vivid and reliable and solve the problem of high requirements on an experimental device. According to the technical scheme adopted by the invention, the realistic animation generation method based on Kinect uses a single Kinect v2.0 to carry out static three dimensional model and color depth skeleton information acquisition on a person. A marker-free motion capture method is used, skeleton motion information of each frame and a three dimensional model after deformation are obtained. After a user designates a new action, on the basis of the built data collection, a video for the new action by the person is generated through the proposed sparse representation-based low rank matrix filling texture synthesis method. The method of the invention is mainly applied to video processing.

A method, system and program product for the detection of removed objects in video images is disclosed. An object (or objects) is within a fixed camera's FOV (Field of View), is designated to be monitored. If an object is not designated, an algorithm can be used to segment and identify all the objects in the image that is acquired by a fixed camera. In such case, all these objects can be monitored. A reference frame around the monitored object that also includes the surrounding background can be determined. The object can be extracted out of the reference frame, thereby creating a “reference with hole” image and an object reference image. The hole in the “reference with hole” image can be filled by either an in-painting algorithm or a texture synthesis algorithm depending on the complexity of the surrounding background.

A method for compressing and synthesizing a bidirectional texture function realizes compression and synthesis of the bidirectional texture function data by principal component analysis and color rectangular texture blocks. The method comprises the following steps; firstly the bidirectional texture function data are grouped according to different viewpoints and compressed by the principal component analysis algorithm; and then data blocks are immediately extracted from the compressed bidirectional texture function data by an error calculation method, one color rectangular texture block is formed by every four data blocks, and multi-dimension texture synthesis is carried out by the color rectangular texture blocks; and at last, the compressed and synthesized bidirectional texture function data are drawn on a 3D model surface. The method can compress large-scale bidirectional texture function data and synthesize new multi-dimension texture images according to the existing texture images. The method has the advantages of fast synthesis and precious effect, solves the consistency problem of the multi-dimension texture synthesis, expands means of the multi-dimension texture sources, and meets the realistic drawing requirements in a virtual environment.

The invention discloses a looseness defect image generation method based on texture synthesis. The method includes the steps of (1) cutting an appropriate defect sample from an X-ray image of an actual workpiece with a loose texture defect; (2) removing a background image in the sample to obtain a defect texture image sample; (3) in a new workpiece image, according to the required area, in any expected position of the workpiece, selecting a rectangular area as a defect generation area; (4) using a texture synthesis method to generate a new defect texture image of the same size as that of the defect generation area; and (5) superposing the generated defect texture image to the position of the workpiece image. The looseness defect image generation method based on texture synthesis realizes rapid generation of a looseness defect image, the shape and area of a generated defect have randomness, an image effect is very close to a real image, and through human eye subjective judgment of an expert, the generated looseness defect image can be accepted as a defect image sample.

The invention discloses a model construction method, and particularly relates to a face image real-time beautifying and texture synthesizing method for three-dimensional face reconstruction. Accordingto the method, through the processes of face feature point extraction by utilizing an ASM algorithm, color conversion, face seamless fusion, eyebrow repair processing, three-dimensional face reconstruction and the like, the automatic detection and beautification can be realized on the face, so that a relatively better face reconstruction result is obtained. Experimental results show that the algorithm provided by the invention can obtain a satisfactory reconstruction effect, including reconstructing a shielded eyebrow region, has the superior operation efficiency compared with other methods,and can achieve the real-time performance. The face image is beautified based on the standard skin color to generate the face texture map, and the face texture map is applied to the three-dimensionalface reconstruction to obtain the personalized three-dimensional face model.