44 results about "Quadtree segmentation" patented technology

The invention provides an encoder, decoder, video frame encoding method and bit stream decoding method. The encoder for receiving a video frame and performing encoding processes to generate an encoded bit stream includes: a fidelity enhancement block, for performing a fidelity enhancement technique on the video frame utilizing a quad- tree partition, and generating fidelity enhancement information including a parameter associated with the quad-tree partition structure; and an entropy coding block, coupled to the fidelity enhancement block, for encoding the fidelity enhancement information, and embedding the encoded fidelity enhancement information into the encoded bit stream. The said method and device make the decoder immediately stores the fidelity enhancement information and execute the fidelity enhancement without waiting for decoding all the bit streams through indicating the position of the fidelity enhancement information by an indicating needle.

The invention discloses a depth learning target detection method based on pre-segmentation and regression, mainly to solve the problems that the existing target detection method is poor in small target detection precision and long in detection time. The method comprises steps: 1) a quadtree segmentation algorithm is used to extract a region of interest in a to-be-detected image; 2) a basic convolution layer and an auxiliary convolution layer are used to carry out feature extraction on the region of interest, and feature graphs of multiple scales are obtained; 3) the position information of a default border is calculated on the feature graphs of multiple scales, a convolution filter is used for detection on the feature graphs of multiple scales, and multiple predicted borders and multiple category scores are obtained; and 4) non-maximum suppression is used for the multiple predicted borders and the multiple category scores, and the final target border position and the category information are obtained. A small target in the image can be quickly and accurately detected, and the method can be used for target real-time detection in unmanned aerial vehicle aerial photographing.

A method and apparatus for constructing a map for a mobile robot to be able to reduce a data amount and increase an approach speed. The method includes: searching for a plurality of feature data occupying an arbitrary space by scanning a surrounding environment of the mobile robot; performing quadtree segmentation on first feature data of the plurality of feature data to generate a plurality of first node information as a result of the quadtree segmentation; determining a position of second feature data of the plurality of feature data with respect to the first feature data; and performing a neighborhood moving algorithm for generating a plurality of second node information of the second feature data according to the position of second feature data by using the plurality of first node information.

Video processing methods and apparatuses for candidate set determination for a current block partitioned from a parent block by quad-tree splitting comprise receiving input data of a current block, determining a candidate set for the current block by prohibiting a spatial candidate derived from any of neighboring blocks partitioned from the same parent block or determining the candidate set for the current block by conducting a pruning process if all the neighboring blocks are coded in Inter prediction and motion information of the neighboring blocks are the same, and encoding or decoding the current block based on the candidate set by selecting one final candidate from the candidate set. The pruning process comprises scanning the candidate set to determine if any candidate equals to the spatial candidate derived from the neighboring blocks, and removing the candidate equals to the spatial candidate from the candidate set.

The invention provides an LOD model generation method based on a linear quadtree. The method comprises the following steps: performing linear quadtree segmentation on to-be-drawn terrain, storing data of the to-be-drawn terrain in a two-dimensional array, reading information of each node in the linear quadtree from the two-dimensional array through an index, storing leaf node information of the linear quadtree, organizing terrain nodes, enabling the terrain nodes in the linear quadtree to be stored in a one-dimensional array, establishing a terrain node evaluation mechanism on the basis that a static error and a dynamic error are considered, performing crack elimination on different segmentation levels of the adjacent terrain nodes, and obtaining an LOD model based on the linear quadtree. The method is applied to a terrain simplification aspect in a computer graphics and image field, and the terrain nodes are organized by utilizing a linear quadtree storage model and through combination use of a two-dimensional filling curve; on the basis that error factors are fully considered, the node evaluation mechanism is established, and a terrain crack problem is eliminated; and terrain drawing efficiency can be obviously improved.

A method and apparatus for video coding using flexible quadtree and binary tree block partition are disclosed. According to this method, a current block is partitioned into multiple final sub-blocks using one or more stages of quadtree partitioning and binary tree partitioning, where at least one sub-block is generated by applying the quadtree partitioning to one previous sub-block generated by the binary tree partitioning. According to another method, one or more horizontal splitting flags and one or more vertical splitting flags are used for determining block partition of the current block into multiple sub-blocks using one or more stages of quadtree partitioning and binary tree partitioning. Multiple sub-blocks partitioned from the current block are determined according to the horizontal splitting flags and the vertical splitting flags. The multiple sub-blocks are then encoded or decoded.

The invention relates to a video transcoding method based on machine learning. The video transcoding method comprises the following steps: modeling a quadtree partition mode of an encoding unit in an original video into multiple binary classifiers with different levels, then selecting an optimal feature set, finally, learning a data set composed of a characteristic vector and an optimal encoding parameter, namely introducing a machine learning method into video transcoding, and converting a parameter determining problem in video encoding into a classification problem. Therefore, a corresponding classifier can be selected according to the size of the current encoding unit, and a classification probability value is compared with a corresponding adaptive threshold to select the optimal encoding parameter for encoding. The adaptive probability threshold is adaptively adjusted for different video scenes, therefore, an optimal transcoding speed and transcoding quality can be obtained to ensure relatively small power consumption in a transcoding process, and the transcoding complexity is effectively reduced on the premise of guaranteeing the transcoding rate distortion performance.

The present invention relates to a method and apparatus for encoding / decoding an image, and to a method and apparatus for recording a quantization parameter differential value based on a quadtree segmentation in a largest coding unit (LCU), and adaptively predicting a quantization parameter value based on context information of a neighboring coding unit (CU). The method and apparatus for encoding / decoding a quantization parameter based on a quadtree structure may effectively present block information having a quantization parameter differential value based on segmentation information of a CU, and may perform an adaptive prediction of a quantization parameter value using the context information of the neighboring CU such as a block size, block segmentation, quantization parameter value, and the like.

The invention discloses a remote-sensing image compression method for discrete wavelet transform based on self-adaptation direction prediction and belongs to the technical field of remote-sensing image processing. The method comprises the steps that 1) 15 directions which are formed by close-end integer and fraction pixels and far-end integer pixels are considered in a combined mode, a direction graph is established; 2) remote-sensing images are subjected to two-dimensional discrete wavelet transform based on self-adaptation direction prediction, namely quadtree division based on direction consistency is used for dividing the remote-sensing images to a plurality of image blocks which are not overlapped, a new direction prediction model is used for obtaining the best transformation direction of each image block, lifting wavelet transform is completed in the direction; 3) a transformed image coefficient is subjected to SPIHT coding, direction information is subjected to self-adaptation arithmetic coding, and a final code stream is obtained. According to the method, visual quality and peak signal-to-noise ratio of the remote-sensing images are effectively improved, computing complexity of direction wavelet transform is lowered, and the method can be widely used for efficient coding compression of the aviation and satellite remote-sensing images.

The invention relates to the technical field of image processing, and provides a defogging method and system for an image containing a large-area sky region. The method comprises the steps of carrying out quadtree segmentation on a photographed foggy image I (x), segmenting the foggy image into a plurality of image regions and obtaining seed points in the plurality of image regions; carrying out sky region growth by taking obtained seed points as bases, obtaining images containing sky regions and non-sky regions, and calculating atmosphere light values A; calculating medium transmission rates t<1 > (x) of the non-sky regions, and moreover, uniformly setting the medium transmission rate of the sky regions to be t <sky >; carrying out boundary fuzzification on the images containing the sky regions and non-sky regions and calculating to obtain gauss weighting parameters; carrying out transmission rate fusion on the sky regions and non-sky regions in the images; and calculating a defogged image J (x) according to the atmospheric scattering model of the foggy image, thereby obtaining a clear image without fog. The method and the system are especially suitable for processing the foggy and hazy image containing the large-area sky region.

The invention discloses an interframe predication quick mode selecting method for high-performance video coding. The method comprises the steps of conducting SKIP mode and Inter2Nx2N mode verification on each coding unit (CU) in video frames, judging whether the SKIP mode is a current optimized mode or not if the error of the rebuilt pixel and the original pixel value of each CU unit does not surpass a threshold, or the error of the rebuilt pixel and the original pixel value is distributed in a preset range or predicted residuals do not surpass preset numerical value, judging the SKIP mode as the optimized mode of the current CU if SKIP mode is the current optimized mode, otherwise, judging whether the the current CU is a smallest coding unit (SCU) or not, conducting four quad tree based segmentation on the current CU if the current CU is not an SCU, and conducting mode verification on sub-units one by one. The method effectively reduces coding complexity of a coding side so as to be adapted to mobile equipment and much video application high in real-time requirement.

The invention provides a fractal image compression method based on the combination of a quad tree and neighborhood searching. The method comprises the following steps of: setting the maximum and minimum depths of the quad tree and an error threshold value; dividing an image into value region blocks according to the minimum depth; extracting a value region block to be matched; searching for a neighborhood definitional domain block corresponding to the value region block to be matched; calculating affinity, matching the affinity and the value region block, and calculating a mean square error of the affinity and the value region block; comparing the mean square error with the threshold value; if the mean square error is small, recording a current value region block, a current definitional domain block and the affinity; if the mean square error is greater, judging whether the current value region block is the maximum depth or not, recording the current value region block, a definitional domain block with the lowest error, and the affinity if the current value region block is the maximum depth, and if the current value region block is not the maximum depth, dividing the value region block into sub-blocks according to the quad tree as value region blocks to be matched, correspondingly reducing the corresponding definitional domain block, and repeating the matching process; and judging whether all the value region blocks to be matched are matched or not, and if not all the value region blocks to be matched are matched, repeating the matching process. Searching time is shortened, and high image recovery quality is ensured.

The invention provides a terrain multi-scale TIN (Triangulated Irregular Network) online visualization method oriented to the virtual earth. The method comprises the following steps of global multi-scale TIN scale parameter mathematical model construction, terrain multi-scale TIN automatic construction and virtual segmentation under characteristic constraints, the online visualization of a multi-scale TIN topography spherical surface and real-time crack elimination, wherein the step of the terrain multi-scale TIN automatic construction and virtual segmentation under characteristic constraintsis realized as follows; the step of the online visualization of the multi-scale TIN topography spherical surface comprises the step of carrying out calculation to obtain visual terrain blocks of different levels according to point-of-sight information; and the step of the real-time crack elimination comprises the following step of carrying out quick decoding during real-time visualization to finish the edge connection of the adjacent boundary of terrain blocks of different levels. By use of the method, while the structural flexibility of the TIN is kept, the multi-scale irregular modeling andthe quadtree segmentation of a large-scale terrain are realized, meanwhile, the seamless real-time rendering of a mixed scale TIN terrain is supported, the terrain data processing and expression accuracy of the virtual earth can be greatly improved, a seamless fusion ability for irregular GF spatial data is also greatly improved, so that the high-fidelity expression requirement of a geographical environment can be met, and each category of reliable geographical analysis application can be supported.

The invention discloses an image enhancement method based on contrast and structural similarity. The method comprises steps: an original single haze image is inputted; robust estimation on atmospheric light is realized through iterative quadtree segmentation; content-based superpixel segmentation technology is adopted for segmenting the image into local area blocks; a cost function is built, an atmospheric propagation chart of each local area block is estimated, and the optimal transmission parameter for each block is obtained; and an atmospheric degradation model is used for image restoration. The de-haze method provided by the invention can effectively improve visibility of a recovery scene image, has the advantages of fast operation speed and good image recovery effects, and can be applied to a system with high real-time performance requirements.

The invention designs a weighted least square filtering-based image defogging method, and research is carried out according to an existing single image defogging algorithm. The method comprises the specific implementation steps of firstly improving estimation of an atmospheric light value by utilizing super-pixel processing and quadtree segmentation methods; secondly according to dark channel prior knowledge and an image space conversion thought, performing multi-time fusion by utilizing weighted least square filtering, thereby obtaining accurate atmospheric curtain parameters; thirdly performing inverse conversion through the obtained atmospheric curtain parameters to obtain an atmospheric transmission function; and finally substituting the obtained atmospheric light value and the atmospheric transmission function into a model for performing restoration processing on a foggy image. According to the method, the atmospheric light value can be accurately estimated, the transmission function can be indirectly obtained, the definition, contrast and detail information of the image can be effectively improved, and especially the processing effect on a close shot is better.

The invention discloses a terrain rendering method which comprises the following steps: acquiring terrain image data, and performing quadtree segmentation on the terrain image data to obtain a plurality of terrain tiles; Creating a texture array of each terrain tile according to the layer-by-layer number, the image width and the image height of the terrain image data; Storing texture images and texture coordinates of all layers of the terrain image data in the current terrain tile through the texture array of each terrain tile; And performing terrain rendering according to the texture image and the texture coordinates in the texture array of each terrain tile. Wherein the texture array is not limited by the layer-by-layer number of the image, so that the limitation of a computer display card on the number of texture units can be avoided, layer-by-layer rendering is not needed in the rendering process, and the rendering efficiency is improved. The terrain rendering device and equipmentand the readable storage medium disclosed by the invention also have the above technical effects.

A method and apparatus for constructing a map for a mobile robot to be able to reduce a data amount and increase an approach speed. The method includes: searching for a plurality of feature data occupying an arbitrary space by scanning a surrounding environment of the mobile robot; performing quadtree segmentation on first feature data of the plurality of feature data to generate a plurality of first node information as a result of the quadtree segmentation; determining a position of second feature data of the plurality of feature data with respect to the first feature data; and performing a neighborhood moving algorithm for generating a plurality of second node information of the second feature data according to the position of second feature data by using the plurality of first node information.

A method and apparatus for constructing a map for a mobile robot to be able to reduce a data amount and increase an approach speed. The method includes: searching for a plurality of feature data occupying an arbitrary space by scanning a surrounding environment of the mobile robot; performing quadtree segmentation on first feature data of the plurality of feature data to generate a plurality of first node information as a result of the quadtree segmentation; determining a position of second feature data of the plurality of feature data with respect to the first feature data; and performing a neighborhood moving algorithm for generating a plurality of second node information of the second feature data according to the position of second feature data by using the plurality of first node information.

The invention discloses an LOD (Level of Detail) technology-based scene rendering method. The method comprises the steps of determining a square rendering terrain; segmenting the rendering terrain by utilizing a quad-tree; updating and traversing the rendering terrain; repairing a cracked terrain; and the like. An LOD technology is a technology for adjusting grid data precision by stage, and the grid data precision is relatively coarse for triangles relatively far away from a point of sight and is relatively fine for triangles relatively close to the point of sight. After use of the LOD technology, the large scene rendering effect is better improved in efficiency and effect, and the game running speed is increased.

The invention discloses a thermodynamic diagram generation method and system considering space density difference, and the method comprises the steps: collecting experiment data, and storing the experiment data in a form of a space point group; in combination with a thermodynamic diagram scale, calculating a real distance corresponding to the screen resolution, and taking the real distance as a side length threshold value of quadtree segmentation; counting the number of the spatial point groups, and calculating a point number threshold value; carrying out quadtree segmentation based on point number threshold and side length threshold constraints on the spatial point group, and extracting block side length information reflecting a point group distribution aggregation degree; calculating adaptive bandwidth values point by point according to the corresponding relation between the side length of the quadtree block and the bandwidth parameter of the nearest neighbor kernel density function;and substituting the adaptive bandwidth value into a nearest neighbor kernel density function, calculating a nearest neighbor kernel density function value, and presenting a result in a Heat-map form. According to the method, quadtree segmentation and the nearest neighbor method are combined, the bandwidth values are calculated point by point, and the real distribution characteristics of the spatial point group are expressed more objectively.

The invention discloses an image defogging method based on dark channel compensation and atmospheric light value improvement, which not only can effectively correct an underestimated dark channel value and weaken the halo effect at the edge of an image scene, but also can accurately obtain the atmospheric light value of an image, so that the recovered image is clearer and more natural, and the details are more abundant. The method comprises the following specific implementation steps: firstly, solving a minimum channel image by utilizing R, G and B color channels of an original image, and obtaining a compensated dark channel image by virtue of a dark channel compensation model; then, calculating the atmospheric light value of the image through the steps of graying processing, quadtree segmentation and the like on the original image; and finally, estimating the image transmissivity by combining the dark channel image and the atmospheric light value, and recovering a fog-free image through an atmospheric scattering model. Experimental results and subjective and objective evaluation prove the feasibility and effectiveness of the method.

The invention discloses a DirectX-based game engine design method. The DirectX-based game engine design method comprises respectively designing a rendering engine, an input system, a sound system, a physical system treating collision and motion, an animation system, an artificial intelligence system, a scene management system and the like, wherein a rendering method of the rendering engine comprises the steps of determining a square rendering terrain, partitioning the rendering terrain by utilizing a quadtree, updating and traversing the rendering terrain and repairing a cracked terrain. An LOD technology is a technology of adjusting grid data precision degree in stages and has the performance that adjusting of a triangle farther from a view point is rough and adjusting of a triangle closer to the view point is exquisite. By adopting the LOD technology, rendering effect on a large-scale scene is respectively greatly improved in efficiency and effect, and running speed of a game is improved.

The present invention provides an image encoding method and an image decoding method. The image decoding method, according to the present invention, comprises the steps of: determining a transformation performing region; segmenting the determined transformation performing region into at least one sub-transform block by using at least one of quadtree segmentation and binary tree segmentation; and performing inverse transformation on the at least one sub-transform block.

The invention provides a UUV near sea surface visible light image defogging method. According to the method, (1), an original foggy image is acquired, and the minimum value of three color channels is solved and is recorded as M(x); (2), an improved method of M(x) and a quadtree segmentation method is utilized to solve global atmosphere light; (3), adaptive surface fuzziness of the M(x) is carried out; (4), an atmosphere scattering function is solved; and (5), a fog imaging physical model is utilized to solve a no-fog image which is outputted. According to the method, adaptive surface fuzziness of the minimum value image of the three color channels is carried out, each pixel has a respective convolution matrix, edges of a target object can be excellently kept, subsequent target identification, positioning and tracking are facilitated, moreover, for solving a problem of a local dark phenomenon after filtering, the global atmosphere light is utilized to roughly divide the image into a dark area and a bright area, and different radiuses are utilized for fuzzy processing. The method is advantaged in that the relatively good defogging effect is realized, and a sky halation phenomenon can be overcome.

The invention discloses an InSAR rapid image registration method based on quadtree segmentation. The method is characterized in that through the combination of an image quadtree segmentation criterionand an FFT maximum correlation coefficient registration principle, rough registration and rough segmentation are carried out on the InSAR main and auxiliary complex images, and then the quadtree treestructure image block segmentation and sub-block FFT maximum correlation coefficient sub-pixel registration are carried out, the sub-block quadtree segmentation and the sub-block sub-pixel registration are carried out on the InSAR complex image one by one, and the global image interpolation of the traditional FFT maximum correlation coefficient registration is avoided, and the rapid sub-pixel registration of the large-scene InSAR complex image is achieved. The global image interpolation of the traditional FFT maximum correlation coefficient registration is avoided, the operation efficiency ofthe large-scene InSAR complex image registration is effectively reduced, and the good registration precision and operation efficiency in the InSAR main and auxiliary complex images with large pixel offset change can be achieved.

The present invention relates to a method and apparatus for encoding / decoding an image, and to a method and apparatus for recording a quantization parameter differential value based on a quadtree segmentation in a largest coding unit (LCU), and adaptively predicting a quantization parameter value based on context information of a neighboring coding unit (CU). The method and apparatus for encoding / decoding a quantization parameter based on a quadtree structure may effectively present block information having a quantization parameter differential value based on segmentation information of a CU, and may perform an adaptive prediction of a quantization parameter value using the context information of the neighboring CU such as a block size, block segmentation, quantization parameter value, and the like.

The present invention provides an image encoding method and an image decoding method. The image decoding method, according to the present invention, comprises the steps of: determining a transformation performing region; segmenting the determined transformation performing region into at least one sub-transform block by using at least one of quadtree segmentation and binary tree segmentation; and performing inverse transformation on the at least one sub-transform block.

The invention discloses an interframe predication quick mode selecting method for high-performance video coding. The method comprises the steps of conducting SKIP mode and Inter2Nx2N mode verification on each coding unit (CU) in video frames, judging whether the SKIP mode is a current optimized mode or not if the error of the rebuilt pixel and the original pixel value of each CU unit does not surpass a threshold, or the error of the rebuilt pixel and the original pixel value is distributed in a preset range or predicted residuals do not surpass preset numerical value, judging the SKIP mode as the optimized mode of the current CU if SKIP mode is the current optimized mode, otherwise, judging whether the the current CU is a smallest coding unit (SCU) or not, conducting four quad tree based segmentation on the current CU if the current CU is not an SCU, and conducting mode verification on sub-units one by one. The method effectively reduces coding complexity of a coding side so as to be adapted to mobile equipment and much video application high in real-time requirement.