Abstract: Embodiments of the present disclosure provide a method and device for selecting a context model of a quantized coefficient end flag. The method comprises: obtaining a scanning position POS of a non-zero coefficient corresponding to current quantized coefficient end flag in a specific scanning order; wherein the scanning position POS is a subscript of the non-zero coefficient in the scanning order; configuring a first context model array, and using a fixed value as the base to calculate the logarithmic value of the value of the scanning position POS plus 1, and according to the logarithmic value, selecting a first context model from the first context model array; and using the first context model to encode or decode a binary symbol of the current quantized coefficient end flag. According to the technical solution of the present application, it is able to improve encoding and decoding efficiency for a quantized coefficient end flag, thereby further improve the efficiency of video encoding and decoding.

Abstract: The present disclosure discloses an inverse tone mapping method, system, device and computer readable medium The method of embodiment of the present application comprises: decomposing the original image into an illumination component and a reflection component, wherein the illumination component represents a global illumination condition of the image, the reflection component representing a color and texture detail of the image; recovering the illumination component to obtain a result of illumination component recovery; recovering the reflection component to obtain a result of reflection component recovery; combining the result of the illumination component recovery and the result of the reflection component recovery to obtain a recovery result image. Compared with the prior art, the inverse tone mapping method according to the embodiment of the present invention can greatly improve the effect of the image recovery.

Abstract: The present application provides methods, systems, devices and computer-readable mediums for encoding and decoding transform. A method of the present application comprises: performing transform-encoding on a prediction residual block separately using a plurality of transform matrices, to obtain transformed residual blocks; based on the transformed residual block, determining, by a rate-distortion optimization decision, a transform matrix or a combination of transform matrices matching the residual characteristics of the prediction residual block from the plurality of transform matrices, and determining a transformed residual block to be outputted in a bitstream. Compared with the prior art, the method of the embodiments of the present invention performs a residual transform using transform matrices that are more closely matched with the residual characteristic, thereby improving the expression of the residual signal and improving the coding efficiency of the residual block.

Abstract: An image feature extraction method for person re-identification includes performing person re-identification by means of aligned local descriptor extraction and graded global feature extraction; performing the aligned local descriptor extraction by processing an original image by affine transformation and performing a summation pooling operation on image block features of same regions to obtain an aligned local descriptor; reserving spatial information between inner blocks of the image for the aligned local descriptor; and performing the graded global feature extraction by grading a positioned pedestrian region block and solving a corresponding feature mean value to obtain a global feature. The method can resolve the problem of feature misalignment caused by posture changes of pedestrian, etc., and eliminate the effect of unrelated backgrounds on re-recognition, thus improving the precision and robustness of person re-identification.

Abstract: A back-propagation significance detection method based on depth map mining, comprising: for an input image Io, at a preprocessing phase, obtaining a depth image Id and an image Cb with four background corners removed of the image Io; at a first processing phase, carrying out positioning detection on a significant region of the image by means of the obtained image Cb with four background corners removed and the obtained depth image Id to obtain the preliminary detection result S1 of a significant object in the image; then carrying out depth mining on a plurality of processing phases of the depth image Id to obtain corresponding significance detection results; and then optimizing the significance detection result mined in each processing phase by means of a back-propagation mechanism to obtain a final significance detection result map. The method can improve the detection accuracy of the significance object.

Abstract: The application discloses a method, system, device and computer-readable storage medium for inverse quantization, wherein, in some embodiments, determining whether the inverse transform coefficient corresponding to the quantized coefficient in the quantized block can be set to 0, according to the size of the quantized block; when the inverse transform coefficient corresponding to the quantized coefficient can't be set to 0, the quantized coefficient is inverse quantized to obtain a corresponding inverse transform coefficient.

Abstract: An embodiment of the present disclosure discloses an intra-prediction-based encoding and decoding method, apparatus and filter, and belongs to a video coding technical field.

Abstract: The present application provides methods, systems, devices and computer-readable mediums for deblocking filter. A method of the present application comprises: determining a filtering boundary, and then determining a filter pixel group based on the filtering boundary; determining a filter strength of the filter pixel group, comprising: parsing separately a pixel value difference states of pixel points on both sides of the filtering boundary in the filter pixel group to obtain two one-sided flatness FL and FR; calculating a comprehensive flatness FS of the filter pixel group, wherein FS=FL+FR; calculating the filter strength according to FS; filtering calculation of pixel points included in the filter pixel group according to the filter strength.

Abstract: The present specification discloses a method, apparatus, and device for video frame interpolation.

Abstract: The embodiment of the present specification discloses an image synthesis method, apparatus and device for free-viewpoint.

Abstract: A back-propagation significance detection method based on depth map mining, comprising: for an input image Io, at a preprocessing phase, obtaining a depth image Id and an image Cb with four background corners removed of the image Io; at a first processing phase, carrying out positioning detection on a significant region of the image by means of the obtained image Cb with four background corners removed and the obtained depth image Id to obtain the preliminary detection result S1 of a significant object in the image; then carrying out depth mining on a plurality of processing phases of the depth image Id to obtain corresponding significance detection results; and then optimizing the significance detection result mined in each processing phase by means of a back-propagation mechanism to obtain a final significance detection result map. The method can improve the detection accuracy of the significance object.

Abstract: A bidirectional image-text retrieval method based on a multi-view joint embedding space includes: performing retrieval with reference to a semantic association relationship at a global level and a local level, obtaining the semantic association relationship at the global level and the local level in a frame-sentence view and a region-phrase view, and obtaining semantic association information in a global level subspace of frame and sentence in the frame-sentence view, obtaining semantic association information in a local level subspace of region and phrase in the region-phrase view, processing data by a dual-branch neural network in the two views to obtain an isomorphic feature and embedding the same in a common space, and using a constraint condition to reserve an original semantic relationship of the data during training, and merging the two semantic association relationships using multi-view merging and sorting to obtain a more accurate semantic similarity between data.

Abstract: Disclosed is a deep discriminative network for person re-identification in an image or a video. Concatenation are carried out on different input images on a color channel by constructing a deep discriminative network, and an obtained splicing result is defined as an original difference space of different images. The original difference space is sent into a convolutional network. The network outputs the similarity between two input images by learning difference information in the original difference space, thereby realizing person re-identification. The features of an individual image are not learnt, and concatenation are carried out on input images on a color channel at the beginning, and difference information is learnt on an original space of the images by using a designed network. By introducing an Inception module and embedding the same into a model, the learning ability of a network can be improved, and a better differentiation effect can be achieved.

Abstract: The present application discloses a cross-media retrieval method based on deep semantic space, which includes a feature generation stage and a semantic space learning stage. In the feature generation stage, a CNN visual feature vector and an LSTM language description vector of an image are generated by simulating a perception process of a person for the image; and topic information about a text is explored by using an LDA topic model, thus extracting an LDA text topic vector. In the semantic space learning phase, a training set image is trained to obtain a four-layer Multi-Sensory Fusion Deep Neural Network, and a training set text is trained to obtain a three-layer text semantic network, respectively. Finally, a test image and a text are respectively mapped to an isomorphic semantic space by using two networks, so as to realize cross-media retrieval. The disclosed method can significantly improve the performance of cross-media retrieval.

Abstract: The present application provides a method and a device of encoding and decoding based on free viewpoint, and relates to the technical field of video encoding. The method includes: generating a planar splicing image and splice information based on multiple single-viewpoint videos at a server side; generating a planar splicing video based on the planar splicing image; generating camera side information of the planar splicing video based on camera side information existing in the multiple single-viewpoint videos; and encoding the planar splicing video, the splice information and the camera side information of the planar splicing video to generate a planar splicing video bit stream, and decoding a planar splicing video bit stream to acquire a virtual viewpoint according to viewpoint information of a viewer at client side.

Abstract: The present disclosure provides a hybrid framework-based image bit-depth expansion method and device. The invention fuses a traditional de-banding algorithm and a depth network-based learning algorithm, and can remove unnatural effects in an image flat area whilst more realistically restoring numerical information of missing bits. The method comprises the extraction of image flat areas, local adaptive pixel value adjustment-based flat area bit-depth expansion and convolutional neural network-based non-flat area bit-depth expansion. The present invention uses a learning-based method to train an effective depth network to solve the problem of realistically restoring missing bits, whilst using a simple and robust local adaptive pixel value adjustment method in an flat area to effectively inhibit unnatural effects in the flat area such as banding, a ringing and flat noise, improving subjective visual quality of the flat area.

Abstract: The present application provides methods, systems, devices and computer-readable mediums for deblocking filter. A method of the present application comprises: determining a filtering boundary, and then determining a filter pixel group based on the filtering boundary; determining a filter strength of the filter pixel group, comprising: parsing separately a pixel value difference states of pixel points on both sides of the filtering boundary in the filter pixel group to obtain two one-sided flatness FL and FR; calculating a comprehensive flatness FS of the filter pixel group, wherein FS=FL+FR; calculating the filter strength according to FS; filtering calculation of pixel points included in the filter pixel group according to the filter strength.

Abstract: An intra-frame and inter-frame combined prediction method for P frames or B frames. The method comprises: self-adaptively selecting by means of a rate-distortion optimization (RDO) decision whether to use the intra-frame and inter-frame combined prediction or not; using a method for weighting an intra prediction block and an inter prediction block in the intra-frame and inter-frame combined prediction to obtain a final prediction block; and obtaining the weighting coefficient of the intra prediction block and the inter prediction block according to prediction distortion statistics of the prediction method. Therefore, prediction precision can be improved, and coding and decoding efficiency of the prediction blocks are improved.

Abstract: Disclosed are a panoramic video asymmetrical mapping method and a corresponding inverse mapping method: by means of the mapping methods, mapping a spherical surface corresponding to a panoramic image or video A onto a two-dimensional image or video B; projecting the spherical surface onto an isosceles quadrangular pyramid with a square bottom plane, and further projecting the isosceles quadrangular pyramid onto a planar surface; using isometric projection on a main viewpoint region in the projection and using a relatively high sampling density to ensure that the video quality of the region of the main viewpoint is high, while using a relatively low sample density for non-main viewpoint regions so as to reduce bit rate.

Abstract: Disclosed is a method for detecting pedestrians in an image by using Gaussian penalty. Initial pedestrian boundary box is screened using a Gaussian penalty, to improve the pedestrian detection performance, especially sheltered pedestrians in an image. The method includes acquiring a training data set, a test data set and pedestrian labels of a pedestrian detection image; using the training data set for training to obtain a detection model by using a pedestrian detection method, and acquiring initial pedestrian boundary box and confidence degrees and coordinates thereof; performing Gaussian penalty on the confidence degrees of the pedestrian boundary box, to obtain confidence degree of the pedestrian boundary box after the penalty; and obtaining final pedestrian boundary boxes by screening the pedestrian boundary boxes. Thus, repeated boundary boxes of a single pedestrian are removed while reserving boundary boxes of sheltered pedestrians, thereby realizing the detection of the pedestrians in an image.