Abstract: An endoscopic image segmentation method based on a single image and a deep learning network achieves real-time and accurate segmentation for a single case, and provides support for making a treatment plan based on an endoscopic image in clinical medicine. The endoscopic image segmentation method first proposes a single image-based training set generation method to automatically generate a training set, and further proposes a lightweight deep learning network EUnet to perform feature fitting on the generated training set, to achieve endoscopic image segmentation and obtain a segmentation result of a lesion region. The endoscopic image segmentation method can significantly improve segmentation accuracy and has advantages of a small volume, high real-time performance, and easy operation.

Abstract: The present application discloses a frontpage news prediction and classification method. Keywords to be queried are firstly input by means of a user interface, and collected news text information on the web pages is saved in a local database; a text representation module performs vector representation by using a Doc2Vec representation algorithm, so as to convert each news text into a low-dimensional text feature vector with a high amount of information; and a similarity network construction module calculates the similarity between news, constructs a news similarity network by taking a calculated similarity matrix as an adjacent matrix of a news related network, determines whether a similarity network is traversed, if so, iteratively calculates an HR value of the vector according to an H-index supporting contribution matrix, performs weight sorting on the news by using the HR value, and predicts top-N pieces of news as front page news.

Abstract: The present application relates to the technical field of semiconductor, and in particular to an inverter. The inverter provided by the present application includes a substrate, a discrete device and a heat conducting component. The discrete device and the heat conducting component are both arranged on the substrate. A part of the heat conducting component is located in an area of the substrate where the discrete device is provided, and another part of the heat conducting component is located in an area of the substrate where the discrete device is not provided. The heat conducting component may rapidly transfer the heat of the overheated area of the substrate where the discrete device is mounted to the less hot area of the substrate, and promote the heat generated by the discrete device to spread evenly to the substrate.

Abstract: The present application discloses a frontpage news prediction and classification method. Keywords to be queried are firstly input by means of a user interface, and collected news text information on the web pages is saved in a local database; a text representation module performs vector representation by using a Doc2Vec representation algorithm, so as to convert each news text into a low-dimensional text feature vector with a high amount of information; and a similarity network construction module calculates the similarity between news, constructs a news similarity network by taking a calculated similarity matrix as an adjacent matrix of a news related network, determines whether a similarity network is traversed, if so, iteratively calculates an HR value of the vector according to an H-index supporting contribution matrix, performs weight sorting on the news by using the HR value, and predicts top-N pieces of news as front page news.

Abstract: The present application relates to the technical field of semiconductor, and in particular to an inverter. The inverter provided by the present application includes a substrate, a discrete device and a heat conducting component. The discrete device and the heat conducting component are both arranged on the substrate. A part of the heat conducting component is located in an area of the substrate where the discrete device is provided, and another part of the heat conducting component is located in an area of the substrate where the discrete device is not provided. The heat conducting component may rapidly transfer the heat of the overheated area of the substrate where the discrete device is mounted to the less hot area of the substrate, and promote the heat generated by the discrete device to spread evenly to the substrate.

Abstract: A deblocking filtering method and a deblocking filter are provided. The method includes: determining, according to a current pixel value of a to-be-filtered pixel in a picture block and pixel values of pixels in a first neighbouring partition of the to-be-filtered pixel, a first filtering offset value of the to-be-filtered pixel; performing prediction on the pixel value of the to-be-filtered pixel according to pixel values of pixels in a second neighbouring partition of the to-be-filtered pixel to obtain a prediction value of filtered value of the to-be-filtered pixel; and determining an actual filtered value of the to-be-filtered pixel according to the current pixel value, the first filtering offset value, and the prediction value of filtered value.

Abstract: An intra-frame encoding method, an intra-frame decoding method, an encoder, and a decoder, where the intra-frame encoding method includes obtaining reference pixel values of a current picture block obtaining a predicted value of a first-type pixel in the current picture block according to the reference pixel values of the current picture block using an intra-frame prediction algorithm, obtaining residual values of the first-type pixel, obtaining transform coefficients according to the residual values of the first-type pixel, quantizing the transform coefficients, reconstructing the first-type pixel according to the quantized transform coefficients and the predicted value of the first-type pixel, and obtaining a reconstructed value of the second-type pixel according to the reconstructed value of the first-type pixel using an interpolation method. Hence, flexible encoding can be implemented, and encoding efficiency can be improved.

Abstract: A flyback power converter includes a transformer, a power switch, a power switch control circuit, a synchronous rectification (SR) switch, an SR switch control circuit, and a signal coupler circuit. The signal coupler circuit includes a primary port and a secondary port, wherein the primary port is electrically connected to the power switch control circuit, and the secondary port is electrically connected to the SR switch control circuit. The primary port and the secondary port receive different signals generated by the power switch control circuit and the SR switch control circuit respectively, and the signal coupler circuit senses and converts the different signals to generate corresponding converted signals at the secondary port and the primary port respectively in different and non-overlapping time periods, without direct contact or direct connection between the primary side and the secondary side of the transformer.

Abstract: A flyback power converter includes a transformer, a power switch, a power switch control circuit, a synchronous rectification (SR) switch, an SR switch control circuit, and a signal coupler circuit. The signal coupler circuit includes a primary port and a secondary port, wherein the primary port is electrically connected to the power switch control circuit, and the secondary port is electrically connected to the SR switch control circuit. The primary port and the secondary port receive different signals generated by the power switch control circuit and the SR switch control circuit respectively, and the signal coupler circuit senses and converts the different signals to generate corresponding converted signals at the secondary port and the primary port respectively in different and non-overlapping time periods, without direct contact or direct connection between the primary side and the secondary side of the transformer.

Abstract: The present disclosure provide a deblocking filtering method and a deblocking filter. The method includes: determining, according to a current pixel value of a to-be-filtered pixel in a picture block and pixel values of pixels in a first neighbouring partition of the to-be-filtered pixel, a first filtering offset value of the to-be-filtered pixel; performing prediction on the pixel value of the to-be-filtered pixel according to pixel values of pixels in a second neighbouring partition of the to-be-filtered pixel to obtain a prediction value of filtered value of the to-be-filtered pixel; and determining an actual filtered value of the to-be-filtered pixel according to the current pixel value, the first filtering offset value, and the prediction value of filtered value.

Abstract: An intra-frame encoding method, an intra-frame decoding method, an encoder, and a decoder, where the intra-frame encoding method includes obtaining reference pixel values of a current picture block obtaining a predicted value of a first-type pixel in the current picture block according to the reference pixel values of the current picture block using an intra-frame prediction algorithm, obtaining residual values of the first-type pixel, obtaining transform coefficients according to the residual values of the first-type pixel, quantizing the transform coefficients, reconstructing the first-type pixel according to the quantized transform coefficients and the predicted value of the first-type pixel, and obtaining a reconstructed value of the second-type pixel according to the reconstructed value of the first-type pixel using an interpolation method. Hence, flexible encoding can be implemented, and encoding efficiency can be improved.

Abstract: Aspects of the subject matter described herein relate to image restoration for compressed images. In aspects, image restoration is accomplished by recovering spectral information from data corresponding to a compressed image. The spectral information is recovered using an algorithm to search through a solution space of possible solutions while constraints are imposed on the solution space to trim undesirable solutions from the space. An algorithm described herein may be iteratively applied to improve the quality of the recovered image.

Abstract: Aspects of the subject matter described herein relate to image restoration for compressed images. In aspects, image restoration is accomplished by recovering spectral information from data corresponding to a compressed image. The spectral information is recovered using an algorithm to search through a solution space of possible solutions while constraints are imposed on the solution space to trim undesirable solutions from the space. An algorithm described herein may be iteratively applied to improve the quality of the recovered image.

Abstract: A motion-compensated video encoding scheme employs progressive fine-granularity layered coding to encode macroblocks of video data into frames having multiple layers, including a base layer of comparatively low quality video and multiple enhancement layers of increasingly higher quality video. Some of the enhancement layers in a current frame are predicted from different quality layers in reference frames. The video encoding scheme estimates drifting errors during the encoding and chooses a coding mode for each macroblock in the enhancement layer to maximize high coding efficiency while minimizing drifting errors.

Abstract: A video encoding system performs packetization of FGS/PFGS encoded video bitstreams by selecting encoded bitstream segments for packetization based on an estimated total contribution to distortion reduction associated with each encoded bitstream segment. The selected bitstream segments are then packetized according to a packet-independence packetization strategy that minimizes inter-packet dependency.

Abstract: It is known to represent a square block of an image which has a maximum horizontal or vertical extension N as N vectors each representing one column or row, and encode it using an N×N discrete cosine transform matrix. We introduce a technique which transforms an input vector, or array, representing a portion of an arbitrarily-shaped object block in which the maximum horizontal or vertical extension is at most N, by adding to it additional padding elements, to produce a vector or array which can be encoded using an N×N discrete cosine transform. The padding maximises the number of transformed elements which are zero. We propose encoding arbitrarily-shaped objects based on this padding technique.