16results about How to "Improve invisibility" patented technology

The invention relates to a face depth forgery active defense method and device based on discrete wavelet transform, and relates to the technical field of image processing, and the method comprises the steps: carrying out the attribute editing of an original image and a disturbance-containing image through a depth forgery model, calculating the mean square error loss and structural similarity loss between the original image and the disturbance-containing image, different weights are endowed and fused into a total loss function; gradient information is obtained based on loss function back propagation, the optimal direction of disturbance application is determined, and an intermediate disturbance value is formed; low-frequency and high-frequency components are embedded into corresponding components of the disturbance-containing image through discrete wavelet transform, and then the image is reconstructed through inverse discrete wavelet transform; checking whether the number of iterations is reached or not, and if not, introducing a JPEG noise layer to continuously iterate and adjust disturbance; and if yes, outputting a final face image containing invisible disturbance information. According to the method, the robustness in the wavelet domain is fully utilized, and the resistance of disturbance to JPEG compression is improved on the premise of ensuring that the disturbance is invisible.

The invention provides a multi-position overlapping image privacy protection method and system based on compressed sensing, and relates to the technical field of compressed sensing and image privacy protection, and the method comprises the steps: obtaining an original image, and dividing privacy parts and non-privacy parts of different parts of the original image; generating a coincident privacy mask for the coincident privacy part, generating a privacy mask for the non-coincident privacy part, confusing all the privacy parts, and performing image reconstruction by using the confused privacy part and the original non-privacy part; sampling encryption of compressed sensing is carried out on the reconstructed image, a compressed sensing ciphertext is generated, corresponding watermark matrixes and watermark embedding matrixes are generated for different privacy parts according to a confusion matrix, the watermark embedding matrixes and the watermark matrixes are embedded into the compressed sensing ciphertext, and encryption of the compressed sensing ciphertext is achieved; and users with different levels of keys are utilized to decode and reconstruct different parts of images of the encrypted compressed sensing ciphertext, so that privacy protection of multi-position overlapped images is realized.

The invention relates to the technical field of digital copyright protection and information security, in particular to a dark watermark generation system and method, and the system comprises an input data module, a robust feature extraction module, a self-adaptive dark watermark generation module, a deep learning embedding strategy module, and a dark watermark embedding module. Wherein the input data module is used for receiving digital content to be protected; the robust feature extraction module is used for generating feature parameters for embedding the dark watermark and determining a robust feature position; the self-adaptive dark watermark generation module is used for mapping the dark watermark information into a characteristic parameter space to generate self-adaptive dark watermark data; the embedding strategy module is used for generating a dark watermark embedding strategy; and the dark watermark embedding module is used for embedding the dark watermark data into the robust feature position of the preprocessed digital content. According to the invention, through combination of robust feature extraction, adaptive watermark generation and a deep learning embedding strategy, high precision and anti-interference performance of watermark embedding are realized.

The invention relates to the technical field of image processing, and discloses a wavelet transform domain dual watermarking method based on image interpolation, which comprises the following steps of: firstly, improving embedding capacity by utilizing improved image interpolation; and then a watermark is ingeniously embedded in a low-frequency region of the original image after three-level transformation by using multilayer decomposition and local characteristics of wavelet transformation through a calculation method of changing a wavelet coefficient to realize a protection effect on image copyright. And secondly, carrying out three-level wavelet transformation on the carrier image embedded with the watermark on the basis, extracting coefficient correlation of a low-frequency part as an authentication watermark, and embedding the authentication watermark into a medium-high-frequency part of the two-level transformation to realize a protection function of image content and successfully locate a tampered position. Compared with the prior art, the method has high robustness and invisibility, successfully solves the problem that most watermark algorithms are single in function, and achieves balance between the robustness and the invisibility.

The application provides a watermark embedding and extracting method, device, equipment, storage medium and product. The watermark embedding method comprises: performing block processing on an image to be embedded with a watermark to obtain a plurality of image blocks, wherein the plurality of image blocks comprise a first number of first image blocks and a second number of second image blocks, and target feature parameters of the first image blocks and the second image blocks are different; performing transformation processing on the plurality of image blocks respectively to obtain transformation domain features of the plurality of image blocks, wherein the transformation domain features comprise a low-frequency subband and a plurality of high-frequency directional subbands; and based on the transformation domain features of the plurality of image blocks, watermark information is embedded into target low-frequency subbands of the first number of first image blocks and target high-frequency directional subbands of the second number of second image blocks to obtain an image containing a watermark.

This invention discloses an image watermarking method and system based on self-supervised visual model-guided decoding. The method includes acquiring watermark information *m* and the original image *I* from each data sample in the training dataset; for each data sample, preprocessing the watermark information *m* to generate a message tensor *M*, and performing degradation distortion processing on the original image *I* to generate a distorted guided heatmap *DGH*, thus obtaining a preprocessed training dataset; and using the preprocessed training dataset to perform self-supervised training on a self-supervised visual model to finally obtain a trained self-supervised visual model. This invention aims to solve the problem that existing anti-printing watermarking technologies overly rely on location masks, resulting in poor decoding robustness when mask prediction is inaccurate. It utilizes the semantic understanding capability of deep visual models to replace traditional geometric synchronization, achieving end-to-end, correction-free robust decoding.

The present invention relates to a vehicle comprising: a vehicle glass comprising a transparent region and an opaque region; a transparent dielectric substrate disposed in the transparent region of the vehicle glass; a first region including an antenna on one side of the transparent dielectric substrate; and a second region including a feed pattern electrically connected to the antenna and a ground conductive pattern. The antenna may include: a first slot formed between a signal pattern and a first ground pattern and configured to radiate a circularly polarized first signal having a first frequency band; and a second slot formed between the first ground pattern and the second ground pattern and configured to radiate a circularly polarized second signal having a second frequency band lower than the first frequency band.

This application provides a digital video watermarking method, apparatus, device, and storage medium, belonging to the field of artificial intelligence technology. The method includes: acquiring and merging training video data (channel dimension information and time dimension information) and original watermark information, inputting them into an original digital video watermarking network for encoding to obtain original watermarked video data; adding noise data to the original watermarked video data to obtain noisy watermarked video data; inputting the noisy watermarked video data into the original digital video watermarking network for decoding to obtain predicted watermark information; constructing target loss data based on the training video data, original watermarked video data, predicted watermark information, and original watermark information, adjusting the original digital video watermarking network to obtain a target digital video watermarking network; and implanting a watermark into the target video data through the target digital video watermarking network to obtain target watermarked video data. This application embodiment can construct a video watermarking method with high invisibility and high robustness.

The utility model discloses an ultra-invisible ankle sock which comprises a sock body. The sock body is integrated; the front end of the sock body is provided with a first opening for the big toe to pass through in a position corresponding to the big toe; a second opening through which the little toe passes is formed in the part corresponding to the little toe; a third opening allowing the heel to penetrate through is formed in the rear end of the sock body. The super-invisible ankle sock is good in invisible effect, and materials can be saved.

The utility model relates to a wig with a hidden label, which comprises wig pieces and the label, a plurality of wig pieces are sequentially arranged along the thickness direction, two adjacent wig pieces are fixedly connected, and the label is completely positioned between the two adjacent wig pieces. The identification on the label records product information, so that the problem that the product information is difficult to trace after an existing wig product is worn is solved; the label is located between the wig pieces, the wig pieces shield the label, the label is prevented from being observed, and the invisible effect of the wig pieces is improved; the wig piece prevents the label from being in contact with a user, foreign body sensation caused by the label is avoided, and the label cannot cause reduction of wearing comfort; according to the wig piece, friction between the label and a user is avoided, label abrasion is reduced, and identification of the label is protected.

The present invention discloses a robust watermarking method for screen-captured images based on an attention mechanism and contrastive learning, including: using an encoder to generate an encoded image containing watermark information; inputting the encoded image and a carrier image into a discriminator, and using the discriminator to output a prediction value; simulating the distortion of the encoded image; inputting the distorted encoded image into a decoder, and extracting the watermark information hidden in the distorted encoded image; training the encoder, the discriminator, and the decoder according to the prediction value of the discriminator and the joint loss function, using the trained encoder and decoder to form a screen-captured watermarking model, and using the screen-captured watermarking model to encode and decode the screen-captured image. The present invention optimizes the process of watermark image encoding, while ensuring the invisibility of the encoded image, effectively enhancing the robustness of the watermark model in real-life scenarios; and better guarantees the integrity of the watermark information when facing screen-captured noise.

The invention discloses a mouse trap, which comprises a main body for bearing a pull rope mechanism; the main body comprises a top plate and a support for supporting the top plate; the pull rope mechanism is in transmission connection with a sling for catching mice; the pull rope mechanism is connected with a sensor through a control circuit, and the pull rope mechanism tightens the sling to lock the mouse on the main body when the sensor detects a mouse signal; the part of the sling under the top plate forms a frame shape so that the sling is invisible to the mouse passing through the lower part of the top plate. The invention improves the bulky electric control net cover technology in the existing invisible mouse trap into the sling technology with extremely small volume, greatly reduces the volume and production cost of the invisible mouse trap, improves the invisibility of the mouse trap, occupies very small area and has low requirement on the flatness of the mouse catching site, improves the usability of the invisible mouse trap in the complex and narrow environment outdoors and in the wild, is more convenient to use in the precise position to save the mouse catching time, has good mouse catching effect, and is also suitable for catching various wild mice living in the wild.

The invention discloses a high-concealment image watermarking method based on Hadamard calculation ghost imaging, and the method comprises the specific steps: obtaining a light intensity sequence # imgabs 1 # of an original watermark image # imgabs 0 #, and rearranging the light intensity sequence # imgabs 2 # into a new watermark image # imgabs 3 #; chaos scrambling is carried out on the new watermark image # imgabs4 #, and a real watermark image # imgabs5 # is obtained; the method comprises the following steps of: decomposing a host image # imgabs6 # into four sub-bands based on Haar wavelet transform, namely a low-frequency sub-band # imgabs7 #, a horizontal high-pass vertical low-pass sub-band # imgabs8 #, a horizontal low-pass vertical high-pass sub-band # imgabs9 # and a diagonal high-pass sub-band # imgabs10 #; the method comprises the following steps: partitioning a diagonal high-pass sub-band # imgabs11 #, sorting all blocks, selecting middle blocks, and combining the middle blocks into a reference image # imgabs12 #; the real watermark image # imgabs13 # is embedded into the reference image # imgabs14 #, and a reference image # imgabs15 # embedded with the watermark is obtained; three interference images which are randomly selected are respectively embedded into a sub-band # imgabs16 #, a sub-band # imgabs17 # and a sub-band # imgabs18 #; and a host image # imgabs19 # embedded with a watermark is obtained through inverse transformation of Haar wavelets. According to the method, the imaging quality of the calculation ghost imaging technology is improved, and high robustness is achieved.

The utility model belongs to the field of building decoration, and provides a wooden door sleeve and a wooden door, the wooden door comprises a door leaf and the wooden door sleeve, the wooden door sleeve comprises a main sleeve piece, an auxiliary sleeve plate and an auxiliary sleeve line, the main sleeve piece comprises a main sleeve plate and a main sleeve line, and the main sleeve line and the main sleeve plate are integrally formed; the main sleeve plate is provided with a first inner side face, the main sleeve line is provided with a second inner side face, and the second inner side face is flush with the first inner side face. The auxiliary sleeve plate is arranged between the main sleeve plate and the auxiliary sleeve line, the auxiliary sleeve plate is connected with the main sleeve plate, and the auxiliary sleeve line is connected with the auxiliary sleeve plate. According to the wooden door pocket, the main pocket line and the main pocket plate are integrally formed to form the main pocket piece, in this way, when the wooden door pocket is installed, the main pocket piece is installed on the wall of a doorway, installation of the main pocket plate and the main pocket line can be completed synchronously, and the number of installation steps of the wooden door pocket is reduced; in addition, the second inner side face of the main sleeve line is flush with the first inner side face of the main sleeve plate, and when the door leaf is arranged on the wooden door sleeve and is in a closed state, the main sleeve plate can be shielded.

The invention discloses furniture with a novel three-in-one connecting piece. The furniture comprises a first component, a second component, an eccentric wheel, a connecting rod, an expansion body and an operating piece. The eccentric wheel, the connecting rod, the operating piece and the expansion body are used for connecting a first component and a second component of furniture; one end of the connecting rod protrudes to form a connecting boss, the other end of the connecting rod is provided with a conical part, the connecting rod is fixedly provided with a driving external thread part between the connecting boss and the conical part, the driving external thread part is connected with an internal thread of the operating piece in a matched mode, and a connecting groove is formed between the operating piece and the connecting boss; a conical channel is formed in the expansion body, and the operating part is used for being driven by a sleeve tool so that the conical part can slide by a certain distance in the conical channel, the fastening boss can expand outwards in the radial direction of the expansion body, and the connecting rod can be locked in a first mounting hole of a first component through the expansion body; after furniture finished products are connected according to the novel three-in-one connecting piece, the invisibility can be improved.

The invention belongs to the technical field of digital image watermarking, and provides a robust watermarking method based on multi-scale attention to solve the problem that an existing deep learning digital image robust watermarking scheme is insufficient in performance under attacks of JPEG, cutting and the like. An encoder-noise layer-decoder structure is adopted, and the encoder realizes feature extraction and cross-group fusion by designing an MSCAF module; meanwhile, constructing a space attention mask layer to guide the watermark to adaptively adjust the embedding strength according to the image features; the decoder uses the MSCAF to extract the multi-scale features so as to improve the watermark extraction precision. Experimental results show that under JPEG attack training of different quality factors, the PSNR and the watermark extraction error rate of the coded image are both superior to those of an existing comparison algorithm; under combined noise attack training, the average peak signal-to-noise ratio of a coded image generated by the method reaches 38.6738 dB, the watermark extraction error rate is lower than 0.6%, and the method is superior to an existing comparison algorithm.