Matches in SemOpenAlex for { <https://semopenalex.org/work/W4385575008> ?p ?o ?g. }
- W4385575008 endingPage "15" @default.
- W4385575008 startingPage "1" @default.
- W4385575008 abstract "Image forgery detection and localization has become a research topic of increasing interest with the enormous spread of manipulated images on the Web. Most previous methods related to image forgery focus on one particular attribute or neglect the importance of multi-scale information. Such approaches are considered not suitable for detecting and locating multi-scale splicing forgeries. This paper presents a novel approach that exploits residual attention and integrates multi-scale local and global information to improve detection accuracy. In the proposed method, we first aggregate multi-level convolutional feature maps extracted by the encoder to enrich the feature representations and improve the ability of the model to locate multi-scale forged areas. Then, we design a residual attention block (RAB) to purify the features, which enhances the response of task-related regions and suppresses noise information. Furthermore, a global feature mining block (GFMB) is proposed to capture the long-range dependencies between different regions of the image, enabling the model to handle complex tampering scenarios effectively. The multi-scale splicing forgery regions are precisely detected and located by utilizing the proposed method. The extensive experiments are conducted on three benchmark datasets, CASIA, COLUMB, and NIST’16. Specifically, our model achieves the F1 score of 84.3%, 87.9%, and 80.8% on CASIA, COLUMB, and NIST’16 test sets, respectively, outperforming state-of-the-art methods." @default.
- W4385575008 created "2023-08-05" @default.
- W4385575008 creator A5010938432 @default.
- W4385575008 creator A5020596965 @default.
- W4385575008 creator A5027078757 @default.
- W4385575008 creator A5090838052 @default.
- W4385575008 date "2023-01-01" @default.
- W4385575008 modified "2023-10-14" @default.
- W4385575008 title "Multi-scale Attention Network for Detection and Localization of Image Splicing Forgery" @default.
- W4385575008 cites W1484985793 @default.
- W4385575008 cites W1903029394 @default.
- W4385575008 cites W1985729543 @default.
- W4385575008 cites W1987786716 @default.
- W4385575008 cites W2071794886 @default.
- W4385575008 cites W2096076665 @default.
- W4385575008 cites W2096754397 @default.
- W4385575008 cites W2104116677 @default.
- W4385575008 cites W2108017094 @default.
- W4385575008 cites W2142694135 @default.
- W4385575008 cites W2169439181 @default.
- W4385575008 cites W2281375574 @default.
- W4385575008 cites W2412782625 @default.
- W4385575008 cites W2515389788 @default.
- W4385575008 cites W2560026874 @default.
- W4385575008 cites W2572561073 @default.
- W4385575008 cites W2752015292 @default.
- W4385575008 cites W2752782242 @default.
- W4385575008 cites W2777769595 @default.
- W4385575008 cites W2802179504 @default.
- W4385575008 cites W2802701183 @default.
- W4385575008 cites W2911605501 @default.
- W4385575008 cites W2912648825 @default.
- W4385575008 cites W2948407537 @default.
- W4385575008 cites W2955058313 @default.
- W4385575008 cites W2963091558 @default.
- W4385575008 cites W2964146055 @default.
- W4385575008 cites W2965545961 @default.
- W4385575008 cites W2975393663 @default.
- W4385575008 cites W2981689412 @default.
- W4385575008 cites W3014806201 @default.
- W4385575008 cites W3133303693 @default.
- W4385575008 cites W3137069976 @default.
- W4385575008 cites W3163631282 @default.
- W4385575008 cites W3168491317 @default.
- W4385575008 cites W3170841864 @default.
- W4385575008 cites W3172522282 @default.
- W4385575008 cites W3194371117 @default.
- W4385575008 cites W3204245437 @default.
- W4385575008 cites W3204810842 @default.
- W4385575008 cites W3211551288 @default.
- W4385575008 cites W4206670089 @default.
- W4385575008 cites W4213350031 @default.
- W4385575008 cites W4214893857 @default.
- W4385575008 cites W4220911046 @default.
- W4385575008 cites W4226043703 @default.
- W4385575008 cites W4282822664 @default.
- W4385575008 cites W4311088269 @default.
- W4385575008 doi "https://doi.org/10.1109/tim.2023.3300434" @default.
- W4385575008 hasPublicationYear "2023" @default.
- W4385575008 type Work @default.
- W4385575008 citedByCount "0" @default.
- W4385575008 crossrefType "journal-article" @default.
- W4385575008 hasAuthorship W4385575008A5010938432 @default.
- W4385575008 hasAuthorship W4385575008A5020596965 @default.
- W4385575008 hasAuthorship W4385575008A5027078757 @default.
- W4385575008 hasAuthorship W4385575008A5090838052 @default.
- W4385575008 hasConcept C111219384 @default.
- W4385575008 hasConcept C11413529 @default.
- W4385575008 hasConcept C115961682 @default.
- W4385575008 hasConcept C121332964 @default.
- W4385575008 hasConcept C124101348 @default.
- W4385575008 hasConcept C13280743 @default.
- W4385575008 hasConcept C138885662 @default.
- W4385575008 hasConcept C153180895 @default.
- W4385575008 hasConcept C154945302 @default.
- W4385575008 hasConcept C155512373 @default.
- W4385575008 hasConcept C185798385 @default.
- W4385575008 hasConcept C205649164 @default.
- W4385575008 hasConcept C2524010 @default.
- W4385575008 hasConcept C2776401178 @default.
- W4385575008 hasConcept C2777210771 @default.
- W4385575008 hasConcept C2778755073 @default.
- W4385575008 hasConcept C28490314 @default.
- W4385575008 hasConcept C31972630 @default.
- W4385575008 hasConcept C33923547 @default.
- W4385575008 hasConcept C41008148 @default.
- W4385575008 hasConcept C41895202 @default.
- W4385575008 hasConcept C52622490 @default.
- W4385575008 hasConcept C62520636 @default.
- W4385575008 hasConcept C99498987 @default.
- W4385575008 hasConceptScore W4385575008C111219384 @default.
- W4385575008 hasConceptScore W4385575008C11413529 @default.
- W4385575008 hasConceptScore W4385575008C115961682 @default.
- W4385575008 hasConceptScore W4385575008C121332964 @default.
- W4385575008 hasConceptScore W4385575008C124101348 @default.
- W4385575008 hasConceptScore W4385575008C13280743 @default.
- W4385575008 hasConceptScore W4385575008C138885662 @default.
- W4385575008 hasConceptScore W4385575008C153180895 @default.