Visual Features Research Articles

MVACLNet: A Multimodal Virtual Augmentation Contrastive Learning Network for Rumor Detection

In today’s digital era, rumors spreading on social media threaten societal stability and individuals’ daily lives, especially multimodal rumors. Hence, there is an urgent need for effective multimodal rumor detection methods. However, existing approaches often overlook the insufficient diversity of multimodal samples in feature space and hidden similarities and differences among multimodal samples. To address such challenges, we propose MVACLNet, a Multimodal Virtual Augmentation Contrastive Learning Network. In MVACLNet, we first design a Hierarchical Textual Feature Extraction (HTFE) module to extract comprehensive textual features from multiple perspectives. Then, we fuse the textual and visual features using a modified cross-attention mechanism, which operates from different perspectives at the feature value level, to obtain authentic multimodal feature representations. Following this, we devise a Virtual Augmentation Contrastive Learning (VACL) module as an auxiliary training module. It leverages ground-truth labels and extra-generated virtual multimodal feature representations to enhance contrastive learning, thus helping capture more crucial similarities and differences among multimodal samples. Meanwhile, it performs a Kullback–Leibler (KL) divergence constraint between predicted probability distributions of the virtual multimodal feature representations and their corresponding virtual labels to help extract more content-invariant multimodal features. Finally, the authentic multimodal feature representations are input into a rumor classifier for detection. Experiments on two real-world datasets demonstrate the effectiveness and superiority of MVACLNet on multimodal rumor detection.

Towards determining perceived audience intent for multimodal social media posts using the theory of reasoned action

Increasing use of social media has resulted in many detrimental effects in youth. With very little control over multimodal content consumed on these platforms and the false narratives conveyed by these multimodal social media postings, such platforms often impact the mental well-being of the users. To reduce these negative effects of multimodal social media content, an important step is to understand creators’ intent behind sharing content and to educate their social network of this intent. Towards this goal, we propose Intent-o-meter, a perceived human intent prediction model for multimodal (image and text) social media posts. Intent-o-meter models ideas from psychology and cognitive modeling literature, in addition to using the visual and textual features for an improved perceived intent prediction model. Intent-o-meter leverages Theory of Reasoned Action (TRA) factoring in (i) the creator’s attitude towards sharing a post, and (ii) the social norm or perception towards the multimodal post in determining the creator’s intention. We also introduce Intentgram, a dataset of 55K social media posts scraped from public Instagram profiles. We compare Intent-o-meter with state-of-the-art intent prediction approaches on four perceived intent prediction datasets, Intentonomy, MDID, MET-Meme, and Intentgram. We observe that leveraging TRA in addition to visual and textual features—as opposed to using only the latter–results in improved prediction accuracy by up to 7.5%\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$7.5\\%$$\\end{document} in Top-1 accuracy and 8%\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$8\\%$$\\end{document} in AUC on Intentgram. In summary, we also develop a web browser application mimicking a popular social media platform and show users social media content overlaid with these intent labels. From our analysis, around 70%\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$70\\%$$\\end{document} users confirmed that tagging posts with intent labels helped them become more aware of the content consumed, and they would be open to experimenting with filtering content based on these labels. However, more extensive user evaluation is required to understand how adding such perceived intent labels mitigate the negative effects of social media.

AI-Empowered Multimodal Hierarchical Graph-Based Learning for Situation Awareness on Enhancing Disaster Responses

Situational awareness (SA) is crucial in disaster response, enhancing the understanding of the environment. Social media, with its extensive user base, offers valuable real-time information for such scenarios. Although SA systems excel in extracting disaster-related details from user-generated content, a common limitation in prior approaches is their emphasis on single-modal extraction rather than embracing multi-modalities. This paper proposed a multimodal hierarchical graph-based situational awareness (MHGSA) system for comprehensive disaster event classification. Specifically, the proposed multimodal hierarchical graph contains nodes representing different disaster events and the features of the event nodes are extracted from the corresponding images and acoustic features. The proposed feature extraction modules with multi-branches for vision and audio features provide hierarchical node features for disaster events of different granularities, aiming to build a coarse-granularity classification task to constrain the model and enhance fine-granularity classification. The relationships between different disaster events in multi-modalities are learned by graph convolutional neural networks to enhance the system’s ability to recognize disaster events, thus enabling the system to fuse complex features of vision and audio. Experimental results illustrate the effectiveness of the proposed visual and audio feature extraction modules in single-modal scenarios. Furthermore, the MHGSA successfully fuses visual and audio features, yielding promising results in disaster event classification tasks.

Application of imaging and spectroscopy techniques for grading of bovine embryos - a review.

Although embryo transfers have grown considerably in the cattle industry, the selection of embryos required for successful pregnancies remains a challenging task. Visual inspection of 7th-day embryos using a stereomicroscope, followed by classification based on morphological features is the most commonly practiced procedure. However, there are inaccuracies and inconsistencies in the manual grading of bovine embryos. The objective of this review was to evaluate the potential of imaging and spectroscopic techniques in the selection of bovine embryos. Digital analysis of microscopic images through extracting visual features in the embryo region, and classification using machine learning methods have yielded about 88-96% success in pregnancies. The Raman spectral pattern provides valuable information regarding developmental stages and quality of the embryo. The Raman spectroscopy approach has also been successfully used to determine various parameters of bovine oocytes. Besides, Fourier Transform Infrared (FTIR) spectroscopy has the ability to assess embryo quality through analyzing embryo composition, including nucleic acid and amides present. Hyperspectral Imaging has also been used to characterize metabolite production during embryo growth. Although the time-lapse imaging approach is beneficial for morphokinetics evaluation of embryo development, optimized protocols are required for successful implementation in bovine embryo transfers. Most imaging and spectroscopic findings are still only at an experimental stage. Further research is warranted to improve the repeatability and practicality to implement in commercial facilities.

Synthetic Media Analysis Using Deep Learning

In an era characterized by the rapid evolution of digital content creation, synthetic media, particularly deepfake videos, present a formidable challenge to the veracity and integrity of online information. Addressing this challenge requires sophisticated analytical techniques capable of discerning between authentic and manipulated media. This research paper presents a comprehensive study on synthetic media analysis leveraging deep learning methodologies. The suggested method integrates advanced deep learning models, including Convolutional Neural Networks (CNNs) like VGG (Visual Geometry Group), alongside recurrent structures such as LSTM (Long Short-Term Memory). These models are trained and evaluated on a meticulously curated dataset, ensuring diversity and relevance in the synthetic media samples. To facilitate experimentation and reproducibility, the dataset is securely hosted on a reliable platform such as Google Drive. Prior to model training, preprocessing steps including frame extraction are employed to extract essential visual features from the video data. The VGG model serves as a feature extractor, capturing high-level representations of visual content, while the LSTM model learns temporal dependencies and contextual information across frames. Following comprehensive experimentation, the proposed method's ability to detect synthetic media is thoroughly assessed, utilizing metrics such as accuracy. This research contributes to the ongoing discourse on digital media forensics by providing insights into the efficacy of deep learning techniques for synthetic media analysis. The findings underscore the importance of continuous research and development in combating the proliferation of synthetic media, thereby safeguarding the authenticity and trustworthiness of online content. Index Terms—CNN, LSTM, VGG

Low-Level Visual Features of Window Views Contribute to Perceived Naturalness and Mental Health Outcomes.

Previous studies have shown that natural window views are beneficial for mental health, but it is still unclear which specific features constitute a 'natural' window view. On the other hand, studies on image analysis found that low-level visual features (LLVFs) are associated with perceived naturalness, but mainly conducted experiments with brief stimulus presentations. In this study, research on the effects of window views on mental health was combined with the detailed analysis of LLVFs. Healthy adults rated window views from their home and sent in photographs of those views for analysis. Content validity of the 'ecological' view assessment was evaluated by checking correlations of LLVFs with window view ratings. Afterwards, it was explored which of the LLVFs best explained variance in perceived percentage of nature and man-made elements, and in ratings of view quality. Criterion validity was tested by investigating which variables were associated with negative affect and impulsive decision-making. The objective and subjective assessments of nature/sky in the view were aligned but objective brightness was unreliable. The perceived percentage of nature was significantly explained by green pixel ratio, while view quality was associated with fractals, saturation, sky pixel ratio and straight edge density. The higher subjective brightness of rooms was associated with a lower negative affect, whereas results for impulsive decision-making were inconsistent. The research highlights the validity to apply LLVFs analysis to ecological window views. For affect, subjective brightness seemed to be more relevant than LLVFs. For impulsive decision-making, performance context needs to be controlled in future studies.

Carved Stone Models as Symbolic Features in the Agrarian Landscape of the Sondondo Valley, Ayacucho, Peru

Carved stone models, or piedras maqueta in Spanish are stone representations of the agrarian landscape widely found in the Valley of Sondondo, in the south-central Peruvian Andes. This article aims to examine the representations, locations, scale, and certain visual features of these carved stone models in the valley. My analysis reveals that carved stone monuments provide a unique documentary source of information on the ancient Andean agrarian landscape of the Valley of Sondondo. The piedras maqueta are situated in different areas of the valley, and their study enables an understanding of their symbolism and ritual significance in relation to the environment, as well as daily agricultural spaces and tasks. Therefore, I explore the sacred landscape as an inseparable part of everyday agricultural practices.

133 The effect of sire lines and endpoint cooking temperature on consumer eating experience and visual appeal in pork loin chops

Abstract Prior research and public perception of pork suggest breed influence visual characteristics like color and size and palatability traits like tenderness and flavor of pork chops. Therefore, the objective of this study was to assess consumer perception of visual appeal and palatability of pork loin chops from three sire lines cooked to two different degrees of doneness (DOD). Boneless pork loins from three sire lines (Duroc, Berkshire, and Synthetic) were aged to 14 d post-mortem. Loin quality was then measured, and 2.54-cm chops were frozen from each loin. Chops from each sire line were overwrap-packaged in sets of two and displayed in a retail case with two representative packages per sire line. Panelists rated each package for color, marbling, and size on a just about right (JAR) scale and then indicated their overall liking of each package. Chops from each sire line were cooked to either 63°C or 71°C, cut into 1-cm cubes and two samples from each chop were served to each consumer. Panelists rated 6 samples, each sire line and DOD, for tenderness, flavor, and juiciness on a JAR scale and then rated overall liking. Data were analyzed as a randomized complete block design (RCBD) in SAS 9.4 with panelist age range serving as an influence effect for all panel data. Berkshire chops did not differ (P > 0.08) in redness from other chops, but required 0.13 kg less force to shear, and had a 16.23 cm2 smaller LEA compared with Synthetic and Duroc chops (P < 0.02). Duroc chops had the greatest marbling (P < 0.04) but did not differ in visual color score from other treatments (P > 0.10). There were no differences in consumer visual evaluations (P > 0.45), with traits from all sire lines averaging between 43 and 57 (JAR = 50). Sire line also did not alter visual liking (P = 0.94). There were tendencies for interactions (P > 0.09) between DOD and sire line for consumer sensory evaluation. In general, consumers rated Berkshire chops cooked to 63°C more favorably than other sire line and DOD combinations. Consumers rated chops with different DOD similarly for tenderness and flavor, but tended to score juiciness of chops cooked to 63°C closer to JAR (P = 0.07). Duroc chops were the least tender and least acceptable (P < 0.05). Berkshire chops were juicer than Duroc and Synthetic chops (P < 0.02). Overall, consumers scored all chops between 3.1 and 12.3 units below the JAR line (100-unit scale) regardless of sire line or DOD for tenderness, juiciness, and flavor with overall acceptability within 10 units of neither like nor dislike, which may be indicative of consumers general negative perceptions of pork loin chops.

TB-Net: Intra- and inter-video correlation learning for continuous sign language recognition

Visual feature extraction is the key to continuous sign language recognition (CSLR). However, current CSLR methods based on single-branch networks only rely on intra-video correlation learning to facilitate visual feature optimization. Obviously, this is not conducive to obtaining more robust visual feature representations. Hence, we pioneered a novel CSLR method via intra- and inter-video correlation learning with a two-branch network, named TB-Net. TB-Net explicitly establishes intra-video correlation between glosses and the most relevant video clips at each branch and then introduces inter-video correlation to enhance visual feature extraction at the branch confluence. Specifically, we introduce a contrastive learning-based inter-video correlation module (IEM), which co-optimizes visual features from both branches by designing inter- and intra-video losses to enhance their generalization. In addition, we propose an intra-video correlation module (IAM) based on a gloss-guided attention feature generator to adaptively build mappings between glosses and video clips, which in turn contributes to the acquisition of preliminary gloss-guided visual features within a single video. Extensive experiments on four public CSLR benchmarks show the superior performance of our method.

Towards Retrieval-Augmented Architectures for Image Captioning

The objective of image captioning models is to bridge the gap between the visual and linguistic modalities by generating natural language descriptions that accurately reflect the content of input images. In recent years, researchers have leveraged deep learning-based models and made advances in the extraction of visual features and the design of multimodal connections to tackle this task. This work presents a novel approach towards developing image captioning models that utilize an external k NN memory to improve the generation process. Specifically, we propose two model variants that incorporate a knowledge retriever component that is based on visual similarities, a differentiable encoder to represent input images, and a k NN-augmented language model to predict tokens based on contextual cues and text retrieved from the external memory. We experimentally validate our approach on COCO and nocaps datasets and demonstrate that incorporating an explicit external memory can significantly enhance the quality of captions, especially with a larger retrieval corpus. This work provides valuable insights into retrieval-augmented captioning models and opens up new avenues for improving image captioning at a larger scale.

A novel deep learning based CBIR model using Convolutional Siamese Neural Networks

Content-Based Image Retrieval (CBIR) is a technique that involves retrieving similar images from a large database by analysing the content features of the query image. The heavy usage of digital platforms and devices has in a way promoted CBIR and its allied technologies in computer vision and artificial intelligence. The process entails comparing the representative features of the query image with those of the images in the dataset to rank them for retrieval. Past research was centered around handcrafted feature descriptors based on traditional visual features. But with the advent of deep learning the traditional manual method of feature engineering gave way to automatic feature extraction. In this study, a cascaded network is utilised for CBIR. In the first stage, the model employs multi-modal features from variational autoencoders and super-pixelated image characteristics to narrow down the search space. In the subsequent stage, an end-to-end deep learning network known as a Convolutional Siamese Neural Network (CSNN) is used. The concept of pseudo-labeling is incorporated to categorise images according to their affinity and similarity with the query image. Using this pseudo-supervised learning approach, this network evaluates the similarity between a query image and available image samples. The Siamese network assigns a similarity score to each target image, and those that surpass a predefined threshold are ranked and retrieved. The suggested CBIR system undergoes testing on a widely recognized public dataset: the Oxford dataset and its performance is measured against cutting-edge image retrieval methods. The findings reveal substantial enhancements in retrieval performance in terms of several standard benchmarks such as average precision, average error rate, average false positive rate etc., providing strong support for utilising images from interconnected devices.

Maize leaf disease recognition using PRF-SVM integration: a breakthrough technique

The difficulty of collecting maize leaf lesion characteristics in an environment that undergoes frequent changes, suffers varying illumination from lighting sources, and is influenced by a variety of other factors makes detecting diseases in maize leaves difficult. It is critical to monitor and identify plant leaf diseases during the initial growing period to take suitable preventative measures. In this work, we propose an automated maize leaf disease recognition system constructed using the PRF-SVM model. The PRFSVM model was constructed by combining three powerful components: PSPNet, ResNet50, and Fuzzy Support Vector Machine (Fuzzy SVM). The combination of PSPNet and ResNet50 not only assures that the model can capture delicate visual features but also allows for end-to-end training for smooth integration. Fuzzy SVM is included as a final classification layer to accommodate the inherent fuzziness and uncertainty in real-world image data. Five different maize crop diseases (common rust, southern rust, grey leaf spot, maydis leaf blight, and turcicum leaf blight along with healthy leaves) are selected from the Plant Village dataset for the algorithm’s evaluation. The average accuracy achieved using the proposed method is approximately 96.67%. The PRFSVM model achieves an average accuracy rating of 96.67% and a mAP value of 0.81, demonstrating the efficacy of our approach for detecting and classifying various forms of maize leaf diseases.

A dynamic neural resource model bridges sensory and working memory

Probing memory of a complex visual image within a few hundred milliseconds after its disappearance reveals significantly greater fidelity of recall than if the probe is delayed by as little as a second. Classically interpreted, the former taps into a detailed but rapidly decaying visual sensory or ‘iconic’ memory (IM), while the latter relies on capacity-limited but comparatively stable visual working memory (VWM). While iconic decay and VWM capacity have been extensively studied independently, currently no single framework quantitatively accounts for the dynamics of memory fidelity over these time scales. Here, we extend a stationary neural population model of VWM with a temporal dimension, incorporating rapid sensory-driven accumulation of activity encoding each visual feature in memory, and a slower accumulation of internal error that causes memorized features to randomly drift over time. Instead of facilitating read-out from an independent sensory store, an early cue benefits recall by lifting the effective limit on VWM signal strength imposed when multiple items compete for representation, allowing memory for the cued item to be supplemented with information from the decaying sensory trace. Empirical measurements of human recall dynamics validate these predictions while excluding alternative model architectures. A key conclusion is that differences in capacity classically thought to distinguish IM and VWM are in fact contingent upon a single resource-limited WM store.

A dynamic neural resource model bridges sensory and working memory.

Probing memory of a complex visual image within a few hundred milliseconds after its disappearance reveals significantly greater fidelity of recall than if the probe is delayed by as little as a second. Classically interpreted, the former taps into a detailed but rapidly decaying visual sensory or 'iconic' memory (IM), while the latter relies on capacity-limited but comparatively stable visual working memory (VWM). While iconic decay and VWM capacity have been extensively studied independently, currently no single framework quantitatively accounts for the dynamics of memory fidelity over these time scales. Here, we extend a stationary neural population model of VWM with a temporal dimension, incorporating rapid sensory-driven accumulation of activity encoding each visual feature in memory, and a slower accumulation of internal error that causes memorized features to randomly drift over time. Instead of facilitating read-out from an independent sensory store, an early cue benefits recall by lifting the effective limit on VWM signal strength imposed when multiple items compete for representation, allowing memory for the cued item to be supplemented with information from the decaying sensory trace. Empirical measurements of human recall dynamics validate these predictions while excluding alternative model architectures. A key conclusion is that differences in capacity classically thought to distinguish IM and VWM are in fact contingent upon a single resource-limited WM store.

Highly memorable images are more readily perceived.

Image memorability, the likelihood that a person will remember a particular image, has been shown to be an intrinsic property of the image that is distinct from many other visual and cognitive features. Research thus far has not identified particular visual features that can sufficiently explain this intrinsic memorability, but one possibility is that more and less memorable images differ in their statistical regularity (i.e., how prototypical or distinctive they are). Statistical regularity is known to affect detection time for images, such that stimuli with higher statistical regularity can be detected with shorter presentation directions. Therefore, in the present study, we probed whether memorability affects how quickly an image can be detected. High- and low-memorability images were presented in an intact/scrambled task wherein participants were asked to indicate whether they saw an intact image or noise, and we estimated the presentation duration necessary for participants to reach 70.7% accuracy. Across two experiments using different stimulus materials, we observed and then replicated that more memorable images are associated with shorter detection thresholds than those for less memorable images. The results support the idea that memorable stimuli may better match stored templates used for image perception and/or recognition. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

The influence of electrode types to the visually induced gamma oscillations in mouse primary visual cortex.

The local field potential (LFP) is an extracellular electrical signal associated with neural ensemble input and dendritic signaling. Previous studies have linked gamma band oscillations of the LFP in cortical circuits to sensory stimuli encoding, attention, memory, and perception. Inconsistent results regarding gamma tuning for visual features were reported, but it remains unclear whether these discrepancies are due to variations in electrode properties. Specifically, the surface area and impedance of the electrode are important characteristics in LFP recording. To comprehensively address these issues, we conducted an electrophysiological study in the V1 region of lightly anesthetized mice using two types of electrodes: one with higher impedance (1MΩ) and a sharp tip (10μm), while the other had lower impedance (100 KΩ) but a thicker tip (200μm). Our findings demonstrate that gamma oscillations acquired by sharp-tip electrodes were significantly stronger than those obtained from thick-tip electrodes. Regarding size tuning, most gamma power exhibited surround suppression at larger gratings when recorded from sharp-tip electrodes. However, the majority showed enhanced gamma power at larger gratings when recorded from thick-tip electrodes. Therefore, our study suggests that microelectrode parameters play a significant role in accurately recording gamma oscillations and responsive tuning to sensory stimuli.

Context Disentangling and Prototype Inheriting for Robust Visual Grounding.

Visual grounding (VG) aims to locate a specific target in an image based on a given language query. The discriminative information from context is important for distinguishing the target from other objects, particularly for the targets that have the same category as others. However, most previous methods underestimate such information. Moreover, they are usually designed for the standard scene (without any novel object), which limits their generalization to the open-vocabulary scene. In this paper, we propose a novel framework with context disentangling and prototype inheriting for robust visual grounding to handle both scenes. Specifically, the context disentangling disentangles the referent and context features, which achieves better discrimination between them. The prototype inheriting inherits the prototypes discovered from the disentangled visual features by a prototype bank to fully utilize the seen data, especially for the open-vocabulary scene. The fused features, obtained by leveraging Hadamard product on disentangled linguistic and visual features of prototypes to avoid sharp adjusting the importance between the two types of features, are then attached with a special token and feed to a vision Transformer encoder for bounding box regression. Extensive experiments are conducted on both standard and open-vocabulary scenes. The performance comparisons indicate that our method outperforms the state-of-the-art methods in both scenarios.

Preparation and properties of micro-arc oxidation ceramic coatings using machine vision technology

In this paper, a machine vision system is combined with non-immersion micro-arc oxidation (MAO) technology, using color as a visual feature, and automatically determining the MAO process through a self-developed color image segmentation program for HSV channel. The robotic arm accurately regulates the machining position and duration. Dense ceramic oxide coatings of similar quality to the artificially prepared coatings were successfully prepared on aluminum (Al) surface. The surface morphology and elemental composition of the samples were characterized using scanning electron microscopy (SEM), energy dispersive spectrometry (EDS) and X-ray diffraction (XRD). The performance of the coatings was tested through thickness tests, hardness tests, and roughness tests. The tribological properties were tested by friction experiments, and its corrosion resistance was tested by electrochemical corrosion tests. The results confirm the effectiveness of a machine vision system for MAO process control, which is capable of preparing high-quality coatings comparable to manual operations. This not only solves the need for precise film formation in different application scenarios, but also provides a more efficient, controllable and innovative solution for the MAO process.

Solving Bongard Problems With a Visual Language and Pragmatic Constraints.

More than 50 years ago, Bongard introduced 100 visual concept learning problems as a challenge for artificial vision systems. These problems are now known as Bongard problems. Although they are well known in cognitive science and artificial intelligence, only very little progress has been made toward building systems that can solve a substantial subset of them. In the system presented here, visual features are extracted through image processing and then translated into a symbolic visual vocabulary. We introduce a formal language that allows representing compositional visual concepts based on this vocabulary. Using this language and Bayesian inference, concepts can be induced from the examples that are provided in each problem. We find a reasonable agreement between the concepts with high posterior probability and the solutions formulated by Bongard himself for a subset of 35 problems. While this approach is far from solving Bongard problems like humans, it does considerably better than previous approaches. We discuss the issues we encountered while developing this system and their continuing relevance for understanding visual cognition. For instance, contrary to other concept learning problems, the examples are not random in Bongard problems; instead they are carefully chosen to ensure that the concept can be induced, and we found it helpful to take the resulting pragmatic constraints intoaccount.

Learning multi-granularity semantic interactive representation for joint low-light image enhancement and super-resolution

Images captured in challenging conditions often suffer from the co-existence of low contrast and low resolution. However, most joint enhancement methods focus on fitting a direct mapping from degraded images to high-quality images, which proves insufficient to handle complex degradation. To mitigate this, we propose a novel semantic prior guided interactive network (MSIRNet) to enable effective image representation learning for joint low-light enhancement and super-resolution. Specifically, a local HE-based domain transfer strategy is developed to remedy the domain gap between low-light images and the recognition scope of a generic segmentation model, thereby obtaining a rich granularity of semantic prior. To represent hybrid-scale features with semantic attributes, we propose a multi-grained semantic progressive interaction module that formulates an omnidirectional blend self-attention mechanism, facilitating deep interaction between diverse semantic knowledge and visual features. Moreover, employing our feature normalized complementary module that perceives context and cross-feature relationships, MSIRNet adaptively integrates image features with the auxiliary visual atoms provided by the Codebook, endowing the model with high-fidelity reconstruction capability. Extensive experiments demonstrate the superior performance of our MSIRNet, showing its ability to restore visually and perceptually pleasing normal-light high-resolution images.