Face recognition: using neural spatial-temporal segmentation to extract face features and perform feature geometry comparison

To assess the echogenicity of the cervical stroma and area surrounding the cervical canal (cervical gland area - CGA) using grey scale histogram (GSH) in pregnancies between 20 and 25 weeks. Cross-sectional study involving 149 pregnancies in the second trimester. Transvaginal sonography was performed to measure cervical length (biometric evaluation), assess the CGA and detect cervical funnelling (morphologic evaluation). The GSH was used for the objective assessment of the cervical stroma and CGA in the mid portion of the cervix. Mean, minimum, maximum and standard deviation measurements were obtained on the GSH and the CGA/stroma ratio was calculated. Intra-observer and inter-observer reproducibility were assessed using the intraclass correlation coefficient (ICC). Mean cervical length was 42.37 ± 7.05 mm (range 12-54 mm). Funnelling was detected in four women (2.7%) and one of these also had absent CGA. The mean GSH stroma value was 103 ± 25 (53-160) and the mean CGA value was 64.5 ± 23 (13-167). The mean CGA/stroma ratio was 0.62 ± 0.14 (0.24-1.29). Intra- and inter-observer reproducibility was excellent with ICC of 0.975 and 0.922 respectively. There are large differences in the texture of the cervical stroma and CGA on GSH allowing the objective differentiation of the two areas. The CGA/stroma ratio allows the objective assessment of the absence or presence of the CGA.

The majority of color histogram equalization methods do not yield uniform histogram in gray scale. After converting a color histogram equalized image into gray scale, the contrast of the converted image is worse than that of an 1-D gray scale histogram equalized image. We propose a novel 3-D color histogram equalization method that produces uniform distribution in gray scale histogram by defining a new cumulative probability density function in 3-D color space. Test results with natural and synthetic images are presented to compare and analyze various color histogram equalization algorithms based upon 3-D color histograms. We also present theoretical analysis for nonideal performance of existing methods.

Objective: to assess the classification capability dealing with CT images, by means of Decision Tree classifier. Methods: the CT images were provided by the hospital was used as the data source. In this study.normal liver, Single cystic hepatic hydatid and Polycystic hepatic hydatid CT images were selected.each 200 pieces. Then the texture features were extracted by gray gradient co-occurrence matrix(GGCM) and gray scale histogram. At last. Decision tree classifier were used for classification, which aimed to verify Which feature is more suitable for decision tree classification, and Parameter estimation is used to evaluate the classifier model. Results: For normal liver, Single cystic hepatic hydatid and Polycystic hepatic hydatid, the classification accuracy of gray level co-occurrence matrix was 71%,69% and 69%, respectively. the classification accuracy of gray scale histogram was 74%,63.5% and 69%.the classification accuracy of comprehensive features was 75%, 70.5% and 80.5%. Conclusion: The classification accuracy of comprehensive feature is 11.5% higher than that of single features, which are more suitable for the classification of polycystic hepatic hydatid images. This algorithm can provide reference for the classification of CT images.

Three hundred and eighteen grey scale histograms of thyroid glands in 121 patients have been analysed. The arithmetic average grey scale value (GWMi) and the most commonly appearing grey value (GWMa) were determined electronically. The normal range of GWMi was 24 +/- 4.6 (x +/- s) grey scale units (GWE), for GWMa it was 26 +/- 5.6 GWe (x +/- s). Macrofollicular thyroid adenomas (8 cases) and thyroids with diffuse autonomy (12 cases) differed with a GWMi of 21 +/- 3.8 and 23 +/- GEW (x +/- s) respectively and a GWMa of 23 +/- 3.9 and 25 +/- 4.8 GWE and were within the normal range (p greater than 0.05). Microfollicular thyroid adenomas (12 cases), immunological thyroid lesions (25 cases) and parathyroid adenomas (6 cases) had reduced GWMi (6 +/- 3.1 and 8 +/- 1.8 GWE respectively) and GWMa (5 +/- 3 and 7 +/- 2 GWE, p less than 0.005). Thyroid carcinomas (9 cases) and lymph node metastases (8 cases) were also significantly below normal, with a GWMi of 11 +/- 2.3 and 11 +/- 3.4 respectively (p less than 0.05). Analysis of grey scale histograms makes it possible to evaluate changes in the thyroid without reference to surrounding tissues.

It is generally agreed that faces are not recognized only by utilizing some holistic search among all learned faces, but also through a feature analysis that aimed to specify more important features of each specific face. This paper addresses a novel decision strategy that efficiently uses both holistic and facial component (eye, nose and mouth) feature analysis to recognize faces. The proposed algorithm first performs a holistic search using the whole image features and selects probable candidates for further processing. Then the facial features of these probable candidates are compared and classified with the stored patterns in the train set. Finally the classification results are fused with a weighted majority voting to form the final decision. Simulation studies justify the superior performance of the proposed method as compared to that of Eigenface method.

Face recognition under large illumination variations using homomorphic filtering in spatial domain

Computed tomography analysis on cyclic fatigue and damage properties of rock salt under gas pressure

Face images obtained by an outdoor surveillance camera, are often confronted with severe degradations (e.g., low-resolution, low-contrast, blur and noise). This significantly limits the performance of face recognition (FR) systems. This paper presents a framework to overcome the degradation in images obtained by an outdoor surveillance camera, to improve the performance of FR. We have defined a measure that is based on the difference in intensity histograms of face images, to estimate the amount of degradation. In the past, super-resolution techniques have been proposed to increase the image resolution for face recognition. In this work, we attempt a combination of partial restoration (using super-resolution, interpolation etc.) of probe samples (long distance shots of outdoor) and simulated degradation of gallery samples (indoor shots). Due to the unavailability of any benchmark face database with gallery and probe images, we have built our own database and conducted experiments on a realistic surveillance face database. PCA and FLDA have been used as baseline face recognition classifiers. The aim is to illustrate the effectiveness of our proposed method of compensating the degradation in surveillance data, rather than designing a specific classifier space suited for degraded test probes. The efficiency of the method is shown by improvement in the face classification accuracy, while comparing results obtained separately using training with acquired indoor gallery samples and then testing with the outdoor probes.KeywordsDegradationFace databaseFace recognitionImage qualitySurveillanceSuper-resolution

We present a new efficient face recognition using two level adaptive classifier scheme. Face recognition is a very popular problem in computer vision area. However, the maturity of face recognition technology still need much progress since its performance is depending on several varying factors such as facial illumination, expression, pose etc. In this paper, we propose an adaptive classifier in order to reduce effect of facial feature variations due to the change of facial expressions. Facial features are eyes, nose, and mouth feature groups each of which is constituted of facial feature points. All facial images are treated by first initial evolutionary classifier using initial feature group level and feature point level parameter set. All rejected facial images based on threshold are employed for generating new feature parameter set using two level chromosome evolutions. The feature space adopted here is Gabor wavelet with weight Gabor parameters. We have performed extensive experiments using international standard FERET face image dataset and we have achieved very encouraging results.

A comparative study on local binary pattern (LBP) based face recognition: LBP histogram versus LBP image

Diagnosis of Laryngopharyngeal Reflux Disease Based on Gray and Texture Changes of Laryngoscopic Images

Fatigue driving can result in delayed driver reactions and lack of concentration, increasing the risk of traffic accidents. Therefore, finding effective fatigue detection methods is crucial for accident prevention and ensuring public safety. Traditional driver fatigue detection methods based on facial features are limited by their reliance on manually designed features and classifiers, leading to low accuracy, slow detection speed, and inefficient performance. Hence, methods based on deep learning have gradually become a trend in fatigue-driving detection research. This paper focuses on driver fatigue detection methods based on facial features using deep learning. In this study, we analyze numerous research outcomes on how to accurately extract, analyze, and integrate various facial features, such as eye features, mouth features, facial expressions, and head postures, to precisely determine the driver's fatigue state. Then, we compare the latest deep learning-based technologies. This paper provides an information-rich repository to help researchers select more promising designs, that can drive the widespread development of fatigue-driving detection technologies.

Most of the appearance based visual feature extraction methods in the lip-reading system treat the mouth image in a whole manner. However, the vision of speech process is three dimensional and treating the mouth image as a whole may lose the speech information. Motivated by the bidirectional PCA (BDPCA) and decomposition methods used in the face recognition domain, in this paper, a modular bidirectional PCA (MBDPCA) based visual feature extraction method was presented. In this method, the original mouth image sequences are divided into smaller sub-images, and two approaches are compared to build the covariance matrix: one is using all the sub-image sets together to build a global covariance matrix; the other is using the different sub-image sets independently to build the local covariance matrices. Then the BDPCA is applied to each sub-image set. Experimental results show that the MBDPCA method has a better performance than both the conventional PCA and BDPCA methods; moreover, further experimental results demonstrate that our lip-reading system provides significant enhancement of robustness in noisy environments compared to the audio-only speech recognition.

The exponential growth of user-generated video content necessitates efficient summarization systems for improved accessibility, retrieval, and analysis. This study presents and benchmarks a multimodal video summarization framework that classifies segments as informative or non-informative using audio, visual, and fused features. Sixty hours of annotated video across ten diverse categories were analyzed. Audio features were extracted with pyAudioAnalysis, while visual features (colour histograms, optical flow, object detection, facial recognition) were derived using OpenCV. Six supervised classifiers—Naive Bayes, K-Nearest Neighbors, Logistic Regression, Decision Tree, Random Forest, and XGBoost—were evaluated, with hyperparameters optimized via grid search. Temporal coherence was enhanced using median filtering. Random Forest achieved the best performance, with 74% AUC on fused features and a 3% F1-score gain after post-processing. Spectral flux, grayscale histograms, and optical flow emerged as key discriminative features. The best model was deployed as a practical web service using TensorFlow and Flask, integrating informative segment detection with subtitle generation via beam search to ensure coherence and coverage. System-level evaluation demonstrated low latency and efficient resource utilization under load. Overall, the results confirm the strength of multimodal fusion and ensemble learning for video summarization and highlight their potential for real-world applications in surveillance, digital archiving, and online education.

Two experiments test the effects of exposure duration and encoding instruction on the relative memory for five facial features. Participants viewed slides of Identi‐kit faces and were later given a recognition test with same or changed versions of each face. Each changed test face involved a change in one facial feature: hair, eyes, chin, nose or mouth. In both experiments the upper‐face features of hair and eyes were better recognized than the lower‐face features of nose, mouth, and chin, as measured by false alarm rates. In Experiment 1, participants in the 20‐second exposure duration condition remembered faces significantly better than participants in the 3‐second exposure duration condition; however, memory for all five facial features improved at a similar rate with the increased duration. In Experiment 2, participants directed to use feature scanning encoding instructions remembered faces significantly better than participants following age judgement instructions; however, the size of the memory advantage for upper facial features was less with feature scanning instructions than with age judgement instructions. The results are discussed in terms of a quantitative difference in processing faces with longer exposure duration, versus a qualitative difference in processing faces with various encoding instructions. These results are related to conditions that affect the accuracy of eyewitness identification.