Other affiliations: University of Kerala
Bio: T. Soumya is an academic researcher from College of Engineering, Trivandrum. The author has contributed to research in topics: Video tracking & Steganalysis. The author has an hindex of 4, co-authored 6 publications receiving 33 citations. Previous affiliations of T. Soumya include University of Kerala.
23 Apr 2015
TL;DR: A daytime coloring approach is proposed to improve the visual perception of night video and shows that edge pixel strength and contrast of the surveillance videos are enhanced compared to other methods.
Abstract: A dark video captured during night surveillance is insufficient to recognize an action. In order to perform various video analysis operations, a night time video enhancement approach is required. A daytime coloring approach is proposed to improve the visual perception of night video. The day image is down sampled and its color features are applied to the night fusion video. The experimental results are compared with context enhancement fusion methods and objective metrics are used to evaluate the performance of the algorithm. The quality measures show that edge pixel strength and contrast of the surveillance videos are enhanced compared to other methods.
TL;DR: A novel night video enhancement scheme based on a hierarchical self-organizing network that automatically groups and enhances the neighboring pixels of dark and light regions in each frame under varied illumination conditions.
Abstract: Night video enhancement techniques are widely used for identifying suspicious activities captured by night visual surveillance systems. However, artificial light sources present in the surroundings deteriorate the visual quality of the video captured during night. This non-uniform illumination reduces the object identification and tracking capability of a real-time visual security system. Thus, a uniform enhancement technique is insufficient for handling such uneven illumination. In this paper, we propose a novel night video enhancement scheme based on a hierarchical self-organizing network. This proposed scheme automatically groups and enhances the neighboring pixels of dark and light regions in each frame. In this scheme, two-level self- organizing neural networks were hierarchically arranged to group similar pixels present in the night video frame. We applied the no-reference-based performance evaluation metrics for measuring the objective quality of the video. The experimental results showed that our proposed approach considerably enhances the visual perception of the video captured at night under varied illumination conditions.
TL;DR: An extensive literature review was conducted and the nighttime visual refinement approaches into nighttime restoration and enhancement were classified and identified the research gap fields to explore future research directions in nighttime visual enhancement techniques.
Abstract: Video surveillance systems substitute manual efforts in various safety critic domains such as border area, assisted living, banking, service stations, and transportation. The multimedia-based surveillance system has a significant role in security and forensic systems because people tend to be easily convinced after observing voice, image, and video. Hence, these videos are strong evidence in the forensic investigation. However, most of the criminal activities such as ATM robbery and assassination are occur at nighttime because of the crime supporting dark environment. Many of the night surveillance systems in military, as well as commercial applications, are equipped with infrared and thermal based night vision systems. Its poor capability of texture and color interpretations are the major issues to ensure secure nighttime video monitoring. Specifically, visual refinements of nighttime surroundings and foreground objects provide a valuable assistance in the nighttime security system. In this scenario, it is highly recommended a review of the state-of-the-art nighttime visual refinement approaches. We conducted an extensive literature review and classified the nighttime visual refinement approaches into nighttime restoration and enhancement. This comparative literary analysis identified the research gap fields to explore future research directions in nighttime visual enhancement techniques. Finally, we discussed various open issues and future directions in the context enhancement based nighttime enhancement research.
TL;DR: A recolorization based night video enhancement to increase the visual perception of surveillance videos by combining day background illumination and tone adjusted night video frames to reduce the darkness of the night video frame.
Abstract: Security surveillance cameras are widely deployed to ensure secure banking, entertainment, and assisted living. Surveillance videos captured by these cameras are considered as forensic evidence for detecting crimes such as ATM robbery and vehicle theft. The videos captured under low lighting conditions are insufficient to identify a theft or robbery happened in the dark regions of a surveillance area. In this paper, we propose a recolorization based night video enhancement to increase the visual perception of surveillance videos. The day background illumination and tone adjusted night video frames are combined to reduce the darkness of the night video frame. Subsequently, chromatic colors of the day image regions are selected corresponding to the dark regions of night frame for the optimization based colorization by using white edge scribbles. The proposed algorithm significantly enhanced the perceptual quality of the video frames compared with existing algorithms. The no-reference based objective evaluation approaches are used for comparing and evaluating the performance of the proposed method with the existing methods. The experimental results indicated that the method improved the visual perception of the night surveillance video compared to the existing methods.
••01 Mar 2020
TL;DR: This paper investigates an overview of the existing methods according to the kind of issue they address, and presents a comparison of the already introduced datasets introduced for the human action recognition field.
Abstract: Within a large range of applications in computer vision, Human Action Recognition has become one of the most attractive research fields. Ambiguities in recognizing actions does not only come from the difficulty to define the motion of body parts, but also from many other challenges related to real world problems such as camera motion, dynamic background, and bad weather conditions. There has been little research work in the real world conditions of human action recognition systems, which encourages us to seriously search in this application domain. Although a plethora of robust approaches have been introduced in the literature, they are still insufficient to fully cover the challenges. To quantitatively and qualitatively compare the performance of these methods, public datasets that present various actions under several conditions and constraints are recorded. In this paper, we investigate an overview of the existing methods according to the kind of issue they address. Moreover, we present a comparison of the existing datasets introduced for the human action recognition field.
TL;DR: A new contrast enhancement algorithm is proposed, which is based on the fact that, for conventional histogram equalization, a uniform input histogram produces an equalized output histogram, and can improve the contrast while preserving original image features.
Abstract: A new contrast enhancement algorithm is proposed, which is based on the fact that, for conventional histogram equalization, a uniform input histogram produces an equalized output histogram. Hence before applying histogram equalization, we modify the input histogram in such a way that it is close to a uniform histogram as well as the original one. Thus, the proposed method can improve the contrast while preserving original image features. The main steps of the new algorithm are adaptive gamma transform, exposure-based histogram splitting, and histogram addition. The object of gamma transform is to restrain histogram spikes to avoid over-enhancement and noise artifacts effect. Histogram splitting is for preserving mean brightness, and histogram addition is used to control histogram pits. Extensive experiments are conducted on 300 test images. The results are evaluated subjectively as well as by DE, PSNR EBCM, GMSD, and MCSD metrics, on which, except for the PSNR, the proposed algorithm has some improvements of 2.89, 9.83, 28.32, and 26.38% over the second best ESIHE algorithm, respectively. That is to say, the overall image quality is better.
TL;DR: Experimental results demonstrate that proposed dehazing technique is able to remove the haze from hazy images as well as significantly improve the image’s visibility.
Abstract: The dehazing problem is an ill-posed and can be regularized by designing an efficient filter to refine the coarse estimated atmospheric veil. The most of existing dehazing techniques suffer from over-saturation, halo artifacts, and gradient reversal artifacts problems. In this paper, a dehazing technique is proposed to remove halo and gradient reversal artifacts problem. In this technique, a notch based integral guided filter is proposed. Moreover, the visibility restoration model is also modified to reduce over-saturation problem. The proposed dehazing technique is compared with seven well-known existing dehazing techniques over ten benchmark hazy images. The experimental results demonstrate that proposed technique is able to remove the haze from hazy images as well as significantly improve the image’s visibility. It also reveals that the restored image has little or no artifacts.
TL;DR: This work implements an improved retinex image enhancement algorithm to enhance the structure layer and uses mask-weighted least squares method to suppress noise and artifact in the texture layer.
Abstract: Nighttime image captured in low- or non-uniform illumination scene always suffers from the loss of visibility and contains various noise and objectionable artifact. When we enlarge the amplitude of the brightness, the noise and artifact will be amplified as well. Hence, we propose a nighttime image enhancement approach based on image decomposition. We decompose the input image into two components: Structure layer contains main information of the image, and texture layer contains details, noise, and artifacts. We implement an improved retinex image enhancement algorithm to enhance the structure layer. To remain details and suppress noise and artifact in the texture layer, we use mask-weighted least squares method. In the final, we fuse these two components to obtain the result. The experimental results demonstrate that the proposed approach can improve the perceptual quality of nighttime images and suppress noise and artifact without excessive reinforcement.