An Analysis of Contrast Enhancement using Activation Functions

Contrast Enhancement is a technique which comes into the part of Image Enhancement. Contrast Enhancement is used to enhance the visual quality of any captured or other image. Contrast Enhancement can be performed with the help of Histogram equalization (HE). In this technique, the image is collected in the gray scale allocation. The image is then partitioning and applying adaptive Histogram equalization (AHE). Fuzzy logic provides a set of logics which enhance the contrast and visibility of any image. In this technique, the visual quality and the contrast of image will change and then compare these results with previous techniques. The performance of several established image enhancement techniques is presented in terms of different parameters like Absolute mean brightness error (AMBE), Peak signal to noise ratio (PSNR), contrast and Visual quality.

A new approach based on masking technique for contrast enhancement of medical image is presented due to the low contrast characteristics. In medical image, preserving mean brightness, average information and noise factor reduction are essential to make input image more appealing visually. The proposed method incorporates spatial and frequency domain techniques to enhance the contrast of the medical image. The mask is formulated effectively between reconstructed approximation coefficients and inverse singular value decomposition (ISVD) to obtain contrast residual. Discrete wavelet transformation (DWT) and singular value decomposition (SVD) have used to decompose the input image. At last maximum contrast enhancement achieved by adding mask with the image obtained through intensity exposure histogram equalization (IEHE). The proposed approach is tested for medical images by comparing its peak signal to noise Ratio (PSNR), absolute mean brightness error (AMBE) and entropy with some existing methods and also measured in terms of visual quality.

Contrast enhancement and brightness preservation of the image are two important issues of image enhancement in research field now-a-days. The objective is to enhance the image uniformly over different parts of the image. General Histogram Equalization doesn’t control degree of enhancement of the image. To overcome this drawback, another variant of Histogram Equalization method namely Adaptive Bi-histogram Equalization using Threshold (ABHET) is being proposed. The proposed method undergoes three steps, such as: Histogram segmentation using threshold, Clipping of histogram using mean value of occupied intensity and histogram equalization of each sub-images. Finally all the sub-images are combined into one complete image. Simulation results show that ABHET outperforms other existing HE-based methods and different image quality measures such as: Peak signal to noise ratio (PSNR), Absolute Mean Brightness Error (AMBE) and Structural Similarity Index (SSIM) are being used to test the robustness of the proposed method in terms of enhancement of contrast and preservation of brightness.

Conventional contrast enhancement techniques often fail to produce satisfactory results for low-contrast images, and cannot be automatically applied to different images because their processing parameters must be specified manually to produce a satisfactory result for a given image. This work presents a colour-preserving contrast enhancement (CPCE) algorithm for images. Modification to images was performed in the HSV colour-space. The Hue component is preserved (unchanged), luminance modified using Contrast Limited Adaptive Histogram Equalization (CLAHE), while Saturation components were up-scaled using a derived mapping function on the approximate components of its discrete wavelet transform. Implementation was done in MATLAB and compared with CLAHE and Histogram Equalization (HE) algorithms in the RGB colour space. Subjective (visual quality inspection) and objective parameters (Peak-signal-to-noise ratio (PSNR), Absolute Mean Brightness Error (AMBE) and Mean squared error (MSE)) were used for performance evaluation. The method produced images with the lowest MSE, AMBE, and highest PSNR when tested, yet preserved the visual quality of the image.

This research mainly based on multilayer perceptron (MLP) neural networks technique of data mining to forecast the wheat crop yield at the district level. There are many statistical and simulation models available, but the proposed algorithm with new activation function provides promising results in a shorter time with more accuracy. Sigmoid and hyperbolic tangent activation functions are widely used in the neural network. The activation functions play an important role in the neural network learning algorithm. The main objective of the proposed work is to develop an amended MLP neural network with new activation function and revised random weights and bias values for crop yield estimation by using the different weather parameter datasets. MLP model has been tested by existing activation functions and newly created activation functions with different cases including weights and bias values. In this research study, we evaluate the result of different activation functions and recommend some new simple activation functions, named DharaSig, DharaSigm and SHBSig, to improve the performance of neural networks and accurate results. Also, three new activation functions created with little variations in the DharaSig function named DharaSig1, DharaSig2 and DharaSig3. In this research study, variable numbers of hidden layers are tested with the variable number of neurons per hidden layer for the agriculture dataset. Variable values of momentum, seed and learning rate are also used in this study. Experiments show that newly created activation functions provide better results compared to ‘sigmoid’ default neural network activation function for agriculture datasets.

Histogram Equalization (HE) is a simple, effective and widely used contrast enhancement technique as it can automatically define the intensity transformation function based on statistical characteristics of the image, but it tends to change the mean brightness of the image to the middle level of the gray level range. HE also produces saturation effects by extremely pushing the intensities towards the right or the left side of the histogram. To surmount these drawbacks, Clipping or Plateau Histogram Equalization techniques for brightness preserving and contrast enhancement have been proposed, but, these are not suitable for automatic systems because of manual selection of threshold level. Self-Adaptive Plateau Histogram Equalization (SAPHE) selects the threshold level automatically, but the process is relatively complicated and sometimes fails in execution. To overcome these drawbacks, a Modified Self - Adaptive Plateau Histogram Equalization with Mean threshold (Modified SAPHE-M) is proposed in this paper and compared the experimental results with Histogram Equalization (HE), Self-Adaptive Plateau Histogram Equalization (SAPHE) and Modified Self-Adaptive Plateau Histogram Equalization (Modified SAPHE) by using image quality measures such as Absolute Mean Brightness Error (AMBE) and Peak-Signal to Noise Ratio (PSNR).

In the field of image processing, improving low-contrast images is essential for enhancing the visibility of features that may be hard to see due to slight intensity variations. This issue often arises with images captured in poor lighting or unfavorable conditions, which can impede tasks such as feature extraction, object recognition, and edge detection. This paper presents an algorithm that employs fuzzy techniques to enhance low-contrast images. The approach utilizes a Pythagorean fuzzy set, which is an advanced version of a fuzzy set. The Pythagorean fuzzy set, by integrating membership and non−membership degrees, offers a more precise representation of image uncertainty. The proposed method constructs an increasing function to calculate the degree of non−membership, thereby amplifying the effectiveness of the enhanced image. By integrating this method with contrast limited adaptive histogram equalization, alongside a color restoration technique and brightness-preserving dynamic histogram equalization, the brightness and natural appearance of the original low-contrast image are maintained. We compare our approach to other methods, including conventional histogram equalization, intuitionistic fuzzy generators, interval−valued intuitionistic fuzzy generators, interval−valued intuitionistic fuzzy with contrast limited adaptive histogram equalization, LightNet, and FlightNet. Experimental results indicate that our method surpasses the techniques currently utilized in terms of visibility and color accuracy. Performance evaluation is performed using metrics such as the absolute mean brightness error, contrast improvement index, and structural similarity index.

In recurrent neural networks such as the long short-term memory (LSTM), the sigmoid and hyperbolic tangent functions are commonly used as activation functions in the network units. Other activation functions developed for the neural networks are not thoroughly analyzed in LSTMs. While many researchers have adopted LSTM networks for classification tasks, no comprehensive study is available on the choice of activation functions for the gates in these networks. In this paper, we compare 23 different kinds of activation functions in a basic LSTM network with a single hidden layer. Performance of different activation functions and different number of LSTM blocks in the hidden layer are analyzed for classification of records in the IMDB, Movie Review, and MNIST data sets. The quantitative results on all data sets demonstrate that the least average error is achieved with the Elliott activation function and its modifications. Specifically, this family of functions exhibits better results than the sigmoid activation function which is popular in LSTM networks.

An artificial neural network(ANN) is a type of computing system that mimics the functioning of the neural networks in a biological brain. In this paper, an ANN has been simulated on an Field Programmable Gate Array (FPGA) for the application of handwritten digit recognition using the sigmoid and hyperbolic tangent(tanh) activation functions. Both software and hardware simulations have been carried out, using python and Verilog HDL respectively. The two activation functions were implemented on an Artix-7 FPGA. A comparison has been made between the sigmoid and tanh activation functions based on speed, accuracy, and hardware required, and it has been inferred that the tanh function is best for the application of handwritten digit recognition as it has 3% higher accuracy and uses 5 less LUTs than the sigmoid activation function.

Structure-water solubility modeling of aliphatic alcohols was performed using the multifunctional autocorrelation method. The molecule is represented by using a set of parameters describing global molecules, and others that take the structural environment of the edge O-C into account. Multiple linear regression (MLR) and multilayer feed-forward artificial neural network architectures are utilized to construct linear and nonlinear QSPR models, respectively. The optimal QSPR model was developed based on a 4-4-1 neural network architecture. The efficiency of the approach is demonstrated through the predictive ability of the ANN and MLR models by the leave-20%-out (L20%O) cross-validation method, demonstrating that the neural model is more reliable than that obtained using MLR. The root mean square errors in the solubility prediction (ln SOL) for the calibration and predictive models were 0.13 and 0.18 respectively. On the other hand, we tested four activation functions: the hyperbolic tangent, sigmoid function or Gaussian functions for the hidden layer and a linear, sigmoid, hyperbolic tangent or Gaussian function for the output layer. The influence and the contribution of each type of descriptor in the model is examined. After omission of a set of descriptors, we calculate the error for the solubility and classify them into discrete categories. The standard error and the percentage of the prediction in the precision interval considered have been estimated. The results imply that the solubility of aliphatic alcohols is dominated by the shape and branching of the molecule. The hydrogen-bonding interactions caused by the C-OH group seem to be a less important factor influencing the solubility. The model was compared with other models; especially that using weighted path numbers, which is considered to be the most accurate QSPR model for predicting the water solubility of aliphatic alcohols.

The remote sensing images acquired from the satellites are low contrast images. The availability of low contrast images and failure of the traditional methods such as Histogram Equalization and Gamma correction in preserving the brightness levels in the image are the main issues in satellite image processing. This paper proposes an optimized contrast stretching using non-linear transformation for image enhancement. The non-linear transformation is influenced by the appropriate choice of control parameters for the sample images since manual tuning for individual images is tedious. A Bat algorithm based tuning is employed for the automated selection of control parameters in the transformation. The performance of the optimization algorithm is compared against other metaheuristic algorithms such as Ant Colony Optimization (ACO) and Particle Swarm Optimization (PSO). It is noted that the bat algorithm based contrast enhancement outperforms the other optimization techniques in terms of metrics such as Peak Signal to Noise Ratio (PSNR), Mean Square Error (MSE), Absolute Mean Brightness Error (AMBE), Entropy and CPU time (Central Processing Unit).

This paper shows that the performance of the Hopfield network for solving optimization problems can be improved by using a new activation (output) function. The effects of the activation function on the performance of the Hopfield network are analyzed. It is shown that the sigmoid activation function in the Hopfield network is sensitive to noise of neurons. The reason is that the sigmoid function is most sensitive in the range where noise is most predominant. A new activation function that is more robust against noise is proposed. The new activation function has the capability of amplifying the signals between neurons while suppressing noise. The performance of the new activation function is evaluated through simulation. Compared with the sigmoid function, the new activation function reduces the error rate of tour length by 30.6% and increases the percentage of valid tours by 38.6% during simulation on 200 randomly generated city distributions of the 10-city traveling salesman problem.

In this paper, we propose a new architecture of Pulse Mode Neural Network (PMNN) with very simple activation function. Pulse mode is gaining support in the field of hardware Neural Networks thanks to its higher density of integration. However, the complexity of the activation functions presents a drawback for hardware implementation of Neural Networks and limits its area of application. In this context, the main idea is to apply a new kind of activation function, simply generated by the product of two sigmoidal functions, which are very simple and already implemented in previous work. Details of important aspects concerning the hardware implementation are given. To verify the performance and capacity of the proposed design, we apply it for approximation of image denoising function. The filtered results are verified in terms of the Peak Signal to Noise Ratio (PSNR). Experimental results reveal that the proposed PMNN filter has a greater ability to recover the informative pixel intensities from the infected image with a recovery of 7.5 dB for Gaussian noise and 5.3 dB for Speckle noise. Besides, such results demonstrate the performance and efficiency of our Neural filter when compared to other conventional filtering techniques. The designed network is implemented on a field-programmable gate array (FPGA) platform and synthesis results are presented and discussed.

Real time video image enhancement approach using particle swarm optimisation technique with adaptive cumulative distribution function based histogram equalization

This paper describes an ongoing project that has the aim to develop a low cost application to replace a computer mouse for people with physical impairment. The application is based on an eye tracking algorithm and assumes that the camera and the head position are fixed. Color tracking and template matching methods are used for pupil detection. Calibration is provided by neural networks as well as by parametric interpolation methods. Neural networks use back-propagation for learning and bipolar sigmoid function is chosen as the activation function. The user's eye is scanned with a simple web camera with backlight compensation which is attached to a head fixation device. Neural networks significantly outperform parametric interpolation techniques: 1) the calibration procedure is faster as they require less calibration marks and 2) cursor control is more precise. The system in its current stage of development is able to distinguish regions at least on the level of desktop icons. The main limitation of the proposed method is the lack of head-pose invariance and its relative sensitivity to illumination (especially to incidental pupil reflections).