Histogram Building Research Articles

Sequential Model for Digital Image Contrast Enhancement

Aims: The manuscript aims at designing and developing a model for optimum contrast enhancement of an input image. The output image of model ensures the minimum noise, the maximum brightness and the maximum entropy preservation. Objectives: * To determine an optimal value of threshold by using the concept of entropy maximization for segmentation of all types of low contrast images. * To minimize the problem of over enhancement by using a combination of weighted distribution and weighted constrained model before applying histogram equalization process. * To provide an optimum contrast enhancement with minimum noise and undesirable visual artefacts. * To preserve the maximum entropy during the contrast enhancement process and providing detailed information recorded in an image. * To provide the maximum mean brightness preservation with better PSNR and contrast. * To effectively retain the natural appearance of an images. * To avoid all unnatural changes that occur in Cumulative Density Function. * To minimize the problems such as noise, blurring and intensity saturation artefacts. Methods: 1. Histogram Building. 2. Segmentation using Shannon’s Entropy Maximization. 3. Weighted Normalized Constrained Model. 4. Histogram Equalization. 5. Adaptive Gamma Correction Process. 6. Homomorphic Filtering. Results: Experimental results obtained by applying the proposed technique MEWCHE-AGC on the dataset of low contrast images, prove that MEWCHE-AGC preserves the maximum brightness, yields the maximum entropy, high value of PSNR and high contrast. This technique is also effective in retaining the natural appearance of an images. The comparative analysis of MEWCHE-AGC with existing techniques of contrast enhancement is an evidence for its better performance in both qualitative as well as quantitative aspects. Conclusion: The technique MEWCHE-AGC is suitable for enhancement of digital images with varying contrasts. Thus useful for extracting the detailed and precise information from an input image. Thus becomes useful in identification of a desired regions in an image.

Exploiting GPUs for Efficient Gradient Boosting Decision Tree Training

In this paper, we present a novel parallel implementation for training Gradient Boosting Decision Trees (GBDTs) on Graphics Processing Units (GPUs). Thanks to the excellent results on classification/regression and the open sourced libraries such as XGBoost, GBDTs have become very popular in recent years and won many awards in machine learning and data mining competitions. Although GPUs have demonstrated their success in accelerating many machine learning applications, it is challenging to develop an efficient GPU-based GBDT algorithm. The key challenges include irregular memory accesses, many sorting operations with small inputs and varying data parallel granularities in tree construction. To tackle these challenges on GPUs, we propose various novel techniques including (i) Run-length Encoding compression and thread/block workload dynamic allocation, (ii) data partitioning based on stable sort, and fast and memory efficient attribute ID lookup in node splitting, (iii) finding approximate split points using two-stage histogram building, (iv) building histograms with the aware of sparsity and exploiting histogram subtraction to reduce histogram building workload, (v) reusing intermediate training results for efficient gradient computation, and (vi) exploiting multiple GPUs to handle larger data sets efficiently. Our experimental results show that our algorithm named ThunderGBM can be 10x times faster than the state-of-the-art libraries (i.e., XGBoost, LightGBM and CatBoost) running on a relatively high-end workstation of 20 CPU cores. In comparison with the libraries on GPUs, ThunderGBM can handle higher dimensional problems which the libraries become extremely slow or simply fail. For the data sets the existing libraries on GPUs can handle, ThunderGBM achieves up to 10 times speedup on the same hardware, which demonstrates the significance of our GPU optimizations. Moreover, the models trained by ThunderGBM are identical to those trained by XGBoost, and have similar quality as those trained by LightGBM and CatBoost.

A Ground Elevation and Vegetation Height Retrieval Algorithm Using Micro-Pulse Photon-Counting Lidar Data

The Ice, Cloud and Land Elevation Satellite-2 (ICESat-2) mission employs a micro-pulse photon-counting LiDAR system for mapping and monitoring the biomass and carbon of terrestrial ecosystems over large areas. In preparation for ICESat-2 data processing and applications, this paper aimed to develop and validate an effective algorithm for better estimating ground elevation and vegetation height from photon-counting LiDAR data. Our new proposed algorithm consists of three key steps. Firstly, the noise photons were filtered out using a noise removal algorithm based on localized statistical analysis. Secondly, we classified the signal photons into canopy photons and ground photons by conducting a series of operations, including elevation frequency histogram building, empirical mode decomposition (EMD), and progressive densification. At the same time, we also identified the top of canopy (TOC) photons from canopy photons by percentile statistics method. Thereafter, the ground and TOC surfaces were generated from ground photons and TOC photons by cubic spline interpolation, respectively. Finally, the ground elevation and vegetation height were estimated by retrieved ground and TOC surfaces. The results indicate that the noise removal algorithm is effective in identifying background noise and preserving signal photons. The retrieved ground elevation is more accurate than the retrieved vegetation height, and the results of nighttime data are better than those of the corresponding daytime data. Specifically, the root-mean-square error (RMSE) values of ground elevation estimates range from 2.25 to 6.45 m for daytime data and 2.03 to 6.03 m for nighttime data. The RMSE values of vegetation height estimates range from 4.63 to 8.92 m for daytime data and 4.55 to 8.65 m for nighttime data. Our algorithm performs better than the previous algorithms in estimating ground elevation and vegetation height due to lower RMSE values. Additionally, the results also illuminate that the photon classification algorithm effectively reduces the negative effects of slope and vegetation coverage. Overall, our paper provides an effective solution for estimating ground elevation and vegetation height from micro-pulse photon-counting LiDAR data.

Real-Time Inter-Frame Histogram Builder for SPAD Image Sensors

CMOS image sensors based on single-photon avalanche-diodes (SPAD) are suitable for 2D and 3D vision. Limited by uncorrelated noise and/or low illumination conditions, image capturing becomes nearly impossible in a single-shot exposure time. Moreover, the depth accuracy is affected by jitter. Therefore, many frames need to be taken to reconstruct the final accurate image. The proposed reconstruction algorithm is based on pixel-wise histogram building. Specifically, a histogram is built on the fly for each pixel of the array from the ongoing acquired frames. This paper presents the design and implementation on FPGA of a real-time pixel-wise inter-frame histogram builder at 1 kfps. The design has been proven with a 64 × 64-pixels SPAD camera. Its remarkable robustness has been demonstrated in harsh conditions, such as 42 kHz of dark count rate (DCR) and high background illumination up to 20 times larger than the DCR. The system has a graphic user interface for 2D/3D imager configuration, image streaming, and pixel-wise histogram streaming.

Fast and compact data acquisition for gas-filled detectors with delay line

This article describes the functionality, implementation and performance of a PCI data acquisition board that can be used in conjunction with gas-filled detectors with delay line readout. The board combines a large on-board 256-MByte histogramming memory with a maximum 10-MHz count rate in continuous operation and integrates the time frame generation, histogram building and buffering functionalities in a single PCI board, resulting in a fast, compact and cost-effective data acquisition solution for the Spanish beamline BM16 at ESRF.

Ion channels—analysis and simulation embedded in a flexible environment

A collection of programs for ion-channel analysis is described. The package contains all standard routines for single-channel analysis and some routines for fluctuation analysis. Additionally, advanced algorithms for histogram building and filter correction are incorporated. The major advantage of this package is its embedding in the MATLAB TM environment, a general purpose program for huge numerical computations. Together with the ion-channel simulation facility this package forms the basis for optimizing analysis methods and readily incorporating new algorithms found in the literature.