The remote sensing scene images classification has been of great value to civil and military fields. Deep learning models, especially the convolutional neural network (CNN), have achieved success in this task, however, they may suffer from two challenges: firstly, sizes category objects are usually different, but conventional CNN extracts features with fixed convolution extractor which could cause failure multi-scale features; secondly, some image regions not be useful during feature...

To improve the intra lossless video coding efficiency by introducing frequency domain compression, this paper presents a new scheme for mode in HEVC. The proposed method utilizes reversible lifting DWT and DC level shifting to transform one frame into four sub-bands of coefficients original tools process. recover transformed signed type loss-lessly, modified reconstruction are applied By implementing HM-16.5 reference software, experimental results show that compared with HEVC coding, can...

Rain streaks blur and degrade the color fidelity of images. This affects road segmentation accuracy in advanced driver assistance systems. To address this problem, a method combining gradient property with rain–fog model is proposed to remove rain single The detection used detect streaks, which aids training for removal determines if exist based on trained patches highest proportion high-frequency layer low-cost computation. In order recover nonrain images without oversmoothing, prior...

Accurate road segmentation is imperative for any effective advanced driver assistance system. Existing state-of-the-art methods are generally based on network architectures; however, high accuracy comes at the expense of abundant annotated data and high-computational cost. The inherently large amount nonroad areas increases running time affects accuracy. We propose a method enhancement vanishing points (VPs) deep learning to segment areas. VP serves as location prior remove redundant (e.g.,...

High-efficiency video coding (HEVC) is a very common application in today’s electronics field. Lossless intraframe compression preferred for many specific applications. Compared to lossy intracompression or interframe encoding, encoding has lower computation complexity but an unsatisfactory compressed bit-rate. Extant optimization algorithms center on improving coded bits space domains and few include considerations accelerating time. We focus the frequency domain reduce bit-rate while...

Fog degrades the quality of road images causing errors in stereo matching and segmentation for Advanced Driver Assistance Systems (ADAS). Accident rates can be reduced if robust efficient algorithms are applied image fog removal. Many studies have been conducted on this subject to date, but existing methods not optimized images. Several daytime models cause local darkness, blurring artifacts result low haze-free images, nighttime altogether inefficient. This study focuses dehazing by...

Objective To evaluate the behaviors and eye movement data of crew in different interface layouts, so as to provide a reference for design full ocean depth manned submersible. Methods To simulate visual stipulations 7 areas submersible by E-prime software, then record analyze correct rates reaction time subjects, continuously their data, i. e. dwell time, intake frequency, average intake, pupil diameter, saccade amplitude, acceleration, speed saccade. Results The were ranked...

A multi-level 2D-IDWT's architecture with high performance and memory efficiency is proposed. The proposed composed of two 1D-IDWT cores one single rearrangement register. Both the take only multiplier delay in their critical path at throughput rate one-input/one-output. register, rather than traditional 1.5N transposing memory, used to complete task. pipeline 2D-IDWT takes path. Moreover,the reduces 27% on-chip memory(only need 2N for 53-IDWT 4N 97-IDWT) by utilizing alternate scanning...

In this paper, we propose a methodology for computing the square root of complex number based on piecewise linear (PWL) approximation method. The proposed method relies software-based segmentor that automatically divides three real functions used in computation into fewest segments with predefined fractional bit width and maximum absolute error (MAE). coefficients, including start point, end slope y-intercept each segment, are stored use implementation hardware design. fully pipelined...

With the increasing size of video sequences, dramatically increased access bandwidth and power consumption for fetching storing reference frames with external memory during motion estimation become bottleneck realizing a practical codec system. To reduce no quality loss frames, lossless frame compression method based on pixelwise adaptive prediction is proposed. The algorithm allocates pixels from first line column horizontal vertical separately. Meanwhile, two direction coefficients are...